Thesis Reference - Archive ouverte UNIGE

Transcription

Thesis
Influence of sleep-wake states on human memory and underlying
neural plasticity: insights from EEG recordings and parasomnia
CONSTANTINESCU, Irina Oana
Abstract
L'objectif de cette thèse est d'intégrer plusieurs niveaux d’observation (comportement,
réseaux neuronaux, potentiels neuraux locaux) de la plasticité du cerveau humain lie à
l’apprentissage et a la formation de la mémoire. Nous avons évalué la réorganisation des
représentations neuronales post-apprentissage au cours du sommeil et de l'éveil. Nous
apportons la preuve, pour la première fois, d’une réactivation comportementale pendant le
sommeil chez les humains, en étudiant des patients présentant des épisodes d’activité
motrice pendant le sommeil. Nous avons aussi montré chez l'homme, qu'une stimulation
faible et répétée de balancement, modifie l'architecture du sommeil et rends le sommeil plus
stable. Les résultats de cette thèse de doctorat ouvre de nouvelles perspectives de
recherches pluridisciplinaires sur le sommeil et la mémoire.
Reference
CONSTANTINESCU, Irina Oana. Influence of sleep-wake states on human memory and
underlying neural plasticity: insights from EEG recordings and parasomnia. Thèse de
doctorat : Univ. Genève et Lausanne, 2011, no. Neur. 60
URN : urn:nbn:ch:unige-161683
Available at:
http://archive-ouverte.unige.ch/unige:16168
Disclaimer: layout of this document may differ from the published version.
[ Downloaded 15/10/2016 at 06:52:48 ]
DOCTORAT EN NEUROSCIENCES
des Universités de Genève
et de Lausanne
UNIVERSITÉ DE GENÈVE
FACULTÉ DE MÉDECINE
Docteure Sophie Schwartz, directrice de thèse
Professeure Margitta Seeck, co-directrice de thèse
TITRE DE LA THÈSE
INFLUENCE OF SLEEP-WAKE STATES ON HUMAN MEMORY
AND UNDERLYING NEURAL PLASTICITY:
INSIGHTS FROM EEG RECORDINGS AND PARASOMNIA.
THÈSE
Présentée à la
Faculté de Médecine
de l‟Université de Genève
pour obtenir le grade de
Docteure en Neurosciences
par
Irina CONSTANTINESCU
de Roumanie
Thèse N° 60
Genève
Editeur ou imprimeur : Université de Genève
2011
To my family, my friends and my mentors
Eu nu strivesc corola de minuni a lumii
şi nu ucid
cu mintea tainele, ce le-ntâlnesc
în calea mea
în flori, în ochi, pe buze ori morminte.
I do not crush the aura of world's wonders
and do not obliterate
within my mind the unrevealed
crossed along my path,
be it on flowers, eyes, lips, or graves.
*Lucian Blaga (1895 – 1961)*
Romanian philosopher, poet, playwright, translator, journalist, university professor and
diplomat. Impressive and multidimensional personality of the European culture,
Blaga also approached the philosophical problematic of science.
*
2
The dream (1937) by Salvador Dali
3
Remerciements
Mes remerciements vont tout d‟abord a Sophie Schwartz, de m‟avoir accueillie il
y a 5 ans dans son groupe en m‟offrant le privilège de découvrir l‟univers passionnant
de la recherche sur le sommeil. Un grand merci pour l‟inépuisable enthousiasme,
générateur des idées innovatrices, pour son support permanent et sa confiance en moi
et en nos projets, pour la grande ouverture d‟esprit, pour la créativité, l‟humanisme,
pour l‟écoute et les conseils précieux qui m‟ont aide a évolué pas seulement dans le
plan scientifique, mais aussi personnel. Un grand merci pour tout, Sophie ! Et merci la
petite Elsa pour son doux sourire.
Je tiens à remercier vivement Margitta Seeck, qui m‟a offert la possibilité
d‟accueillir des précieuses données des patients épileptiques implantes dans l‟unité
d‟évaluation prechirurgicale de l‟épilepsie, et qui a accepte d‟être co-conductrice de
cette thèse.
Je remercie chaleureusement Isabelle Arnulf, qui dirige
le laboratoire du
sommeil, à l‟hôpital Pitié-Salpêtrière à Paris, qui m‟a offert l‟occasion de m‟impliquer
dans un projet passionnant sur les patients parasomniaques. Merci pour
l‟enthousiasme communicatif et pour son support. Merci aussi a Delphine Oudiette,
pour les discussions très constructives sur le sommeil et la cognition, et pas seulement,
mais également pour sa bonne humeur et sa promptitude.
Mes remerciements s‟adressent en particulier à Marzia De Lucia pour son aide
précieuse sur les données intracrâniennes, sa rigueur scientifique et sa grande patience
avec moi.
Je remercie également Laurence Bayer pour sa collaboration fructueuse sur le
projet « hamac », son calme, son écoute, et sa confiance dans notre projet.
Merci a Virginie Sterpenich pour son écoute, sa désarmante perspicacité et ses
conseils éclairés sur la dernière ligne droite de la thèse.
Merci a Gilles Pourtois, pour sa vivacité, son enthousiasme contagieux, et ses
conseils avisés sur les données EEG.
Je tiens à remercier Michel Mühlethaler pour ses conseils et son soutien.
4
Un grand merci a Patrik Vuilleumier, pour son son accueil parmi les « Labnics »,
son support constant, ses conseils précieux a travers ces années.
Mes remerciements s‟adressent à Stephen Perrig pour son enthousiasme, et les
échanges constructifs sur le sommeil pendant la durée de la thèse.
Je tiens également à remercier chaleureusement Philippe Peigneux et Dimitri
Van De Ville d‟avoir accepte d‟être rapporteurs de cette thèse et suis heureuse de
pouvoir soumettre ce travail a leur critique.
Je remercie également Pierre Maquet pour ses conseils avisés.
Mes remerciements vont aussi vers Laurent Spinelli et Denis Brunet pour leur
aide précieux concernant les données EEG.
Je remercie vivement mes collègues, Karsten, Stéphanie, Amal, Camille, Karim,
Hamdi, Wiebke, Agustina, Tonia, Julie pour leur support, leur gentillesse et la bonne
humeur a travers ces années. Je n‟oublie pas l‟équipe du laboratoire du sommeil de
Belle-Idée pour leur support et leur accueil.
Je tiens à remercier le Professeur Landis pour son grand support et pour ses
encouragements a poursuivre les travaux de neurosciences.
Je remercie Martine Collart pour son soutien avisé.
Je remercie également le Professeur Lücking, Université de Freiburg,
Allemagne, pour son soutien et sa confiance en moi à travers mes années d‟étude.
Merci a mes mentors de l‟Université de Iasi en Roumanie pour leur chaleureux
support et l‟ouverture d‟esprit.
Un grand merci a mes amis qui ont su m‟écouter et me soutenir pendant ces
années, qui ont été près de moi et ont eu confiance en moi.
Je tiens enfin à adresser toute ma reconnaissance et mon affection à ma chère
famille et a Xavier pour leur énorme soutien inconditionné, leur attention, leur
patience et leur confiance en mes rêves.
5
Resumé
L'objectif du présent travail est d'étudier l'influence des états de veille-sommeil
sur la mémoire humaine et les mécanismes sous-jacents de la plasticité neuronale en
utilisant des enregistrements electro-encephalographiques (EEG) et des modèles
neurologiques comme la parasomnie.
Mieux comprendre comment les états du sommeil et de l'éveil interfèrent avec
des processus de la mémoire pourrait également révéler des indices nouveaux et
fondamentaux sur le fonctionnement du cerveau humain. Malgré un grand nombre
d'études, il ya encore des questions ouvertes sur ce sujet, qui a motivé le présent
travail. Nous avons observé des processus physiologiques et cognitifs liés à la
consolidation des connaissances récemment acquises dans différents états de
vigilance, en appliquant un schéma convergente des études expérimentales. L'objectif
était d'intégrer plusieurs niveaux d‟observation (comportement, réseaux neuronaux,
potentiels neuraux locaux) de la plasticité du cerveau humain. Nous avons utilisé le
concept de réactivations neuronales liées à l‟expérience récemment acquise pour
étudier la dynamique du cerveau sous-tendant la mémoire. Nous avons évalué la
réorganisation des représentations neuronales au cours du sommeil post-formation et
de l'éveil, qu‟elle soit produite spontanément ou induite par la présentation des stimuli
externes, associes. Plus précisément, nous avons étudié la dynamique neuronale liées à
l'expérience acquise chez des patients épileptiques pharmaco-résistants, pendant la
procédure d'apprentissage des séquences de mouvements. Nous avons également
étudié la réactivation induite et la réorganisation des traces mnésiques pendant l'éveil
post-apprentissage en utilisant des enregistrements EEG du scalp, de haute densité
chez l'homme. La réactivation au niveau du comportement des événements récemment
acquis a été évaluée pendant le sommeil post-apprentissage chez des patients
parasomniaques, présentant des épisodes d‟activité motrice pendant le sommeil. Nous
apportons la preuve, pour la première fois à notre connaissance, d‟une réactivation
comportementale pendant le sommeil chez l‟homme. Sur le plan méthodologique, la
présente thèse combine de manière innovante les différentes techniques EEG
6
(enregistrements intracrâniennes, des enregistrements de haute densité au niveau du
scalp, des enregistrements des rythmes de sommeil) qui fournissent une fenêtre
unique, directe et précise, sur le fonctionnement cérébral de l'homme. En outre, nous
avons donné à des techniques déjà existantes de traitement du signal de nouvelles
applications; parmi les techniques appliquées, la présente thèse met en évidence la
valeur unique de l'EEG intracrânien pour la cartographie du cerveau humain.
Une des façons les plus courantes et archaïques pour faciliter le sommeil est le
balancement : nous nous endormons irrésistiblement dans une balançoire et, depuis
des temps immémoriaux, le bercement aide l‟endormissement des bébés. Pourtant,
aucune explication claire physiologique à cet effet n‟a pas encore été fournie. Dans
notre étude, nous avons montré chez l'homme, en utilisant un lit-balançoire pendant
les siestes l'après-midi, qu'une stimulation faible et répétée de balancement, modifie
l'architecture du sommeil et a un effet bénéfique sur le sommeil en rendant le sommeil
plus stable. Notre travail montre que le sommeil peut être amélioré d‟une manière
instrumentale et motive le développement de nouveaux dispositifs pour aider à dormir.
Les résultats de cette thèse de doctorat contribuent à une meilleure
compréhension des réorganisations de l'activité cérébrale liés à l'apprentissage et
ouvre de nouvelles perspectives de recherches pluridisciplinaires sur le sommeil et la
mémoire.
7
Abstract
The aim of the present work was to study the influence of sleep-wake states on
human memory and underlying neural plasticity by using EEG recordings and
neurological models as parasomnia.
Understanding how sleep and wakefulness states impact memory processes may
reveal new and fundamental cues on human brain functioning. Despite a large number
of studies, there are still opened questions on this topic, which motivated the present
work.
We observed physiological and cognitive processes related to the consolidation
of recently acquired knowledge across different vigilance states, by applying a
convergent diagram of studies. The aim was to integrate multilevel (behavior, brain
networks, local field potentials) views of plasticity in the human brain. We used the
concept of experience-related neural reactivations to study brain dynamics subtending
memory. We assessed re-shaping of neural representations during post-training sleep
and wakefulness, either produced spontaneously or induced by presentation of
learning-related external cues. More precisely, we studied experience-related neural
dynamics in pharmaco-resistant epileptic patients while procedurally learning
sequences of movements. We also studied induced reactivation and reorganization of
neural traces during post-training wakefulness at scalp level by using high-density
recordings in humans. Experience-related behavioral re-enactment during posttraining sleep was assessed in parasomnia patients. We bring evidence, for the first
time to our knowledge, of a learning-related behavioral replay in humans.
Methodologically, the present thesis combines in an innovative manner different EEG
techniques (intracranial recordings, high-density scalp recordings, sleep rythms
recordings) which provide a particularly direct and accurate window onto human brain
function (Michel 2009). Furthermore, we conferred to already existent signal
processing techniques new applications: in project 1, a multivariate decoding
algorithm captured learning-dependent changes in intracranial EEG signal at the
single trial level. Among the techniques applied, the present thesis highlights the
unique value of the intracranial EEG for human brain mapping.
8
One of the most common and archaic way to facilitate sleep is rocking: we
irresistibly fall asleep in a rocking-chair and, since immemorial times, we cradle our
babies to sleep. Yet, no clear physiological explanation has been provided to this
observed effect. In our study, we showed by using using a swinging-bed during
afternoon naps that a low and repeated stimulation mimicking rocking modifies sleep
architecture in humans and has a beneficial effect on sleep by rendering sleep more
stable. Our work shows that sleep may be improved instrumentally and motivates the
development of new devices to help sleep.
The present results may contribute at further understanding of learning-related
reorganization of brain activity and opens new perspectives of multidisciplinary
research.
9
INDEX
1. INTRODUCTION ........................................................................ 8
2. LEARNING AND MEMORY SYSTEMS ........................................... 9
2.1. Memory systems: theoretical and experimental approaches ..... 10
2.2. Perceptual and motor skill learning ........................................... 14
Perceptual learning ................................................................................. 14
Motor learning ......................................................................................... 15
Attentional influences on learning ........................................................... 16
2.3. The temporal dynamics of perceptual and motor skill learning.. 17
3. ANATOMO-FUNCTIONAL CORRELATES OF MEMORY ................ 19
3.1. Memory and brain plasticity ...................................................... 19
3.2. Brain systems for memory processes ........................................ 21
Hippocampus and related structures ....................................................... 21
Implications of the hippocampus in memory functions ........................... 22
Amygdala and emotional memory ........................................................... 24
Diencephalon and declarative memory deficits ....................................... 24
Motor areas, basal ganglia, cerebellum and memory for motor skills ...... 24
Prefrontal cortex and memory consolidation ........................................... 25
3.3. Distinct steps in memory consolidation ..................................... 25
4. THE INFLUENCE OF SLEEP-WAKE STATES
ON MEMORY PROCESSES ....................................................... 28
4.1. Sleep functions: a beneficial role for memory consolidation? .... 28
Aspects of sleep physiology..................................................................... 28
Sleep functions ........................................................................................ 30
Napping in humans ................................................................................. 31
4.2. Sleep and wakefulness:
recent hypotheses on memory consolidation ............................. 32
4.2.1. Neural reactivations during sleep .................................................. 32
Reactivations at the cellular and neuronal networks levels ..................... 32
Reactivations at the brain areas level ..................................................... 36
4.2.2. Neural reactivations during wakefulness....................................... 38
4.3. Sleep homeostasis hypothesis ................................................... 39
5. RATIONALE OF THE PRESENT WORK: MAIN QUESTION
AND METHODOLOGICAL APPROACHES .................................... 40
Experimental part
OVERVIEW .................................................................................. 42
THESIS EXPERIMENTAL DESIGN ................................................. 42
Experiment 1
Neurophysiological evidence from human intracranial
recordingsof sequenced knowledge consolidation
CONTEXT .................................................................................... 43
METHODOLOGICAL HIGHLIGHTS ................................................ 43
SUMMARY OF RESULTS ............................................................... 44
CONCLUSION .............................................................................. 44
10
Article submitted to NeuroImage ............................................... 45
ABSTRACT ........................................................................................ 46
1. INTRODUCTION ............................................................................ 46
2. MATERIAL AND METHODS............................................................. 48
2.1. Patients ..................................................................................... 48
2.2. Behavioral task and experimental procedure ............................ 49
2.3. Intracranial EEG recording ........................................................ 50
2.4. Data analysis ............................................................................. 51
2.4.1. Multivariate decoding approach to iEEG ........................................ 51
Hidden Markov Model of single-trial iEEG ................................................ 52
2.4.2. Accuracy estimation ...................................................................... 54
2.4.3. Estimating contacts with higher classification power .................... 55
2.4.4. Event-related potentials ................................................................ 55
3. RESULTS ....................................................................................... 55
3.1. Behavioral results...................................................................... 57
3.2. Results from the multivariate decoding
approach to single-trial iEEG...................................................... 57
3.2.1. Accuracy estimation ...................................................................... 57
3.2.2. Estimating contacts with higher classification power .................... 58
3.3.3. Relation between classification accuracy
and spatio-temporal pattern of single-trial events ......................... 59
3.3. Intracranial event-related potentials......................................... 60
4. DISCUSSION................................................................................. 61
4.1. Learning-related changes in the stability
of momentary neural states ....................................................... 62
4.2. Role of the hippocampus in sequence learning .......................... 63
4.3. Multivariate decoding approach to single-trial iEEG data ........... 64
5. CONCLUSIONS .............................................................................. 66
Acknowledgments ............................................................................ 66
Reference list ................................................................................... 66
Figures ............................................................................................. 71
Experiment 2
Evidence of overt replay of a recently learned
motor sequence during human sleep
CONTEXT .................................................................................... 75
CONCLUSION .............................................................................. 75
Article submitted to PLoS Biology ............................................... 76
ABSTRACT ........................................................................................ 76
INTRODUCTION ................................................................................ 77
METHODS ......................................................................................... 78
Ethic Statement ................................................................................ 78
Subjects ........................................................................................... 78
Behavioral task and experimental procedure ................................... 79
Sleep and nocturnal behavior monitoring ......................................... 81
Statistical analyses of motor performance ....................................... 81
Assessment of sequence replay during sleep ................................... 82
RESULTS ........................................................................................... 82
Sleep and cognitive performances .................................................... 82
11
Dreams content after the post-training night of sleep ...................... 83
Behaviors exhibited during post-training sleep ................................ 83
RBD patients ........................................................................................... 83
Sleepwalkers .................................................................................... 84
Evidence for behavioral replay during sleep ..................................... 84
DISCUSSION..................................................................................... 85
Acknowledgements .......................................................................... 87
References ....................................................................................... 87
Figures ............................................................................................. 89
SUPPLEMENTAL DATA ....................................................................... 92
Questions to judges .......................................................................... 92
Experiment 3
Experience-dependent induced reactivations
during post-training wakefulness
CONTEXT .................................................................................... 94
CONCLUSIONS ............................................................................ 95
Article in preparation.................................................................. 95
INTRODUCTION ................................................................................ 95
EXPERIMENTAL PROCEDURE ............................................................ 96
Subjects ........................................................................................... 96
Stimuli and task ............................................................................... 97
Behavioral task and experimental procedure ................................... 97
Electrophysiological data acquisition.............................................. 100
Event-related potential analysis ..................................................... 101
Segmentation analyses .................................................................. 101
RESULTS ......................................................................................... 102
Behavioral results .......................................................................... 102
ELECTROPHYSIOLOGICAL RESULTS ................................................ 103
ERPs ............................................................................................... 103
Perception condition....................................................................... 103
Mental imagery condition ............................................................... 104
Scalp topography analysis .............................................................. 105
DISCUSSION................................................................................... 108
SUPPLEMENTARY DATA .................................................................. 109
Instructions .................................................................................... 109
Eyes Opened ................................................................................... 110
Eyes Closed .................................................................................... 111
References ..................................................................................... 112
Experiment 4
Instrumental modulation of electrophysiological
features of sleep
CONTEXT .................................................................................. 114
METHODOLOGICAL HIGHLIGHTS .............................................. 114
SUMMARY OF RESULTS ............................................................. 114
CONCLUSION ............................................................................ 115
12
Article submitted in Current Biology ......................................... 116
SUPPLEMENTAL DATA ..................................................................... 120
SUPPLEMENTAL EXPERIMENTAL PROCEDURES ............................... 120
Participants .................................................................................... 120
Protocol .......................................................................................... 121
EEG data analyses .......................................................................... 121
References and Notes ..................................................................... 122
Supplemental References ............................................................... 123
DISCUSSION ............................................................................. 124
MULTIMODAL APPROACH TO MEMORY CONSOLIDATION.......... 124
NOVELTY OF THE TECHNIQUES ................................................. 126
CONCLUSIONS AND PERSPECTIVES .......................................... 127
BIBLIOGRAPHY ........................................................................ 129
13
1. INTRODUCTION
The aim of this preamble is to provide a rationale for the present research work.
It offers an overview of the cognitive and neural processes underlying learning and
memory, as well as the influence of sleep on these processes. This introduction
therefore highlights a selection of key issues which support the theoretical questions
and the methodological approaches of the thesis.
Neuroscience has provided lately some astonishing breakthroughs in our
understanding of sleep and memory (Kandel 2009), from noninvasive large-scale
imaging of the human brain during sleep to revealing the molecular machinery of
complex processes like cerebral plasticity. This evolving knowledge about brain
functioning together with the recent development of investigation techniques foster
translations between distinct levels of description, such as the molecular/cellular level,
the macroscopic systems level, and the behavioral level. The main motivation for the
present work comes from these new possibilities to study the interaction between
sleep and memory both at the neural and behavioral levels.
The introduction begins with a conceptual framework for learning and memory.
In the second part, some key aspects of memory consolidation and supporting
plasticity mechanisms are introduced, with a special emphasis on interacting neural
circuits. In the third part of the introduction, recent contributions on the role of sleep
and wakefulness in memory formation are reviewed. Some aspects concerning local
brain activity and regulatory processes during sleep are also detailed, as well as their
possible functional links with memory and plasticity processes. To conclude the
introduction, the research questions and hypotheses investigated in this thesis are
briefly presented.
After the introduction, the experimental part reports in detail each of the four
projects conducted in the context of this thesis: the hypotheses tested, the
methodological approaches used, the main findings and their interpretation. A general
discussion concludes the thesis by providing an integrated overview of all the results
and opening some new perspectives for further research.
14
2. LEARNING AND MEMORY SYSTEMS
Learning is the process of acquiring new information, while memory
refers to the persistence of learning in a state that can be revealed at a
later time. Memory is the usual consequence of learning.
Squire, Memory and Brain, 1987, p.3
Almost every aspect of our thinking and behavior depends on the acquisition and
retention of new knowledge: our sense of identity, how we perceive the world, how
we react to various stimuli, perform simple or complex activities, such as eat or drive,
organize our daily agenda. Thus, learning and memory play a fundamental role in
cognitive functioning and behavioral adjustment.
Learning most often refers to the gradual acquisition of the information, while
memory refers more to specific processes manipulating the established information,
which may impact further behavior. Remembering – the direct effect of memoryenables one to compare the newly learned information with the already existent
memory content and decide the relevance of the information for the system (thus
maintaining the system‟s efficiency). Therefore, although memory holds information
about the past, its main function is to allow predictions and adaptive responses to
events yet to come (Schacter and Addis 2007; Schacter, Addis et al. 2008).
La mémoire du passé n'est pas faite pour se souvenir du passé, elle est
faite pour prévenir le futur. La mémoire est un instrument de prédiction.
Alain Berthoz, www.automatesintelligents.com/interviews/2003/octobre/berthoz.html
Because many elements in the future are produced according to probabilistic
laws, the machinery that encodes memories in the brain has to operate fast and
recurrent updating of the stored information. This built-in capacity to anticipate
change is reflected at multiple brain levels (Addis, Wong et al. 2007; Bar 2009). As
stated by Bar et al (2009), memory-related brain dynamics actually reflect ongoing
generation of predictions, which relies on acquired experience and aim at providing
permanent and adequate adjustment to our interaction with the changing environment.
15
2.1. Memory systems:
theoretical and experimental approaches
The topic of memory is broad and accumulating. Much theoretical and
experimental effort has been allocated in the last years in the attempt to clearly define
and characterize the neural systems that are recruited during the acquisition of
different tasks and subsequent memory processes (Squire 2004).
Theoretically, memory is not currently viewed and studied as a unitary process.
The concept of memory system was first promoted by Tulving et al. (1985), who
defined it as “a set of correlated processes, more closely related to one another than
they are to processes outside the system” (p.386).
More precisely, memory systems can be characterized according to the kind of
processed information and to the contributing brain mechanisms (Schacter and
Tulving, 1994). According to these aspects, early work on memory postulated two
distinct systems. On one hand, a system that deals with explicitly acquired
information, generically termed declarative memory. Declarative memory, usually
defined as the memory for facts and events (“know what”), can be further divided into
episodic memory, reflecting autobiographical events, and semantic memory, dealing
with general knowledge about the world. The other classically accepted memory
system operates with implicitly acquired information, in the absence of conscious
awareness and is termed procedural memory (Eustache and Desgranges 2008).
Procedural memory (“know how”) is mainly based on the acquisition of perceptual
and motor skills.
Another way to look at memory processes is by considering the time axis.
Therefore, R. Atkinson and R. Shiffrin proposed in the late 60ies a “multi-store”
model, which assumed that human memory is formed through three consequent
stages: sensory memory, short-term memory and long-term memory (Atkinson 1968).
According to the authors, sensory memory corresponds approximately to the initial
200-500 milliseconds after a stimulus is perceived. Short-term memory refers to
memory for information currently “held” in mind and has limited capacity. It allows
recall for a period of several seconds to a minute without rehearsal, allowing the
temporarily saving of a limited quantity of information. Long-term memory refers to
information that is stored for an extended period of time. It has a potentially unlimited
duration and capacity. In the present thesis we do not deal with short-term memory,
16
but with long-term memory aspects, which implies changes in widely distributed
neural connections throughout the brain. Among all areas, the hippocampus seems
particularly important for the transformation of information from short-term to longterm memory. In humans, studies in neurological patients with focal brain damage
remain critical in the attempt to define which brain regions are contributing to certain
memory processes. Furthermore, as lesion studies address disruption instead of
engagement of memory systems, they can reveal structures that are indispensably
required for a certain function. Maybe the first influential observation which turned
into a landmark for the development of the neuroscience of memory systems, was the
case of the patient H.M. (Squire 2009). In 1953, an experimental brain surgery
intended to control severe seizure disorder, left patient H.M. unable to form new
memories. A finger-sized piece of the temporal lobes on both sides of the brain,
including most of the hippocampus, amygdala, and nearby parahippocampal gyrus,
was removed. Patient H.M. could remember facts he had learned and names of people
he had met before the surgery, but virtually nothing after it. For half a century, the
findings from testing the patient H.M. highlighted memory as a distinct cerebral
function, accessible for evaluation, independently from other perceptual and cognitive
capacities. Since the case of patient H.M., several theories suggested that memory
involves several successive steps leading to stable memory. In particular, it was
proposed that new memories formed by the hippocampus are later transferred in the
cerebral cortex for long-term storage (Scoville 1957; Scoville 2000). A large body of
studies deals with long-term memory in the attempt to understand the basis of
information storage and use over time. Studies in amnestic patients showed that
damage limited to the hippocampal structure results in selective deficits in recollection
and relational memory.
Among the techniques used to approach cognitive and neurophysiological aspects
of memory, EEG studies represent a prominent neuroimaging tool. Different effects
related to memory processes, such as explicit/implicit, old/new, recognition/familiarity,
have been addressed by analyzing modulations in the electrical activity of the brain,
time-locked to an event (such as a stimulus to be encoded). Event-related brain
potentials (ERPs) have been employed in memory research to identify neural activity
associated with both encoding of information and the later retrieval of stored
information. Different correlates between amplitudes and/or latencies of event-related
17
brain potentials and cognitive tasks are quantified depending on the memory paradigm
applied (Voss 2008; Friedman 2000; Wieser 2003).
For example, it has been shown in terms of ERP activity, that implicit and
explicit learning of event sequences yield different neural representations in the brain
(Russeler 2000). In this EEG study, event-related brain potentials of 21 subjects were
recorded while they performed a choice reaction time task. In this task, a repetitive
sequence of eight successive stimuli (e.g. a sequence of letters) was presented on a
computer screen and subjects had to react as quickly and accurately as possible when
a stimulus appeared on the screen, by pressing a button, either with the index or
middle finger, of either left or right hand, according to the stimulus position on the
screen. There were therefore four finger press possibilities and eight “standard”
stimuli, which meant that the same finger coded for two stimuli. The regularity of the
sequence was unknown to the subjects. Within the regular event sequence, a
“perceptual” or a “motor” deviant stimulus replaced sometimes an expected stimulus:
a “perceptual” deviant was a different letter from the expected one in the sequence,
but which required the same motor response, while a “motor” deviant was a letter
from those requiring a response with the opposite hand. Post-experimental free recall
and recognition test, assigned subjects to either an explicit or implicit group,
according to the either explicit or implicit knowledge of the sequence regularity. The
ERPs showed different brain potential modulation for different types of stimuli
(perceptual and motor deviants as compared to expected stimuli) between the two
groups. In the group of explicit learners, a larger N200 component (negative
component peaking at 200 ms post-stimulus) was evoked by both types of deviants
(perceptual and motor) as compared to standard stimuli; a larger P300 (positive
component peaking at 300 ms post-stimulus) was evoked by “motor” deviants only as
compared to “perceptual” and “standard” stimuli. In the group of implicit learners, the
N200 and P300 components remained unaffected. In both groups of subjects the
lateralized readiness potential (LRP) which accompanied “motor” deviants revealed a
different modulation as compared “standard” and “perceptual” stimuli. This meant
that in both groups, the preparation for the next motor response dependent on the
previous one, therefore on response-response association. These results suggest that
while implicit learners acquire knowledge about response dependencies only, explicit
learners acquire knowledge about both response and stimulus dependencies.
18
Moreover, the study of EEG microstates of temporal stability in the brain
electrical activity opens promising perspectives concerning the further understanding
of the memory architecture (Lehmann 2010; Pascual-Marqui 1995; Van de Ville
2010). Functional microstates (Michel et al. 1992; Lehman 1987) refer to time
segments of stable potential map configuration supposed to reflect different steps of
information processing. According to the microstate model, the brain activity can be
seen as a sequence of non-overlapping microstates of variable duration and strength.
A growing number of studies evaluate functional microstates in the spontaneous EEG
and their influence by specific mental conditions, such as memory impairment in
Alzheimer disease. Dierks et al (1997) studied alterations of EEG microstates in
Alzheimer disease patients compared to healthy controls. The main findings were
shorter segment duration and increased number of segments in patients with memory
impairment compared to healthy controls.
Another complementary neurophysiological approach to memory related
dynamics is represented by the study of synchronous brain oscillations such as theta or
gamma activity during subsequent stages of memory formation (Axmacher 2006; Kirk
2003). While EEG provides a high temporal resolution of recordings, it does not allow
the unequivocal location of the neural generators responsible for the scalp-recorded
potentials. One option is to integrate the ERP findings with the results of functional
neuroimaging studies and thus to identify the intracerebral generators of the memory
effects. Direct intracranial recordings in animals or humans (in the context of
presurgical investigation), which provides a combination of precise spatial and
temporal information (Guo 2005; Fell 2008). For example, one intracranial study on
seven epileptic pharmaco-resistant patients (Seeck 1997), required precise
discrimination between repeated and non-repeated faces. The patients failed to show
explicit knowledge of previously seen faces, as measured by the accuracy of motor
responses. However, all subjects showed a differential modulation of the intracranial
potential to repeated versus non-repeated faces, thus suggesting implicit
discrimination between the two types of stimuli. Some considerations need to be taken
into account when testing pharmaco-resistant epileptic patients. Although the
intracranial recordings performed in these patients provide unique opportunities to
observe cognitive processes at a very high spatial and temporal resolution, the neural
responses collected may be subject to considerable individual differences (Halgren
19
1978). Epileptic patients represent a heterogenous population and the presence, the
severity and the localization of the epileptic disturbances may influence the
interpretation of the cognitive testing; maybe the most contributive variable for the
memory research in these patients is the extent to which the implanted tissue (most
frequently the temporal lobe) is affected. Also, convulsivant medication may have
significant effects on memory functions in these patients. The left and right-sided
focus of epileptic discharges has also been related to the presence of more or less
important verbal /non verbal memory deficits (Kapur 1994). The important advances
in signal processing methods allow a more and more accurate distinction between
pathological and physiological data from testing these patients.
2.2. Perceptual and motor skill learning
Skill learning represents “a more or less permanent change in behavior which
occurs as a result of practice” (Kimble 1961, p.6). Experimentally, perceptual skill
learning is defined as improvement in sensory discrimination after practice (Karni and
Sagi 1991); motor skill learning is generally evaluated by improved efficiency and
speed of task execution (Poldrack 2005; Doyon and Benali 2005).
Perceptual learning
Perceptual learning essentially implies detecting discriminatory issues between
initially very similar stimuli (Mitchell 2009). The behavioral improvement is
underlined by stable learning-related changes at the neural level, which persist after
the active stimulation. A broad literature in the perceptual field deals with visual tasks
and practice-induced improvement in visual performance (Ahissar and Hochstein
2004). This involves training-related increased expertise to detect and extract the taskrelevant information from a mixture of input signals. A simple example is provided by
the experimented radiologists who can readily identify abnormalities in highly
complex images, whereas this task is rather impossible for an untrained person
(Skrandies and Fahle 1994; Sasaki, Nanez et al. 2010). Experimentally, it has been
shown that experience-dependent changes in neural patterns may happen at the
earliest cortical stage of visual processing, i.e., in primary visual cortex (Karni and
Sagi 1991; Schwartz, Maquet et al. 2002; Furmanski, Schluppeck et al. 2004; Li,
20
Piech et al. 2004; Pourtois, Rauss et al. 2008). This has been shown by using cell
recordings in animals (monkeys) and brain imaging in humans. These data suggest
that perceptual learning is highly localized as it essentially depends on selective local
changes, specifically driven by the task performance (Seitz and Watanabe 2005).
However, these findings do not rule out the involvement of a broader spectrum of
areas, as the strength of neural connections between low-level visual areas and higher
integrative areas might also undergo task-specific modulations (Schwartz, Maquet et
al. 2002; Ahissar and Hochstein 2004). This in agreement with general learning
theories, which state that the interactions between bottom-up sensory inputs and topdown goal-directed influences are critical for consolidating memory traces and
therefore lead to a better performance (Gilbert, Sigman et al. 2001). In the above
studies dealing with visual tasks, it has been proposed that perceptual learning recruits
a rather highly localized network.
Motor learning
Motor learning generally refers to the gradual process by which more or less
complex motor behaviors come to be performed without effort through repeated
practice and/or interactions with the environment (Willingham 1998). Motor skill may
refer to either simple, repetitive movements (such as when playing the „„ball and cup”
(Milner, Fogel et al. 2006), or to more complex sequences of movements (such as
when playing the piano) (Krings, Topper et al. 2000). The learning of sequences of
movements and the underlying neural dynamics across vigilance states represents one
of the research topics addressed in this thesis. Motor sequences can be explicitly
known or not, but even in the latter case, the regularity of the motor sequence can still
be learned, even though it remains implicit. The multitude of tasks used to
experimentally investigate motor skill learning evaluate either the incremental
acquisition of movements that subjects have to execute as quickly and accurately as
possible such as finger tapping tasks (Walker 2005; Karni, Meyer et al. 1995; Maquet,
Laureys et al. 2000; Maquet, Laureys et al. 2003; Hotermans, Peigneux et al. 2006), or
the ability to compensate for environmental changes such as motor adaptation tasks
consisting in adapting the motor reaction to externally manipulated conditions
(Shadmehr and Holcomb 1997; Huber, Ghilardi et al. 2004).
21
At the brain level, different anatomical structures are involved in motor learning:
the motor cortex, the cerebellum and the basal ganglia seem to play a key role in skill
learning, whether it is about sequential motor learning or about motor adaptation
tasks. Depending on the stage of learning, the primary motor cortex, supplementary
motor cortex, the cerebellum and the putamen are mainly activated at an early stage,
while the supplementary motor area, precuneus and prefrontal cortex are rather active
at a more advanced stage of learning.
The task demands (e.g. motor or more cognitive demands) (Doyon, Song et al.
2002; Doyon, Penhune et al. 2003) also modulate the anatomo-functional
contributions, with the cortico-striatal pathways classically known to mediate learning
of implicit motor behaviors (Doyon, Bellec et al. 2009), while the cortico-cerebellar
loop is rather important for motor adaptation. Recent reports suggest that both implicit
and explicit learning of a sequence of movements requires the contribution of the
hippocampus and its associated medio-temporal limbic areas (Schendan, Searl et al.
2003; Albouy, Sterpenich et al. 2008). Therefore, this type of motor acquisition would
require not only the cortico-striatal and cortico-cerebellar systems, but also
hippocampal regions. It seems that these differently contributing brain areas interact to
create new memories coding for the acquired behavior.
Attentional influences on learning
Attention contributes to the improved performance after perceptual learning by
enhancing the selection of task-relevant features of the stimulus. It has been proposed
that attention modulates early stages of perceptual processing in a task-relevant
manner (Schwartz, Vuilleumier et al. 2005; Klemen, Buchel et al. 2009). It is equally
accepted, in the visual domain at least, that attentional processes may be engaged in a
top-down control of early stages of sensory information processing (Lamme and
Roelfsema 2000). However, attention may not be mandatory for learning as
demonstrated by recent studies showing performance improvements outside the focus
of attention (task-irrelevant learning) (Watanabe, Nanez et al. 2001; Watanabe, Nanez
et al. 2002; Seitz and Watanabe 2005).
22
2.3. The temporal dynamics of perceptual
and motor skill learning
In any type of learning, be it perceptual or motor, the factor time represents a
critical dimension, because during learning the brain learns new routines gradually,
through practice and interactions with the environment. At the behavioral level, the
gradual increase in accuracy and speed level does not only benefit from the intrapractice sessions, but also from the inter-practice periods, with no additional training
(Karni and Sagi 1991; Karni, Meyer et al. 1995; Karni, Meyer et al. 1998; Doyon,
Penhune et al. 2003; Press, Casement et al. 2005; Halsband and Lange 2006). At the
brain level, the neural representations of the newly acquired skill, whether referring to
perceptual or to motor behaviors, undergo time-dependent re-shaping, leading to
better performance.
The time-course of skill learning can be divided into several distinct stages. A
well-studied, „typical” example is that of a visuo-motor learning task, namely the
serial reaction time task. Subjects are facing a computer screen where visual cues
could appear successively at four different spatial locations, arranged horizontally.
Each of the four possible positions of the visual cue on the screen corresponds to one
of four response buttons on a response pad. When a visual cue appears on one of the
position, the subjects are instructed to react as quickly and accurately as possible by
pressing the spatially corresponding response button. The succession of the visual
cues appearance follows a sequential order, pre-defined or probabilistically
established. Firstly, considerable improvement occurs within one session of intense
training; this is an initial, “fast” stage, where the performance is under sensory (e.g.
visual) control. A second, intermediate, “slow” stage follows, where there is gradual
learning of the sensori-motor (e.g. visual-keypress response) associations across
repetitive training. A third “consolidation” stage following after a certain time interval
without training (Ungerleider, Doyon et al. 2002) leads to further performance
increases in the absence of additional training. A following, “automatic” stage is
proposed, where the task is correctly executed with minimal effort (Doyon and Benali
2005; Halsband and Lange 2006). Thus, it takes time for long-lasting plastic brain
changes necessary for consolidating the motor routine to occur.
23
The individual contribution of the brain regions involved in motor skill learning
varies across different stages of the acquisition: the cerebellum seems more involved
in the early stages of acquisition, while the basal ganglia is important for later stages
of learning (Doyon and Benali 2005). Plasticity-related changes during motor skill
learning occur at both intra- and inter-system levels. Besides, it has been shown, by
using functional brain imaging, that motor representations may shift from the
associative cortex to the striatum during the explicit learning of a motor sequence
(Hikosaka, Nakamura et al. 2002). A transfer of activity from the cerebellar cortex to
the dentate nucleus may accompany the implicit acquisition of a known sequence of
movements (Jueptner, Frith et al. 1997).
Perceptual learning, much like motor skill learning, is accompanied by timedependent changes across distinct cortical areas, over shorter or longer time scales.
Yotsumoto et al. (2008) trained subjects during few weeks on a visual task and have
showed an increase in both execution performance and brain activity in early vision
regions corresponding to the trained visual quadrant. Once performance reached an
optimal level and remained constant, the brain activation in the corresponding areas
decreased. These results suggest that distinct temporal phases characterize perceptual
learning. Mukai et al. (2007) showed in participants who were trained on a visual
learning task, that once a certain level of performance is reached, the activation in
early visual areas as well as in higher-level regions associated with attentional
processes decreased. It has been shown that stimulus specificities found in perceptual
learning were maintained across time (Parkosadze, Otto et al. 2008). For instance
improvement for vertical bisection stimuli (a central element bisecting an interval) did
not transfer to horizontal bisection. So, the time-related improvement for this
perceptual task is specific for the stimulus type. Similarly, it has been shown in the
motor domain, by studying intermanual transfer of a visuo-motor sequence, that what
is acquired is the specific regularity of the motor sequence and not a spatial pattern,
and intermanual transfer of implicit learning was found only in a mirror image of the
originally trained sequence (Wachs 1994).
24
3. ANATOMO-FUNCTIONAL CORRELATES OF MEMORY
3.1. Memory and brain plasticity
It is known that activity-dependent modulation of synaptic efficiency (e.g.
synaptic strengthening) and also structural changes (e.g. changes in cell dimensions
and number and in axonal length) represent the neurobiological bases of memory
formation. The activity-driven molecular and cellular changes support consequent
remodeling of functional neural circuits at larger scales in the brain (Bruel-Jungerman,
Davis et al. 2007). While neural plasticity is most evident during development,
experience can reshape neural networks throughout all life span.
The plasticity mechanisms which contribute conjointly to the formation of longterm memory are represented by long-term potentiation (LTP, rapid and enduring
synaptic strengthening), long-term depression (LTD, weakening of synaptic strength),
synaptogenesis (growth of new synapses) and neurogenesis (formation and growth of
new neurons) (Bruel-Jungerman, Davis et al. 2007).
One of the most extensively investigated forms of synaptic change related to
learning and memory is the LTP. For example, it has been shown in rats that the
synaptic strengthening (LTP) implying a modulation of amygdaloidal synaptic
transmission is critical for fear conditioning (Shaban, Humeau et al. 2006). Further
demonstration of the LTP as a necessary condition for the maintenance of memory is
brought by studies where suppression of LTP after learning, disrupts a previously
established memory trace (Pastalkova, Serrano et al. 2006).
The LTD mechanism based on the rule of “use it or lose it”, by weakening
unused connections, may promote the consolidation of patterns of intensely used
specific neural connections, which could contribute to the memory storage.
Synaptogenesis, i.e. the actual formation of new viable synapses, has been a subject of
debate: does this process reflect a true synaptic growth or just morphological changes
at the existent synaptic level? For instance, exposure to a “rich” environment or
certain learning paradigms have been shown to determine an increase in synapses
number (Markham and Greenough 2004). Several studies promote the idea that
hippocampal neurogenesis may be triggered by learning of specific tasks and
furthermore, that the formation of new neurons is related to the task difficulty and
25
hippocampal involvement in the learning process (Leuner, Mendolia-Loffredo et al.
2004). Moreover, it has been shown that a blockage of neurogenesis in the adult
hippocampus specifically disrupts the eye-blink conditioning, which is a form of
learning depending on the hippocampus (Shors, Miesegaes et al. 2001). Plasticity
mechanisms engaged in learning and memory processes are also associated with rapid
gene regulation in different brain areas, depending on the type of the informational
input. For example, the expression of a class of immediate early genes (IEGs) seems
particularly important for triggering transcriptional control mechanisms subserving the
stabilization of different types of memory, especially fear related memories, olfactory
and spatial memories (Davis, Bozon et al. 2003; Bailey, Wade et al. 2009; Loebrich
and Nedivi 2009).
All in all, the plasticity mechanisms highlighted above contribute to memory
related functional and structural changes across different brain areas. As an example,
the primary visual cortex (V1) of the adult mammalian brain has provided one of the
clearest evidence of the experience-dependent plasticity processes at cortical circuits
level, which could be modeled according to a variety of manipulations, such as
perceptual learning and visual deprivation (Schwartz, Maquet et al. 2002; Karmarkar
and Dan 2006).
The concept of learning as “organized knowledge which grows and becomes
better organized” (McDermott 1985) also suggests that learning mechanisms may
operate at some relatively abstract levels, which would imply accumulating,
organizing and re-organizing information, leading to improved processing efficiency
and thus changes in behavior. At the brain level, learning can be viewed as
“experience-dependent lasting modification in neural representations” (Dudai 1989).
This definition highlights two key principles: first, the changes in brain activity are
“experience-dependent” (therefore, the changes are caused by experience) and
secondly, this definition promotes the notion of “neural representation”, in the sense
of an acquired “code” or “trace” of the learned knowledge (for example, newly
encountered stimuli, or the forming of new associations between already learned
stimuli). Furthermore, the formed neural representations are of dynamic nature in the
sense of flexible processing rather than fixed encoding of information. Thus, learning
and memory involve cognitive and neural reorganization at multiple levels of
integration (which is a key notion for the experimental part of the thesis).
26
3.2. Brain systems for memory processes
In the following sections, I will briefly overview same anatomical and functional
characteristics of different brain regions involved in memory, with a special emphasis
on the hippocampus as a key contributor to memory formation. As shown above, there
are distinct memory systems. Until recently, it was considered that these distinct
memory systems also involve selective brain systems: procedural memory would
involve cortical regions (motor or sensory cortices), the striatum, and the cerebellum
among others (Grafton, Mazziotta et al. 1992; Doyon, Gaudreau et al. 1997;
Ungerleider, Doyon et al. 2002; Doyon, Penhune et al. 2003; Monchi, Petrides et al.
2006), whereas declarative memory would involve the medial temporal lobes, in
particular the hippocampus (Squire and Zola-Morgan 1991; Squire 1992; Reber and
Squire 1994; Squire and Zola 1996; Squire 2004; Squire, Stark et al. 2004). However,
recent studies challenge the above schema. For instance, there is evidence for a
hippocampus role in the learning of a sequence of movements (Albouy, Sterpenich et
al. 2008) (Schendan, Searl et al. 2003), which is classically considered as a procedural
task (see also below).
Hippocampus and related structures
Together with adjacent perirhinal, entorhinal and parahippocampal cortices, the
hippocampus is often referred to as the medial temporal lobe (MTL) memory system
(Squire and Zola-Morgan 1991). The organization of the MTL system involves
bidirectional pathways between the cerebral cortex and the hippocampus, and these
pathways are largely conserved across mammalian species (Manns and Eichenbaum
2006).
Neurons within the hippocampus form a network different from that found
anywhere else in the nervous system (Andersen, Morris et al. 2007). Firstly, the
relatively simplified two- or three-layered architecture, with its strict layering of
synapses in the dendritic arborisation is a characteristic proper to the hippocampus.
Therefore, the hippocampal neuroanatomy sustains a unique internal network
organization, taking into account the large unidirectional information flow through
hippocampal circuits and the highly distributed organization of within hippocampus
associative connections. This high degree of neuronal interconnectivity within the
27
hippocampus allows the consequent integration and comparison of the information
(Amaral, Ishizuka et al. 1990). These anatomical properties may help to understand
why the hippocampus is one of the only few brain regions that receive highly
processed, multimodal information from different neocortical regions. Across species,
the cortical association areas do not target directly the hippocampus, but instead
connect to a sum of related areas within the parahippocampus region. Therefore, the
flow of information from cerebral cortex is first directed to the parahippocampal
region, whose outputs converge to the hippocampus; the outputs of hippocampal
processing are directed back to the parahippocampus, which, in turn, send its outputs
to the same original cortical areas: therefore, the cerebral cortex, the parahippocampal
region and the hippocampus can be viewed as a hierarchy of connectivity (Lavenex
and Amaral 2000; Eichenbaum and Lipton 2008).
Small neuronal ensembles in the hippocampus form local circuits which may
generate spatiotemporal patterns of spiking activity, therefore converting relatively
unstructured inputs into specific patterns of activation (Takahashi and Sakurai 2009). In
other words, the local connectivity within the hippocampus may provide the mechanism
by which any initial input can “organize” the response to any subsequent input.
Implications of the hippocampus in memory functions
It is known that the hippocampal structure plays a key role in the episodic
memory formation (Alvarez and Squire 1994; Eichenbaum 2004; Ergorul and
Eichenbaum 2004; Squire 2004), as part of the explicit declarative memory system.
However, recent neuroimagery results provide a new perspective upon hippocampusrelated memory processes, by demonstrating the direct hippocampal involvement in
the processing of non-declarative memory (Schendan, Searl et al. 2003; Albouy,
Sterpenich et al. 2008). In a functional magnetic resonance study, Albouy and
colleagues showed by training subjects on an implicit oculomotor sequence learning
task and testing them at different time intervals, that both the hippocampus and
striatum interact and are involved in the consolidation of motor sequence memory,
thus bringing support for the hippocampal involvement in long-term formation of
procedural memory. These results confirm the cell-level evidence for the hippocampal
capacity to associate temporally fragmented, but structured information to form
integrative, coherent memory representations (Burgess, Maguire et al. 2002).
28
Moreover, recent theoretical approaches suggest a potential role of hippocampal
circuitry in representing sequences of events, based on its anatomical and functional
capacity to mediate three prominent cognitive features: associative representation,
sequential organization and relational networking (Wallenstein, Eichenbaum et al.
1998; Eichenbaum 2004). Rich experimental animal data supports this perspective
(Agster, Fortin et al. 2002; Fortin, Agster et al. 2002; Lee and Wilson 2002; Ergorul
and Eichenbaum 2004) and provides further evidence for firing patterns in the
hippocampus time-locked to individual sequential events (Louie and Wilson 2001;
Foster and Wilson 2006). Navigation studies in rodents report specialized neural
networks co-mapping visual and spatial features into flexible representations
(Ekstrom, Kahana et al. 2003), thus confirming the hippocampus relevance for
navigation, path integration and cognitive mapping (McNaughton, Battaglia et al.
2006). Furthermore, while initial studies showed the role of hippocampal neurons in
signaling relative to the position of the animal in the environment (“place cells”), thus
forming a spatial cognitive map of the environment (O'Keefe and Dostrovsky 1971),
the idea also emerged that spatial information encoded in the “place cells” actually
reflects a more general category of relational information, which depends on the
hippocampus (Eichenbaum 2001). It has been therefore proposed that the
hippocampus has the capacity to build and store “configural” associations between
separate events. Studies on associative memory provided further insights concerning
the hippocampus capacity to form flexible new associations between different stimuli
irrespective of modality (Henke 2010). Chadwick et al (2010) presented three distinct
short videoclips of every day life episodes to healthy human subjects and than asked
them during an fMRI scanning session to recall as much details as possible of each
episode a number of times. The authors have than directly examined neural
computations within hippocampus related to the three episodic representations and
were able to indentify from the pattern of fMRI hemodynamic response whioch
specific episode of memory has been recalled. They therefore showed that neural
traces of recently encoded episodic memories are detectable from the BOLD patterns
of fMRI.
The role of hippocampus in relational memory is a central research question
addressed in the personal part of the thesis.
29
Amygdala and emotional memory
Since the demonstration in monkeys by Mishkin et al. (1978) that conjoint
lesions of amygdala and hippocampus resulted in greater impairments on declarative
memory tests than did individual lesions of each structure, the contribution of
amygdala to memory either as a proper storage area or as a modulator of mnemonic
processes has become an important topic of research. Fear conditioning, a special
form of procedural memory based on emotional learning and on Pavovlian
conditioning, has been the initial paradigm use to study amygdala function in memory
(LeDoux 2003). Recent long-term memory consolidation theories implying amygdalohippocampal connections, demonstrate that amygdala also plays an enhancing role in
the encoding and long-term recall of emotional stimuli (McGaugh 2004; Sterpenich,
Albouy et al. 2009).
Diencephalon and declarative memory deficits
Damage to diencephalic regions (e.g. the thalamus, the hypothalamus) is
currently believed to impact memory not by actually disrupting memory processing
per se, but rather by altering communication between the medial temporal lobe and
other regions, responsible for memory processes. For instance, the mammillothalamic
and the amygdalofugal tracts are two pathways through which structures of the medial
temporal lobe are connected to memory structures. This explains the presence of
amnesic elements in patients with Korsakoff's syndrome, who do not show significant
damage of medial temporal lobe structures, but rather present damage to diencephalic
structures, including mammillary bodies and the dorsomedial nucleus of the thalamus
(Hampstead and Koffler 2009).
Motor areas, basal ganglia, cerebellum
and memory for motor skills
There is extensive evidence for the involvement of motor brain regions such as
supplementary motor area, primary motor and premotor cortices in the memory for
motor behaviors (Doyon, Penhune et al. 2003). Moreover, it has also been suggested
that areas previously acknowledged to subserve pure motor executive and control
functions (e.g. the basal ganglia), are also mediating the memory formation of learned
30
sensory-motor associations (White 2009). The other aspects related to motor skill
learning and contributing areas has already been tackled in previous sections, and will
be further discussed in the main experimental part.
Prefrontal cortex and memory consolidation
The dorsolateral prefrontal cortex may play a critical role in memory
consolidation processes, particularly for hippocampal-dependent spatial and
contextual information (Frankland and Bontempi 2005). Because of its dense
interactions with both MTL structures and the basal ganglia, the prefrontal cortex may
support hippocampal-prefrontal interactions in memory processes. Namely, it may
participate to the extraction of regularities based on internal representations, so as to
improve behavioral control (Robertson 2007; Shima, Isoda et al. 2007). In line with
this idea, the next section will detail memory consolidation models, based on
reactivation of hippocampal-prefrontal neural patterns across different vigilance states
to allow stabilization of the acquired knowledge and subsequent performance
improvement (Peyrache, Khamassi et al. 2009).
3.3. Distinct steps in memory consolidation
The term “consolidation” was first associated with memory processes by the
German researchers Müller and A. Pilzecker, in their work, Experimentelle Beiträge
zur Lehre vom Gedächtnis [Experimental contributions to the science of memory].
Zeitschrift für Psychologie Ergänzungsband, 1, 1–300 (1900). The authors noticed
that learning new stimuli immediately after training on series of non-sense syllables
interfered with the recall of the items learned first. They postulated therefore that it
takes time for the memory trace to become stable and resistant to interferences.
Memory consolidation defines a continuum of experience-driven processes, both
at synaptic and systemic level, that render an initially fragile memory trace resistant to
interferences through stabilization. In other words, it operates a spatial and temporal
reorganization of recently acquired information into a stable representation (Dudai
2004; Buzsaki and Chrobak 2005). At the synaptic level, memory consolidation
involves a cascade of molecular events, implying the restructuring of existing
synapses (Dudai 2004) and the formation of new connections (Malenka and Nicoll
31
1999), within the first few hours after learning. The changes at the synaptic level
determine modulations in gene expression and consequent synthesis of new proteins
contributing to structural changes (Bozon, Davis et al. 2003). System-level
consolidation is a slower process that involves the gradual reorganization of neural
representation of new memories over extended periods of time (Frankland and
Bontempi 2005). One prevalent view of such memory consolidation process is that the
hippocampus would act as a temporary store for newly acquired information, which is
then gradually transferred, for permanent storage, on a distributed cortical network
(McClelland, McNaughton et al. 1995; Squire and Alvarez 1995; Frankland and
Bontempi 2005). Figure 1 is adapted from and illustrates distinct steps in memory
consolidation: first, the hippocampus integrates newly acquired information from
different cortical modules (Figure 1 A); successive reactivations take place within the
hippocampal-cortical network, with time (Figure 1 B); the gradual strengthening of
cortico-cortical connections renders the newly formed memories independent of the
hippocampus (Figure 1 C).
Figure 1. Time-dependent memory consolidation model
(A,B,C explained in the text)
Thus, the initial encoding of the sensory, cognitive and motor information takes
place in different primary cortical (visual, auditory, olfactive and somatosensory) and
associative areas. The hippocampus integrates the different inputs coming from
cortical modules and maps them into coherent representations. Then, successive
reactivations of hippocampal-cortical networks allow the reinforcement of the
32
network intraconnections, either by strengthening the pre-existent links or by creating
new ones (Frankland and Bontempi 2005). The progressive enrichment of the corticocortical connections by synchronous reactivations renders the new traces gradually
independent of the hippocampus and facilitates their integration into the pre-existent
cortical memory representations. Thus, the changes in the connectivity between the
hippocampus and the neocortex happen first, then changes in connectivity among the
different cortical regions can occur (Dudai and Eisenberg 2004; Frankland and
Bontempi 2005).
Recently, a third step in the process of memory consolidation has been
identified: memory reconsolidation. It has been postulated that, once stabilized, a
memory trace is maintained in a “latent state” and has the capacity to return to an
“active state”, whenever the recall of the original information is needed. When
reactivated, the memory re-enters a dynamic and fragile state, requiring further active
processes of stabilization (additional consolidation) via synaptic changes and protein
synthesis (Nader, Schafe et al. 2000; Alberini 2005).
Another complementary view on memory consolidation processes is provided by
the Multiple Trace Model (MTM). MTM is principally based on the distinction
between semantic memory and episodic memory, arguing that while the hippocampus
is necessary for the retention and retrieval of episodic memories, semantic memories
are created by multiple traces organized in the neocortex through the process of
consolidation and do not specifically depend on the hippocampus. According to this
model, acquired information is saved in a distributed network, including the
hippocampus and cortical regions (Nadel and Moscovitch 1997). The MTM is
supported by three main arguments: (1) episodic memory is completely abolished if
the medial temporal lobe (containing the hippocampus) is damaged (development of
retrograde amnesia); (2) the hippocampus is active when recalling episodic,
autobiographic events (Gilboa, Winocur et al. 2004); (3) the forgetting rate in amnesia
is proportional with the volume of hippocampal lesion (Cipolotti, Shallice et al. 2001).
Taken together, these observations point out the hippocampus critical
involvement in the formation of long-term memory.
33
4. THE INFLUENCE OF SLEEP-WAKE STATES
ON MEMORY PROCESSES
Although the functions of sleep remain largely unknown, one of the most
exciting hypothesis is that sleep contributes importantly to processes of
memory and brain plasticity.
Walker and Stickgold, Sleep, memory, and plasticity, Annu Rev Psychol, 2006
If the mechanisms underlying the formation of new memories are quite well
understood, the precise timing of the gradual transformation of hippocampusdependent recent memories into cortex-dependent remote memories, are still a matter
of debate. One prevalent hypothesis is that memory consolidation involves the
reactivation of neural traces acquired during wakefulness, and that sleep may offer a
permissive condition for such reactivation to occur (Ribeiro, Mello et al. 2002; Walker
2005; Dang-Vu, Desseilles et al. 2006; Peigneux, Orban et al. 2006; Karlsson and
Frank 2009). Therefore, memory consolidation cannot be fully understood without
considering both wake and sleep brain states.
.
4.1. Sleep: a beneficial role for memory consolidation?
Aspects of sleep physiology
Human sleep consists of cyclic succession of distinct stages, characterized
depending on the EEG patterns, eye movement and mucle tonus: non rapid eye
movement sleep (NREM sleep with stages 1, 2, 3, 4), generally characterized by high
amplitude and low frequency EEG activity, slow eye movements and relatively
diminshed muscle tonus and rapid eye movement sleep (REM sleep) which is
characterized by low amplitude and high frequency EEG activity, rapid eye
movements and abolition of muscle tonus. Sleep is not a uniform state, as distinct
neurobiological processes cyclically alternate throughout a night of sleep according to
an ultradian rythm of 90-100 minutes. Nocturnal sleep in humans can be divided into
an early sleep (in which SWS predominates) and late sleep, in the second part of the
night (in which REM sleep predominates). Each sleep phase, namely NREM and
REM sleep, is characterized by dramatic changes both at electrophysiological and
anatomo-functional level (Walker 2008). The NREM sleep presents some specifical
34
EEG oscilations: spindles, delta waves and slow wave activity. Sleep spindles reflect
brain oscillations within 11-15 Hz (sigma band) and of less than 0.5 ms duration.
Delta waves are high amplitude oscillations, with a frequency between 1 and 4; the
slow waves are characterized by a frequency less than 1Hz. Stage 2 of sleep, named
light slow wave sleep, contains sleep spindles and some specific phasic
graphoelements named K complexes; stages 3 and 4 of NREM sleep are characterized
by the predominance of slow wave sleep (named SWS). At cellular level, the different
sleep oscillations ae sustained by different brain structures: sleep spindles are
generated by thalamo-reticular neurons (Steriade 2005). The slow waves are initiated
by the neorcortex.
Sleep duration and intensity are regulated by two physiological processes:
homeostatic process and circadian rythm. The circadian rythms generate a close to
24h biological rythm, entrained via the suprachiasmatic nucleus, brain structure which
plays the role of an internal clock. External stimuli such as light, food intake and
social constraints train the internal clock to regulate with high precision the circadian
rythm of sleep. The homeostatic control reflects the growing sleep pressure
accumulated with time spent awake. The homeostatic pressure may be quantified by
measuring the slow wave activity (Borbely and Achermann 2005). Sleep is therefore a
highly controlled process, through the complex and permanent interaction of circadian
and homeostatic mechanisms (Borbely 1982; Franken and Dijk 2009).
Perturbations in sleep architecture and physiology generate sleep disorders, such
as parasomnias (sleepwalking, nightmares, REM sleep behavior disorder, etc), in
which wake and sleep elements co-exist, and thus provoke undesirable physical or
verbal overt behaviors during sleep, probably by inappropriate activation of systems
or programs normally differently modulated during sleep, such as autonomic nervous
system, or programs, such as cognitive, behavioral, or motor program stimulation
(Guilleminault 1994).
35
Figure 2. Sleep architecture A. Sleep stages in a young adult.
B. Different stages of sleep and wakefulness
(EEG: electroencephalogram, EOG: electroculogram, EMG: electromyogram)
Sleep functions
From Thomas Edison‟s view of sleep as “a waste of time” (he believed long
sleep is a sign of laziness), to Francis Crick, suggesting that sleep allowed the brain to
“take out the trash” by removing non relevant information not needed for long term
storage, several hypotheses proposed a role of sleep in memory consolidation.
Therefore, beside the extended number of documented implications of sleep in
metabolic recovery (Spiegel, Leproult et al. 1999), cellular energy conservation
(Benington and Heller 1995), thermoregulation (McGinty, Alam et al. 2001), brain
detoxification (Inoue, Honda et al. 1995), tissue restoration (Everson, Laatsch et al.
2005), hormonal balance (Van Cauter, Spiegel et al. 2008) and genetic reprogramming
(Jouvet 1998), a potentially critical contribution of sleep to plasticity mechanisms and
memory stabilization processes have gained increasing scientific interest (Maquet
2001; Peigneux, Laureys et al. 2004; Stickgold 2005; Diekelmann, Wilhelm et al.
2009). It has been therefore suggested that sleep is required both for brain maturation
at earlier stages in life and for consolidating memory traces throughout the life span of
an individual (Maquet 2001).
36
Maybe the first systematic study on sleep and impact on memory was conducted
by Jenkins and Dallenbach (1924), who trained two subjects on lists of nonsense
syllables at different times during the day. When tested at various time intervals after
learning, recall was much better after a period of sleep than a comparable period of
wake, showing that retention is much higher after period of sleep than after period of
wakefulness. The result was interpreted in the context of the interference theory, based
on the assumption that sleep is a passive state, lacking sensory interference and
therefore, preventing any external disruption of the ongoing memory processes.
Some researchers argue that these specific changes impact differently the
memory traces, depending on the sleep phase where they occur: this view is reflected
by “the dual-process” theory (Plihal and Born 1999) which stands for a contribution of
NREM sleep on more declarative aspects, while REM sleep is considered to play a
more prominent role on procedural tasks (although this strict rule is challenged by
studies which show that sleep stage 2 also contributes to learning of some procedural
tasks (Fogel, Smith et al. 2007) and that REM sleep plays a role in the consolidation
of emotional declarative memory (Wagner, Gais et al. 2001)). Another view is the
“double step” hypothesis (Giuditta, Ambrosini et al. 1995), which posits that both
sleep stages are required for memory consolidation, regardless of the memory system
involved. According to this theory, the alternation itself of the two different sleep
phases may favor memory processes (Stickgold, Whidbee et al. 2000).
Napping in humans
Beside the night sleep, humans also may perform naps, which are usually
defined as short periods of sleep during day time. It is not yet clear if the same
molecular and physiological processes that occur during night sleep also occur during
napping. Yet, the behavioral and cognitive benefits of a nap are already welldocumented. Usually, naps have been used as a countermeasure for sleepiness and
fatigue, as they improve arousal and alterness. Napping has been also shown to
increase memory performance, even in the absence of fatigue and sleepiness. Studies
on the benefit of napping on performance have demonstrated that short day time sleep
episodes have selective enhacement on various forms of memory (Mednick 2003;
Schmidt 2006; Backhaus 2006). Furthermore, these studies have showed that the nap
related performance improvement is related to the specific stages of sleep within the
37
nap. Mednick et al (2002) applied a visual perception task to show that a 60 minute
midday nap, rich in SWS, reverses deterioration of a perceptual task. Naps including
both SWS and REM leed to an improvement in performance similar to that of a full
night of sleep (Mednick 2003).
The area of research in napping is still young. Future research concerning the
functions of nappig needs to address physiological and medical benefits of day time
naps that have been studied for nocturnal sleep (Stickgold and Walker 2009).
4.2. Sleep and wakefulness:
recent hypotheses on memory consolidation
4.2.1. Neural reactivations during sleep
Sleep might provide a permissive condition for neurophysiological mechanisms
underlying brain plasticity to occur. This observation emerge from different studies in
animals and in humans, assessing memory processe at different brain levels: cellular,
neuronal assemblies, brain areas, behavioral (Agster, Fortin et al. 2002; Hobson and
Pace-Schott 2002; Ribeiro, Mello et al. 2002; Massimini, Huber et al. 2004; Aton,
Seibt et al. 2009).
Reactivations at the cellular and neuronal networks levels
At cellular level, much light has been brought by studies of individual
hippocampal “place” neurons in rats. These cells exhibit elevated rates of firing
whenever the animal is in a specific location in environment, therefore coding for the
space map. It has been initially shown that the same firing patterns expressed during
waking behavior are spontaneously reactivated during the following period of sleep.
Practically, the same cells involved in a spatial task were specifically reactivated
during subsequent sleep (Pavlides and Winson 1989). In another seminal paper, the
hippocampal place cells presenting highly correlated activity during spatial
exploration in a labyrinth showed high firing correlation during the deep NREM sleep
represented by slow wave sleep (SWS), following the training period (Wilson and
McNaughton 1994). Importantly, the temporal order of firing during post-training
sleep was similar to the firing order during previous wake experience. The temporally
ordered sequence of neuronal bursts was replayed within the CA1 region of the
38
hippocampus during SWS periods, in a “compressed” manner of approximately 20fold the initial neuronal discharge rate (Skaggs and McNaughton 1996; Lee and
Wilson 2002). The temporally structured replay of a recently learned behavior is not
only appearing during SWS, but also during REM sleep, as reflected by long temporal
sequences of synchronized firing patterns in the hippocampus cells during this sleep
phase (Louie and Wilson 2001). Therefore, one can conclude that both NREM and
REM sleep contribute in a non-exclusive manner to memory consolidation processes.
It has been proposed that SWS may promote mainly post-acquisition neuronal
reverberations, while REM sleep is a favoring transcriptional features intervening in
long-term memory storage (Ribeiro and Nicolelis 2004).
The reinstatement of specific patterns of activity during sleep, consistent with
wake behavior patterns, was also in songbirds, especially during song-recognition
behavior (Dave and Margoliash 2000; Gentner and Margoliash 2003; Shank and
Margoliash 2009). By using the birdsong and the comparison during sleep between
vocal production (imprinted in premotor areas and spontaneously replayed during
sleep) and stored auditory feedback, as “off-line” models for learning, the authors
showed that bursts of activity during sleep induced adaptive changes in premotor
networks as part of song learning.
The replayed patterns of activity in the hippocampus reflect specific oscillations:
on one hand, gamma oscillations (40-100 Hz) and theta rhythm (4-7 Hz), which
characterize the hippocampus activity during exploratory behavior and REM sleep.
On the other hand, the resting periods and SWS are characterized by irregular bursts
of spike-like activity calles “sharp waves” and very fast oscillations, termed “ripples”
(140-200 Hz) which represent pattern reinstatement at CA1 hippocampal level
(Kudrimoti, Barnes et al. 1999). The sleep – memory consolidation assumes an
interaction between neocortex and hippocampus during sleep. More specifically,
during learning at wakefulness, the information si encoded simultaneously into
neocortical networks and in a temporary fashion in the hippocampus. During
subsequent SWS, neuronal activity related to recently acquired information is grouped
by the slow oscillations (frequency less than 1 Hz) into up-states (neuronal activity)
and down-states (neuronal diminished activity). This grouping mechanism generates
within the thalamo-cortical network spindle activity and in the hippocampus it
generates sharp-wave ripples, which contribute to the transfer of information from the
39
hippocampus to the neocortex. This transfer is initially associated with an increase in
thalamo-cortical spindle activity. The synchronization orchestrated by the cortical
slow oscillations between spindles within thalamocortical network and hippocampal
sharp-wave ripples seems to be important for the memory long-term storage within
neocortex (Born 2010).
Figure 2. Model of the consolidation of declarative memory|
as an interaction between the neocortex and the hippocampus (after Born 2010)
Taking into consideration these different types of oscillations and their specific
“windows” of appearance, Lorincz et al. (2000) elaborated a computational model
based on two complementary phases, aimed at explaining the information trafficking
within the hippocampal-neocortical axis and consequent long-term memory
stabilization. The model advocates that during gamma and theta oscillations, the
neocortex sends its inputs to the hippocampus through the enthorinal region. The
opposite phenomena (i.e. the hippocampus sends information to the neocortex via the
enthorinal cortex) coincide with the ripples and sharp waves activity. In its most
general form, the model proposes that memories are stored within the hippocampus
during wake experience and transferred to the neocortex during sleep, where sharp
waves and ripples are suggested to drive Hebbian synaptic changes (increase in
synaptic efficacy induced by repetitive and persistent stimulation of the post synaptic
cell by the presynaptic cell) in the neocortical target areas. From a computational
perspective, the averaged firing rates of neurons reflecting several hundred
milliseconds of behavior provide a “rate” code of the recorded inputs. The influential
Hebbian laws of learning and synaptic plasticity require a precise temporal code of
40
spikes sequence over very short timescales (generally less than several milliseconds).
It has been established for the hippocampal pyramidal neurons that their rate of
spiking and temporal codes are highly correlated, which can be critical for
understanding the fast synaptic stabilization of temporally ordered sequences of events
occurring at large behavioral scales (Mehta, Lee et al. 2002).
Neuronal reactivations occur equally in other brain areas besides the
hippocampus. Siapas et al. (Siapas and Wilson 1998) have provided a clear
demonstration of temporal correlations between the hippocampal fast ripples and
spindle (7-14 Hz) oscillations present in the prefrontal cortex during NREM sleep.
The authors proposed that the correlated activity of hippocampal and neocortical
regions may be important for the gradual transfer of memory traces from the
hippocampus to neocortex within the dynamics of memory consolidation.
Post exposure brain reactivations have been concomitantly proven during sleep
in the posterior parietal neocortex and in the CA1 region of the hippocampus,
conferring arguments for the idea that representations in the two areas may reflect the
same experience and specific information transfer between the hippocampus and
necortex (Qin, McNaughton et al. 1997). Along the same lines of evidence, Sirota et al
(2003) found that the somatosensory cortex and hippocampus show oscillationmediated temporal correlations during sleep. Ji and Wilson (2007) described during
SWS in rats, the reactivation of the same temporally organized spiking patterns
evoking previous waking experience, in the visual cortex and the hippocampus.
Euston et al. (2007) elegantly showed in the prefrontal cortex the compressed
reactivation of spatiotemporal waking patterns, which could again reflect the memory
consolidation processes and neural pattern rehearsal in an “off-line” manner. In
support of a role of hippocampal-prefrontal interactions in memory processes,
Peyrache et al. (2009) recently demonstrated that following the acquisition of new
rules, prefrontal activity reflected neural patterns that occurred during active behavior
and that this reactivation occurred predominantly when cortico-hippocampal
interaction was enhanced. Thus, learning can influence the selection of neural patterns
subsequently reactivated in hippocampal-prefrontal networks.
The above mentioned studies bring valuable insights into the plastic changes that
underlie memory consolidation across distributed brain networks, as reflected by the
41
coordinated reactivation of memory traces in the hippocampus and different
neocortical regions.
Reactivations at the brain areas level
Functional brain imaging studies in humans found that following intensive
training on perceptual or motor tasks, the same brain regions active during task
performance showed changes in cerebral blood flow (Maquet 2001; Peigneux,
Laureys et al. 2004) during sleep. Notably, by using positron emission tomography
(PET), it has been shown that during REM sleep some brain areas were more active in
subjects previously trained on a visuo-motor serial reaction time task than in nontrained, control subjects (Maquet, Laureys et al. 2000). The elegant experimental
procedure consisted, first, in determining in a group of healthy subjects, the brain
areas engaged in the performance of a sequence of events formed on a probabilistic
grammar (probabilistic serial reaction time task). Then the subjects were scanned
during post-training sleep, in order to identify the brain regions active during REM
sleep. Another, group of subjects were also scanned during sleep, but without being
trained on the task before sleep. The trained group showed during REM sleep
compared to REM sleep of untrained subjects the activation of some some brain areas
(left premotor area, bilateral cuneus, inferior part of the thalamus and mesencephalon)
which were also previously found to be involved in the task execution during
wakefulness. These results provided therefore the first evidence of off-line
reprocessing of a recently established memory trace in human sleep. Behaviorally, the
task execution improved after sleep in the trained group. The reshaping of this visuomotor network by reactivations during REM sleep could therefore sustain the gain in
performance observed during the following day (Laureys, Peigneux et al. 2001).
Post-training reactivations in the hippocampus have been also found during
NREM sleep in humans. Healthy subjects were trained on a spatial task, consisting in
learning to navigate in a virtual city (Peigneux, Laureys et al. 2004). The hippocampus
activation was observed during the task performance and also during the SWS period
after training. Moreover, individual hippocampus activity during sleep positively
correlated with overnight improvement in performance, therefore arguing that
hippocampus activity reinstated during sleep may play a critical role in memory trace
consolidation.
42
Observations in rats led to the hypothesis that neurotransmitters modulation such
as low cholinergic drive during SWS might be critical for the replay of newly acquired
memory traces in the hippocampus and their long-term neocortical storage (Hasselmo
2008). In their study, Gais et al. (Gais and Born 2004), administered a cholinesterase
inhibitor (physostigmine), and blocked SWS-related consolidation of hippocampusdependent declarative memories (word pairs test) in human subjects and had no effect
on the consolidation of a procedural mirror tracing task, which was not supposed to
relie on the hippocampus. These results show that the optimal re-organization of
memory traces is also actively modulated by neurotransmitters, which could facilitate
selective plasticity processes during sleep.
The studies introduced so far show spontaneous reactivation of experiencedependent patterns of neural activity. By using stimulations applied during sleep,
studies could also demonstrate externally induced neural reactivations (Marshall,
Helgadottir et al. 2006; Rasch, Buchel et al. 2007).
Adapting a cue-dependent recall strategy to a sleep-memory consolidation
paradigm, Rasch and colleagues (2007) showed that the presentation during slow
wave sleep of an odor, which had previously been associated with a declarative
memory task, induced neural reactivations and subsequent declarative memory
improvement. Furthermore, re-exposure to the same odor during REM sleep or
wakefulness or omission of the odor presentation during prior learning did not impact
memory performance. Neuroimaging data confirmed the hippocampus activation
during SWS in response to odor presentation. This study suggests active reprocessing
of recently acquired memory traces, triggered by a relevant context during sleep.
The same research group (Marshall, Helgadottir et al. 2006) also showed in
humans that by externally inducing cortical slow oscillations (and increased sleep
spindles), during early nocturnal NREM sleep, it is possible to facilitate the retention
of hippocampus-dependent declarative memories. The effect was proven to be
selective for the slow oscillation frequency (0.75 Hz), while a 5 Hz stimulation
decreased slow oscillations and had no impact on the memory performance (Marshall,
Helgadottir et al. 2006).
At the cognitive and behavioral level, it seems that selective cues presented to
human subjects during sleep may target the reactivation and strengthening of
individual memories (Rudoy, Voss et al. 2009). Moreover, recent human dreaming
43
studies propose that mental imagery observed during dreaming may reflect off-line
reactivations of previously trained cognitive experiences and therefore, may subserve
memory consolidation processes (Wamsley, Perry et al.; Wamsley and Stickgold).
Taken together these studies suggest that sleep offers a favorable functional
neurophysiological context for synaptic changes to occur. A still unanswered issue is
whether changes in regional brain activity found in humans reflect offline experiencedependent processes, or use-dependent non-specific neurophysiological effects. In the
next sections, I discuss these aspects in more details.
4.2.2. Neural reactivations during wakefulness
Animal models of memory formation suggest that part of the post-training
consolidation process takes place during wakefulness, in the periods of behavioral
inactivity that follow the training on a new task (Buzsaki 1989). This means that recent
memory traces are not only maintained during post-training wakefulness, but are also
altered during this period of time (Frankland and Bontempi 2005). Convergent cellular
data in animals and imaging data in humans, argue that following exposure to a new
task, coordinated reactivations of practice-related neuronal patterns emerge in the
waking period immediately preceding sleep (Pavlides and Winson 1989; Kudrimoti,
Barnes et al. 1999; Peigneux, Orban et al. 2006). Wilson et al showed that during awake
periods immediately after spatial learning, fragments of the spatial experience are
replayed in a temporally reversed order within the rat hippocampus (Foster and Wilson
2006). However, the persistence of the temporal patterning of neural ensembles during
post-training wakefulness is restricted to short time windows, which could reflect
merely a consequence of intensive activation of learning-related neural networks, rather
than an active role in the maintenance of newly acquired information (Pavlides and
Winson 1989; Hoffman and McNaughton 2002).
Evidence in favor of experience-dependent off-line processing of newly acquired
information during post-training wakefulness comes from a functional magnetic
resonance imaging study in humans (Peigneux, Orban et al. 2006). The authors
showed that post-training modulation of brain responses during active wakefulness
occurs in the same brain areas which were specifically involved in the task
performance. Furthermore, the post-training activity correlated with behavioral
performance and the observed modulations followed a different time course for the
44
hippocampus-dependent task than for hippocampus-independent task. This study
argues therefore that during post-training wakefulness learning-dependent changes in
spontaneous regional brain activity might also occur and contribute to the processing
and stabilization of memories.
***
As tested in the already mentioned imagery study by Maquet et al. (2000),
experience-dependent brain activity assumes offline reprocessing of a recently
acquired memory trace and reshaping of the areas selectively involved in the
processing of the newly learned information. Furthermore, experience-dependent
mechanisms taking place during sleep impact behavior in a stable and prolonged
manner, as demonstrated by Stickgold et al (2000).
The use-dependent perspective describes the restoration of optimal synaptic
function after intensive neuronal usage during previous waking hours, as it is
postulated by the downscaling hypothesis (Tononi and Cirelli 2003).
Considering the above issues, it is debatable if memory trace consolidation relies
on the spontaneous reinstatement of specific patterns of activity or/and on local
homeostatic changes (e.g. synaptic downscaling).
4.3. Sleep homeostasis hypothesis
The notion of “local sleep” has become a fundamental property of local neuronal
networks in different brain regions involved in learning and memory processes
(Huber, Ghilardi et al. 2004; Huber, Ghilardi et al. 2006). It was shown that SWA is
regulated as a function of previous wakefulness (it increases with accumulating
wakefulness and decreases during sleep) and therefore may be considered a marker of
homeostatic sleep regulation (Borbely 1982; Tononi and Cirelli 2003). According to
the hypothesis proposed by Tononi and Cirelli (2003), SWA is tied to synaptic
downscaling, in order to preserve synaptic weight after learning and sustain improved
performance.
Using high-density EEG, it has been shown that local increases in slow-wave
activity (SWA) during sleep after intensive learning of a motor adaptation task
correlates with improved performance (Huber, Ghilardi et al. 2004). The authors
proposed that by triggering local synaptic changes with learning of a new task,
increased SWA will be induced in the specifically involved brain regions. Thus, 12h
45
of experimental sensorimotor deprivation (arm immobilization) was also found to
decrease local SWA over the involved cortical regions (Huber, Ghilardi et al. 2006).
In other words, increased synaptic strength after learning tasks would induce
stronger synchronization within selective neuronal assemblies and thereby would lead
to increased local SWA. This mechanism would prevent the “saturation” of brain
plasticity and therefore, would allow more efficient shaping of memory
representations after sleep as reflected by a better performance (Walker 2009).
***
The present framework of memory consolidation give support to the observation
by William James that “the tactile and muscular feelings of a day of skating or riding,
after long disuse of exercise, will come back to us through the night … These revivals
show that profound rearrangements and slow settlings into a new equilibrium are
going on in the neural substance, and they form the transition to that more peculiar
and proper phenomenon of memory” (James, W. Principles of psychology, 1890,
p.647).
5. RATIONALE OF THE PRESENT WORK: MAIN QUESTION
AND METHODOLOGICAL APPROACHES
The aim of the present research is to bridge the gap between animal cellrecording data and human neuroimaging data by studying the temporally-structured
dynamics of experience-specific neural and behavioral activity across wakefulness
and sleep in humans.
As the research framework implies the investigation of spatial and temporal
dynamics of brain network activity, we considered that the most adequate tool to
approach the main question is represented by electrical neuroimaging (EEG). The
critical advantage of the EEG as a modern functional imaging method is the capacity
to capture neural activity within millisecond temporal resolution. We will use both
scalp EEG and intracranial electroencephalographic (iEEG) recordings in humans.
The latter provides a unique opportunity to assess neural activity at both high temporal
and spatial resolution, in particular for deep regions that cannot be easily accessible to
surface EEG recordings. By processing event-related activity and using spatiotemporal analysis of EEG voltage maps, the electrical neuroimaging techniques
46
applied in the present studies, provides a direct assessment of the dynamics
subtending learning and memory functions.
Thus, the main aim of the experiments described hereafter was to assess
learning-related neural reactivations by using the highly informative intracranial
recordings in pharmaco-resistant epileptic patients and large-scale scalp recordings in
healthy volunteers, and to reveal sleep-related behavioral reactivations by using video
recordings during sleep in patients with specific parasomnia (sleepwalking and REM
sleep behavior disorder, RBD). To address these issues, we developed and applied
different experimental strategies. On the one hand, we studied the sequential encoding
and processing of learned events by using versions of serial reaction time tasks,
adapted for the evaluation of implanted epileptic patients and for the testing of
sleepwalkers and RBD patients. On other hand, we used an association learning task
to test the gradual integration of multisensory components into a coherent memory
trace, measured by high-density scalp recordings in healthy volunteers. Finally, we
also applied external sensory stimulations during short periods of daytime sleep, i.e.
naps, to instrumentally modulate electrophysiological features of sleep, and further
evaluate their relevance for memory processes.
***
The main goal of present research work is to investigate the shaping of neural
representations of acquired information across wakefulness and sleep dynamics, and
thus evaluate the potential contribution of plasticity mechanisms to memory
architecture.
47
Experimental part
OVERVIEW
The experiments performed during the present work deal with human wake and
sleep data subtending brain plasticity processes, in a multi-level manner – from overt
behavior to neural populations electrical activity. The thesis design promotes a
transdisciplinar view on brain functioning, through the convergent theoretical and
methodological approaches used. This contributes to the originality of the present
thesis in cognitive neuroscience and provides a valuable perspective for research
continuation.
THESIS EXPERIMENTAL DESIGN
48
Neurophysiological evidence from
human intracranial recordings
of sequenced knowledge consolidation
Experiment
1.
CONTEXT
It has been proven that skill for performing sequenced motor activities continues
to develop after practice has ended, therefore suggesting an “off-line” process of
consolidation (Doyon et al., 2009; Hotermans et al., 2006; Maquet, 2001). Important
support for off-line memory reprocessing comes from cell-recording data in rodents
showing during both sleep and wakefulness, hippocampal replay of the same
sequentially-structured pattern of neural activity as found during previous active
behavior (Louie and Wilson, 2001; Wilson and McNaughton, 1994; Foster and
Wilson, 2006). At the macroscopic systems level, functional brain imaging studies in
humans showed that intensive sequence learning can lead to regional changes in
cerebral blood flow and in the scalp EEG activity, particularly within the
hippocampus (Albouy et al., 2008; Hazeltine et al., 1997; Schendan et al., 2003).
However, the exact contribution of the medial temporal structures to the consolidation
of newly learned sequences of events remains not fully understood in humans.
In the present context, the aim of this first study of the thesis was to combine
elements from animal cellular studies in vivo and human brain neuroimaging
approaches, to assess consolidation of recently acquired structured information and
the underlying pattern of distributed neural activity. Sleep and wakefulness data were
therefore acquired in drug-resistant epileptic patients with implanted electrodes
covering several brain areas while they repetitively performed fixed visuomotor
sequences known to benefit from sleep (serial reaction time task, SRTT) (Maquet,
Laureys et al. 2000).
METHODOLOGICAL HIGHLIGHTS
Intracranial electroencephalographic (iEEG) recordings in humans allow a
particularly valuable level of observation of brain processes in humans by capturing,
with high temporal and spatial resolution, local field potentials (LFP) of distinct
neuronal populations.
49
We implemented a classification method based on a statistical mixed-model and
trained on single motor responses, by considering all implanted areas in each patient.
By using the same statistical algorithm, but dropping in turn each of the recorded area
and re-evaluating the corresponding classification performance, we were able to
identify the areas which critically contributed to classification accuracy. Therefore, the
designed classification algorithm could capture learning-dependent changes in iEEG
signal related to single events, which renders this approach particularly adequate to
study datasets from individual patients.
SUMMARY OF RESULTS
The classifier correctly assigned single iEEG trials from the trained sequence as
belonging to either the initial training phase (day 1, before sleep) or a later
consolidated phase (day 2, after sleep), whereas it failed to do so for trials belonging
to a control condition (pseudo-random sequence). In both patients, the hippocampus
provided a critical contribution to the classification accuracy. Fronto-striatal regions
also contributed to the classification result in one patient with electrodes in this area.
Complementary average event-related potential (ERP) analysis performed within
the same hippocampal sites suggested amplitude modulation as a function of training
phase, selectively for the trained sequence around 200 ms prior to keypress.
CONCLUSION
Together, these findings provide, for the first time to our knowledge,
electrophysiological evidence from intracranial recordings in humans for the role of
medial temporal regions in learning-related modulation of the neural representation of
single events within sequentially-organized knowledge.
50
Article submitted to NeuroImage
Title: Decoding sequence learning from single-trial intracranial EEG activation
Article Type: Regular Article
Section/Category: Cognitive Neuroscience
Irina Constantinescu (1,2)*, Marzia De Lucia (3)*, Virginie Sterpenich (1,2), Gilles
Pourtois (4), Margitta Seeck (5), Sophie Schwartz (1,2,6)
*Contributed equally to this work
(1) Department of Neuroscience, University of Geneva, Switzerland
(2) Geneva Neuroscience Center, University of Geneva, Switzerland
(3) EEG Brain Mapping Core, Center for Biomedical Imaging of Geneva
and Lausanne, Switzerland
(4) Department of Experimental clinical and health psychology, University of Ghent,
Belgium
(5) Department of Clinical Neurology, Geneva University Hospitals, Switzerland
(6) Swiss Center for Affective Sciences, University of Geneva, Switzerland
Corresponding author:
Sophie Schwartz
Department of Neuroscience, University Medical Center
Michel-Servet 1, 1211 Geneva 4, Switzerland
Phone: +41 22 3795376 Fax: +41 22 379 5402
Short title: Decoding learning-related neural responses
Keywords: intracranial EEG; depth electrodes; local field potentials; single-trial
classification; multivariate decoding; Hidden Markov Model; serial reaction time task;
human memory; learning; hippocampus
51
ABSTRACT
Continuous intracranial EEG (iEEG) data were acquired during two sessions (one
before and one after a night of sleep) in two patients with depth electrodes implanted
in several brain areas, while they performed a deterministic visuomotor sequence
(serial reaction time task, SRTT). To identify brain areas implicated in the learning of
temporally-structured information, we implemented a multivariate decoding algorithm
at the single trial level of iEEG. The algorithm is based on a Hidden Markov Model
(HMM) that captures spatio-temporal properties by identifying series of discrete states
of stable voltage configuration. Our results show that the decoding algorithm correctly
classifies single iEEG trials from the trained sequence as belonging to either the initial
training phase (day 1, before sleep) or a later consolidated phase (day 2, after sleep),
whereas it failed to do so for trials belonging to a control condition (pseudo-random
sequence). Accurate single-trial classification was achieved by taking advantage of
distributed pattern of neural activity. However, across all the contacts, only the
hippocampus provided significant contribution to the classification accuracy for both
patients, and one fronto-striatal region for one of the patients. We also show that
learning increased the temporal stability of discrete neural states. Together, these
human intracranial findings demonstrate that a multivariate decoding approach
captures learning-related changes at the level of single-trial iEEG, and provide new
evidence for the involvement of medial temporal (as well as fronto-striatal) regions in
the learning of single visuomotor events within sequentially-organized knowledge.
1. INTRODUCTION
A large variety of our daily life activities are organized into sequences of
movements, whose precise and fluent execution depends on practice (e.g. talking,
getting dressed, riding a bike, using a computer mouse, typing). Finger tapping tasks
and serial reaction time tasks (SRTT) have proven to be extremely useful to
investigate skills learning (e.g. Fischer et al., 2002; Karni et al., 1995; Pascual-Leone
et al., 1999; Peigneux et al., 2003; Walker et al., 2002). In these tasks, subjects are
required to perform rapid sequences of finger movements, much like when playing the
piano. Sequence learning typically results in a better performance for trained
sequences of movements compared to non-trained or random sequences. Importantly,
52
performance improvement may develop over time after practice has ended,
particularly during sleep, suggesting an „off-line‟ process of consolidation
(Diekelmann and Born, 2010; Diekelmann et al., 2009; Doyon et al., 2009; Hotermans
et al., 2006; Korman et al., 2003; Maquet, 2001; Press et al., 2005; Stickgold et al.,
2001). Strong support for off-line memory reprocessing comes from cell-recording
data in rodents showing that temporally-structured hippocampal activity recorded
during spatial navigation may be replayed during subsequent periods of sleep (Euston
et al., 2007; Louie and Wilson, 2001; Wilson and McNaughton, 1994; see also Dave
and Margoliash, 2000) and wakefulness (Foster and Wilson, 2006). However, while a
number of functional MRI (fMRI) and scalp EEG studies investigated sequence
learning in humans, the implication of the human hippocampus in the consolidation of
freshly learned sequences is still debated (Albouy et al., 2008; Fletcher et al., 2005;
Hazeltine et al., 1997; Schendan et al., 2003; Seidler et al., 2005).
The goal of the present study is twofold: (1) to test the implication of medial
temporal lobe (MTL) regions in visuomotor sequence learning by using human
intracranial EEG recorded from several brain regions including the hippocampus; and
(2), at a methodological 4 level, to demonstrate the feasibility and advantages of using
a multivariate decoding approach to iEEG data collected in individual patients. Here,
we recorded iEEG from distributed cerebral locations in two epileptic patients to
identify selective and spatially-localized changes in neural activity related to
visuomotor learning. Critically, depth-electrode recordings may allow the direct
evaluation of hippocampus contribution to sequence learning. Other techniques in
human studies are either constrained by limited access to deep regions (scalp EEG) or
lack sufficient temporal resolution to capture variations at the level of singleresponse
event-related activity within an SRT sequence (fMRI). Thus, while iEEG recordings
in humans remain uncommon, they provide information about changes in neural
activity at high temporal and spatial resolution, therefore bridging the gap between
animal cell-recording data and human neuroimaging findings.
To investigate the neural representation of single events (trials) within a trained
sequence, we developed a multivariate decoding algorithm that first classifies
learning-related changes in neural activity by exploiting distributed patterns of neural
activity. We then estimate which electrodes mostly contribute to the discrimination
power in order to identify brain regions implicated in learning-related changes. This
53
data-driven approach is challenging because it aims at extracting distinctive features
of neural activity during a complex cognitive paradigm at the single-trial level and
with minimal a priori constraints (for similar approaches, see De Lucia et al., 2007a,
b; Murray et al., 2009; Tzovara et al., submitted). This effort is counterbalanced by the
benefit of a statistical assessment of datasets from single individuals, which is
particularly appropriate for the study of clinical cases with heterogeneous anatomical
and functional characteristics.
2. MATERIAL AND METHODS
2.1. Patients
We tested epileptic patients with depth-electrodes implanted in several brain
regions for presurgical evaluation purposes. Here we report the findings in two
patients, patient M.R. (male, left-handed, aged 35) and patient C.S. (female, righthanded, aged 25) who completed the experimental protocol without any major
epileptic activity over the whole period of recording (about 18 hours over 2 days, see
Figure 1a). During the whole experimental protocol, including one night of sleep, both
patients were free of any medication. The ictal semiology consisted of generalized
seizures alternating with partial seizures for patient M.R and of complex partial
seizures for patient C.S. None of the patients had any detectable hippocampal damage
(including sclerosis): the magnetic resonance imaging (MRI) exam was normal for
patient M.R and showed bilateral occipital periventricular heterotopia without any
hippocampal abnormality for patient C.S. Both patients had a good general clinical
status, no cognitive impairment, and no history of sleep disorder. None of the patients
had formerly exerted any activity that could influence performance on the main
experimental task (SRTT, see below), such as playing a musical instrument or
extensive practice with typing on a keyboard. The patients could execute the task
without difficulty (see Results). Both patients judged the quality of the night sleep
during the experiment as good (St Mary‟s Hospital Sleep Questionnaire and verbal
interview) and polysomnographic sleep scoring showed good sleep efficiency (see
Supplementary Material). Both patients provided written informed consent to
participate to this study, which was approved by the ethical committee of the Geneva
University Hospital.
54
2.2. Behavioral task and experimental procedure
The patients were tested on a serial reaction time task (SRTT). During testing,
the patients faced a computer screen (19 inches) displaying four grey rectangles (2.15
x 4.30 degrees of visual angle, each) arranged horizontally next to each other (2.15°
gaps) on a homogeneous, light blue background, and a fixation cross at the center of
screen (Figure 1b). The four rectangles and the fixation cross remained on the screen
throughout the experiment. On each trial, one of the four grey rectangles briefly
changed its color to yellow during 200 ms. The patients had to react as quickly and
accurately as possible by pressing on a response-pad the spatially corresponding key
to the flashed position within a 1000 ms interval. The patients responded with the four
fingers of their non-dominant hand (right for M.R. and left for C.S.). Whenever they
did not press the correct key or pressed no key within the response interval, an error
feedback was displayed (background color turning to dark blue during 200 ms). The
next trial started 500 ms after the end of the response period. The inter-stimulus
interval was fixed and set to 1700 ms. On each trial in the sequence, reaction times
(RT) and response accuracy were recorded. The SRTT was programmed using the EPrime software (Psychology Software Tools, Inc., Pittsburg, PA) to allow high
temporal precision for stimulus presentation, RT data collection, and marking of the
events on the iEEG recording.
Two types of sequences were presented: a structured (S) sequence, requiring
eight keypresses in a specific pre-determined order and a control (C) pseudo-random
sequence. For both patients, the S-sequence used was 2-4-3-1-4-2-1-3, where 1 refers
to the left-most rectangle and 4 to the right-most rectangle on the screen. The Csequence consisted of series of eight keypresses (as for the S-sequence) formed by
randomly shuffling two sub-sequences of 1-2-3-4 positions. Hence, both sequences
differed in the succession of single positions, but the C-sequence shared some
common features with the S-sequence because the four distinct positions were flashed
once in each of the two quartets forming the 8-position sequence. A 7 short pause of
2500 ms was inserted after the last trial of each sequence, i.e. after 8 keypresses. The
patients were not informed about the regularity of the S-sequence.
The patients were tested during two sessions on two successive days, including
one night of sleep between the sessions (Figure 1a). On day 1, a training session
55
started at 7.45 p.m. and lasted about 30 min. On day 2, a test session started at 9.30
a.m. and lasted 30 min. Each session consisted in three blocks of the S-sequence (20
repetitions each) and one equivalent block of the C-sequence. The C-sequence was
introduced in both sessions to allow the assessment of sequence-specific changes in
behavior and neural activity, independent of non-specific improvements in executing
the visuomotor task. The C-sequence was performed after two S-blocks to ensure that
the patients had acquired good practice with the SRT task. The C-block was also
followed by a last S-block to limit interference of the C-sequence on the consolidation
of the S-sequence. To obtain information about explicit knowledge of the S-sequence,
we asked the patients at the end of the second session on day 2 whether they had
noticed any regularity regarding the succession of the positions (or keypresses). We
also invited them to reproduce the motor sequence on the keypad.
2.3. Intracranial EEG recording
We tested both patients while they underwent continuous long-term intracranial
EEG (iEEG) recordings. Electrophysiological activity was recorded over arrays of
depth electrodes surgically implanted to identify the epilepsy focus. Intracranial EEG
was recorded (Ceegraph XL, Biologic System Corps.) using electrode arrays with 8
stainless contacts each (AD-Tech, electrode diameter: 3 mm, inter-contact spacing: 10
mm), orthogonally implanted in several brain regions. Patient M.R. had 64 contacts (8
arrays), covering, bilaterally, the anterior and posterior hippocampal regions, the
amygdala, and the fronto-orbital cortex. For patient C.S., a total of 88 contacts (11
arrays) covered anterior and posterior hippocampal regions, the amygdala, the frontalcaudate area, and the occipital cortex, bilaterally, as well as the right fronto-orbital
cortex (Figure 2a). For each patient, precise electrode location was determined by the
co-registration of a post-operative computed tomography scan (CT) with a highresolution anatomical MRI. For the iEEG recordings, the reference in both patients
was a scalp electrode, subcutaneously implanted, located at position Cz and the
ground another scalp electrode at position FCz of the 10-20 international EEG system.
The iEEG signal was sampled at 512 Hz and band-pass filtered between 0.1-200 Hz.
DC correction and a 50 Hz notch filter were applied to the data. Intracranial EEG was
recorded continuously during the training period on day 1 and testing period on day 2,
56
as well as during the night between the two sessions. Electrooculography using two
electrodes placed on the eyes‟ external canthi and electromyography using chin
muscle electrodes were also performed during the sleep periods. Sleep data were
scored manually by two trained sleep scorers on 30s epochs based on Rechtschaffen
and Kales standard scoring criteria (Iber et al., 2007).
2.4. Data analysis
We first analyzed iEEG data from all implanted regions using a classification
method based on single-trial responses. In short, this approach aims at classifying
single trials of iEEG (epochs corresponding to single keypresses) in the S-sequence as
belonging to day 1 or to day 2. We then performed a localization procedure to detect
which areas contributed significantly to the classification accuracy. Additionally, we
computed event-related potentials of iEEG data from the areas that mostly contributed
to the classification to test for possible amplitude-related modulations on the averaged
signal. We also used this amplitude modulation value for classifying single-trials
belonging to day 1 and day 2. We performed this test to check whether amplitude
modulation was sufficient to obtain significant classification accuracy.
2.4.1. Multivariate decoding approach to iEEG
We implemented a multivariate approach to investigate the consolidation of a
sequential motor skill based on distributed brain signals. This method presents several
advantages for the analysis of intracranial recordings with respect to classic average
event-related potentials. First, it exploits single-trial measurement of different duration
therefore accounting for possible latency shifts in the electrophysiological response to
similar stimuli and that are not time-locked to the visual cue or motor response.
Second, it provides classification estimations for different conditions, whose accuracy
can be tested statistically. Finally, by a recurrent procedure it allows to assess the
contribution of each measurement site to the classification accuracy. The distinctive
features of this method are particularly appropriate for the study of single-patients
datasets, whose measurement quality can be affected to a different degree and at
different site locations, making it difficult to statistically assess group-level effects.
Moreover, in the context of intracranial data, the averaging of iEEG epochs rarely
57
generates classical ERP components, thus preventing a straightforward use of the
knowledge derived from scalp EEG studies. While it provides a stringent test for the
classification of single-trial data together with an identification of anatomical sites for
conditions-dependent changes, this method takes advantage of several single-trial
features simultaneously, including signal amplitude and spatio-temporal pattern of
activity (see Hidden Markov Model of single-trial iEEG section for details).
As described in more details below, the main steps for the multivariate decoding
approach are: (i) the HHM model of single-trial iEEG, (ii) the accuracy estimation,
and (iii) the localization estimation.
Hidden Markov Model of single-trial iEEG
To model the neural response in the S-sequence, we used a selection of contacts
for each of the two patients, choosing the most distant contacts on each electrode
array. This data selection procedure was motivated by the need to reduce the
computational time in training the model and is justified by the observation that close
contacts measure highly correlated brain activity. The number of selected contacts N
was 16 for M.R. and 22 for C.S. (for Talairach coordinates of these contacts, see
Supplementary Table). To achieve faster RTs, motor learning should predominantly
modulate the electrical activity preceding the keypress. We therefore defined each
individual trial (corresponding to iEEG epochs) as starting from the fixation cross and
ending when the keypress was recorded. Because we aimed at characterizing neural
responses when the patients performed the S-sequence, irrespective of the specific
finger used, we pooled all the trials from the S-sequence in separate datasets, one for
each of the two days. This approach of pooling together all the events corresponding
to different fingers is also essential for allowing the use of the same model when
classifying single trials of the C-sequence. Obviously, the model it is not informed by
any feature of the sequence order.
For each patient, we therefore created two datasets one for day 1 and one for day
2 including all the single trials (epochs) from the S-sequence. Each of these epochs
comprises a set of voltage measurements Mtf at each time-frame tf, with Mtf = [v1(tf),
v2(tf),…,vN(tf)] (Figure 2b,c). The aim of the HMM algorithm is to model the time
series of vectors Mtf within each epoch {M1, M2,…, ML}, where L is the length of one
epoch.
58
In an HMM, one considers a probabilistic model of two sets of random variables
{H1,…, HQ} and {M1, M2,…, ML} (Rabiner, 1989). The variables {H1,…, HQ}, called
hidden states, represent a stochastic process whose evolution over time cannot be
observed directly. The property of this process can only be inferred through the
realizations of the variables {M1, M2,…,ML}, that is to say, in our case, the EEG
signal recorded from several electrodes. The HMM is characterized by the initial state
probability vector π of elements πi = P(H1 = i); the state transition matrix A, with aij =
P(Htf = i|Htf−1 = j) and the set of emission probability density function B = {bi(Mtf) =
p(Mtf|Htf = i)}, which is, in our case, modeled by a Gaussian mixture (Gaussian
mixture model; GMM).
The first step of the algorithm is a GMM estimation of the set of voltage
measurement for a given condition. Thus, the N-dimensional vectors {Mtf}, representing
the instantaneous voltage measurement across the entire set of electrodes (Figure 2b),
were extracted and pooled together regardless of the timing at which they were observed
and of the trial to which they belonged (Figure 2c). The GMM estimation itself was
initialized by a K-means clustering method (Bishop, 1995), which provides a first guess
of the mean of each cluster, the prior probability for each of the clusters, and associated
covariance. The latter were calculated considering, for each of the Q Gaussians, the set of
vectors {Mtf} closest (on the basis of Euclidean distance) to each mean. The GMM
parameters were estimated by an expectation-maximization algorithm for mixture of
Gaussians (Dempster et al., 1977). The resulting GMM parameters were then used to
initialize an HMM model with Q hidden states. This second step allows characterizing the
temporal structure of the response as a series of states. The HMM model was estimated
using Baum-Welch algorithm (Baum et al., 1970), a machine learning standard procedure
which compute the values of π, A and B defined above. Both steps in the model
estimation were computed using a toolbox for Matlab developed by Kevin Murphy
(http://www.cs.ubc.ca/~murphyk/Software/HMM/hmm.html).
In order to select the number of clusters Q providing the best classification
between the trials on day 1 and day 2, we considered a range of possible values of Q
between 3 and 8. We therefore obtained six HMM (one model for each of the Q
values considered) for each of the two datasets and each of the patients.
59
2.4.2. Accuracy estimation
To select which HMM provided the best discrimination power between day 1
and day 2, we validated the models across ten splits of the data. For each split, nine
tenth of the data extracted from the S-sequence were used for training the models, the
remaining trials were used for testing. For each test, we computed a Receiver
Operating Characteristic (ROC), whose underlying area provides an unbiased measure
of the classification accuracy (Swets et al., 1979). After selecting the parameters
corresponding to the largest area lying beneath the ROC curve, we tested the model on
the C-sequence. A significantly lower performance in classifying trials belonging to
the C-sequence would confirm that the classification performance between day 1 and
day 2 for the S-sequence was attributable to sequence-specific effects related to
learning and neural plasticity, and not to non-specific task adaptation effects (e.g.
more efficient visuomotor mapping with practice) nor to general differences in
measurement conditions (e.g. different states of the patient). For testing the model on
the C-sequence, we considered ten splits of the whole set of single trials and ran a test
for each split. It is worth noting that the attempt to classify data belonging to the Csequence is a valid test because the model estimation was based on single trials from
the S-sequence on day 1 and day 2, disregarding any information about the grammar
of the sequence itself. As such, this test is essential for claiming that the algorithm
captures neural changes between the day 1 and day 2 pertaining to the learning of the
sequence, rather than reflecting other learning factors such as an improvement in the
mapping between visual cues and response keys (which would also affect the control
sequence), or non-specific effects including motivation or fatigue.
As an additional estimation of the quality of the trained model, we assessed the
accuracy obtained for new test data by measuring the ratio of trials correctly classified
and the total number of trials when keeping the same priors as those obtained during
the training. Finally, to exclude that the discrimination power reached when
classifying the S-sequence could trivially be due to shorter trial duration on day 2
compared to day 1, we performed a control test by classifying the two datasets based
on the reaction times (RTs) only (i.e. based on the single-trial duration).
ROC curve areas based on the ten estimations in the cross validation procedure
were compared using t-tests (p<0.01). To further assess any possible relation between
60
the estimated model accuracy and the RTs, we computed the Pearson correlation
between the difference of the logarithm of the likelihoods of each of the two models
(hereafter called discrimination function) and the RTs on a trial-by-trial basis in the
test dataset.
2.4.3. Estimating contacts with higher classification power
To test for the role of specific brain areas in the classification performance, we
estimated the algorithm accuracy after dropping couple of contacts belonging to each
electrode array, one at a time, and for each patient separately. At this stage, each
couple of electrodes is considered together because they can reflect correlated activity
along the strip and in order to shorten the computational time. For each of this
electrodes selection, we estimated the parameters providing the highest classification
accuracy for the S-sequence as we did above. We then compared the resulting
classification performance with that obtained previously when keeping all the
contacts. The same analysis was carried out with the data from the C-sequence based
on the models estimated in the structure sequence.
A further analysis was carried out focusing on those couple of electrodes that
produced a significant drop of the classification accuracy. In the following we will
refer to the string of contacts including this couple of electrodes as target string. In
this further test, we trained two set of new models for day1 and day2 taking into
account the same number of initially chosen selected contacts (i.e. 16 for M.R. and 22
for C.S.) but replacing one contact within the target string with another contact in the
same string. The goal is to test how robust is the drop in classification accuracy
against different choice of the electrodes along the strip and the critical contribution of
each single contact while keeping the same number of total electrodes.
2.4.4. Event-related potentials
To characterize the average time-course modulation of the electrical activity in the
areas found relevant by the single-trial classification approach, we computed intracranial
ERPs (iERPs) for the contacts along the strips corresponding to these areas: for patient
C.S., the right anterior hippocampus and right frontal-caudate region; for patient M.R.,
the left posterior hippocampus (Figure 5). Intracranial data were preprocessed and
61
analyzed using the Cartool software (http://brainmapping.unige.ch/Cartool.htm).
Individual epochs were timelocked to the presentation of the visual cue, as was done
for the single-trial approach above. The duration of the epochs was 1000 ms, with a
200 ms baseline. Single-trial EEG epochs were analyzed offline, after removing all
epochs including co-occurring artifacts or epileptic spikes (16.7% of all epochs
removed for patient C.S. and 32.2% of all epochs removed for patient M.R., using
stringent visual inspection). Note that this percentage of rejected epochs is acceptable
when considering similar procedures in the human intracranial literature (Pourtois et
al., 2007; Pourtois et al., 2010a; Pourtois et al., 2010b). We also excluded epochs
corresponding to errors (wrong key pressed or misses; less than 2.25% in both
patients). These preprocessing steps resulted in a total of 821 trials for the S-sequence
(390 trials for day 1 and 431 trials for day 2) and 244 trials for the C-sequence (128
trials for day 1 and 116 trials for day 2) in patient C.S, and in 599 trials for the Ssequence (324 trials for day 1 and 275 trials for day 2) and 237 trials for the Csequence (128 trials for day 1 and 109 trials for day 2) in patient M.R. Individual
epochs were low-pass filtered using a 30 Hz cutoff.
Four different iERPs were computed, corresponding to the S-sequence and
C-sequence for day 1 and day 2, for the considered contacts. The amplitude of the
electrical signal at each time-point across all the trials was used as the dependent
variable for within-subject statistical comparisons We tested for any modulation of
response amplitude between day 1 and day 2 separately for each sequence type using
non-parametric statistical analyses based on randomization tests (e.g. Pourtois et al.,
2007). The randomization method tests for differences in any variable (here amplitude
at each time-point) without any assumption regarding data distribution, by comparing
the observed dataset with random shufflings of the same values over many iterations.
Here, the randomization implied an iterative shuffling of 10‟000 times, which
provides a robust estimation of the probability that the amplitude difference at any
time-point is observed by chance. We report effects that survived a threshold of
p<0.05 with a minimal temporal stability of at least 10 consecutive time-points (>20
ms at 512 Hz sampling rate; Guthrie and Buchwald, 1991; Pourtois et al., 2008).
62
3. RESULTS
3.1. Behavioral results
For each patient, we first averaged the reaction times (RT) over the 8 keypresses
of each sequence in each block (i.e. 20 sequences in each block). Note that only hits
were included in the analysis and that the patients made very few errors (incorrect key
pressed or misses) in each block (% of all keypresses ±SD; Patient M.R.: 2.31 ±2.27;
Patient C.S: 0.50 ±1.07). The data were then subjected to an ANOVA with Day (day
1, day 2) and Blocks (3 blocks for the S-sequence, 1 block for the C-sequence) as
factors. Additional t-tests were performed whenever interactions were significant.
Both patients showed a main effect of Day (patient M.R.: F(1,152)=4.89, p<0.05;
patient C.S.: F(1,152)=15.72, p<0.001) due to the RTs being faster on the second than
first day (mean ±SD; M.R. Day 1: 597.11 ms ±66.26, Day 2: 575.10 ms ±66.01,
T(158)=2.11, p<0.05; C.S. Day 1: 400.76 ms ±38.28, Day 2: 379.30 ms ± 41.65,
T(158)=3.38, p<0.001). Critically, the analysis revealed a Day by Block interaction
(M.R.: F(3,152)=3.62, p<0.05; C.S.: F(3,152)=2.72, p<0.05) due to greater
performance improvement for the S-sequence blocks than the C-sequence blocks on
day 2 (RT difference between day 1 versus day 2 for M.R. S-sequence: 24.35 ms,
T(118)=2.21, p<0.05, C-sequence: 14.99 ms, T(38)=0.72, p=0.49; for C.S.
S-sequence: 28.91 ms, T(118)=5.16, p<0.001, C-sequence: -0.90 ms, T(38)=-0.07,
p=0.95). Consistent with previous studies (Albouy et al., 2006; Destrebecqz et al., 2005)
none of the patient exhibited any explicit knowledge of the S-sequence or of three
successive positions (i.e. no fragmentary rule knowledge). They reported that they did
not notice any regularity in the succession of the yellow rectangles or keypresses.
3.2. Results from the multivariate decoding
approach to single-trial iEEG
3.2.1. Accuracy estimation
For patient M.R., the best classification performance was obtained for Q=3 on
day 1 and Q=4 on day 2. The ROC area was 0.80 ±0.07 (mean ±SD; Figure 3a). For
patient C.S., the selected parameters were Q=6 and Q=7 for day 1 and day 2,
respectively. The average ROC curve area was 0.90 ±0.06 (Figure 3b). The accuracy
63
obtained when keeping the same priors as those obtained in the training, measured as
the ratio of the number of trials correctly classified to the total number of trials, was
0.70 ±0.06 and 0.81 ±0.05 respectively. In both patients, the accuracy for classifying
the S-sequence was significantly higher than that for classifying the control sequence
(p<0.01), because the average ROC curve area for the C-sequence was 0.66 ±0.04 and
0.66 ±0.04 for day 1 and day 2, respectively. The corresponding accuracy when
keeping the same bias as in the training was 0.60 ±0.06 and 0.63 ±0.04. When
classifying the trials in the S-sequence based on their reaction times only, we obtained
a classification accuracy considerably lower than that for the S- and C-sequences in
both patients (Figure 3). In addition, the correlation between the discrimination
function and the RTs evaluated on each of the ten test datasets was also negligible and
never exceeded 0.37 (n.s.).
3.2.2. Estimating contacts with higher classification power
When leaving out in turn each electrode array (2 contacts each), the accuracy for
classifying the S-sequence dropped significantly (p<0.01) for one pair in patient MR
and for two pairs in patient CS. For patient M.R., the critical array corresponded to
left posterior hippocampus (Figure 4a and Figure 5a). For patient C.S., accuracy
decreased when dropping arrays including the right anterior hippocampus and the
right frontal-caudate array (Figure 4b and Figure 5b,c). For none of the other arrays,
we obtained a significant drop in the classification accuracy. We report the results
using model parameters that provided the highest classification performance for the Ssequence. Importantly, for none of the parameters choice did we obtain higher
classification accuracy when dropping any of these areas. This last result provides a
quality check on the contribution from each electrode included in the present analysis,
because dropping an electrode with noisy signal might improve the classification
accuracy. When applying the same procedure on the data from the C-sequence, we
found that, in both patients, many areas were associated with a significant drop in the
accuracy when they were not included in the model estimation (Figure 4a,b) but,
critically, the accuracy for the S-sequence remained always significantly higher than
that of the C-sequence. Again, as for the S-sequence, for none of the parameters
choice was any increase in the discrimination power observed. All together these
64
results show that in both patients the electrode including the hippocampus was
critically contributing to the classification accuracy.
We further analyzed how specific was the contribution of the hippocampus by
estimating two new sets of models for day 1 and day 2 by considering all the initially
chosen contacts (i.e. 16 for M.R. and 22 for C.S.) but replacing the one in the
hippocampus (LPH1 and RAH1 for MR and CS respectively) with one external to the
hippocampus within the same target string. For patient MR, the ROC area dropped to
0.78±0.07 which was significantly lower than the area obtained using all the initially
chosen contacts (p<0.05). For patient CS, the ROC curve area dropped to 0.88±0.04,
significantly lower than when including hippocampus (p<0.05).
We can conclude that for both patients within the string that included
hippocampus, the contact placed in the hippocampus was contributing more to the
classification performance other that the other contact. Note also that these effects
were found for hippocampal contacts in the hemisphere contralateral to the executing
hand in both patients.
3.3.3. Relation between classification accuracy
and spatio-temporal pattern of single-trial events
We investigated which HMM models‟ parameters were most informative in
classifying single-trials from day 1 and day 2 and what signal properties they
reflected. To this aim, we quantified the contribution to the classification accuracy of
each of the parameters that provided the best classification accuracy for the two HMM
models (i.e. priors, means, covariances, and transition probabilities). This was done by
swapping the values of each of these parameters between the two models and
repeating the testing as explained in the section „Accuracy estimation‟. We found that
the values of the covariances were the most informative for discriminating day 1 and
day 2 and that none of the other parameters were significantly impacting the
classification accuracy.
The next step was to gain a better insight on the neurophysiological interpretability
of the HMM models‟ parameters captured during the day 1 and day 2. As mentioned
above (section „Hidden Markov Model of single-trial iEEG‟), the HMM model can be
used to segment the single-trials into series of momentary states, that is to say
momentary states of stable voltage configurations. We therefore investigated how
65
differences in the covariance matrices (which were found above to play a significant
role in the classification) may be reflected in the spatio-temporal pattern of these
momentary states. We found that the average duration of these states was longer
during day 2 and in comparison to day 1 only for those single-trials belonging to the
C-sequence in both patients. The average duration of these momentary states for
subject MR was 97.1 ms and 110.8 ms during day 1 and day 2 respectively and
significantly differed (p<0.05; Fig. S1a). The average duration for subject CS was
62.7 ms and 71.1 ms during day 1 and day 2 (significantly different, p<0.05; Fig.
S1b). Importantly, this difference in the average duration was no longer significant
when considering single trials belonging to the random sequence in both subjects
(during day 1 and day 2 average duration was 74.9 ms and 75.3 ms for MR; 60.1 ms
and 64.3 ms for CS). We concluded that the pattern of momentary states and their
typical temporal duration was a function of the learning stage of the subject. In
addition, when omitting the hippocampus contribution, for patient MR the difference
in the average duration for the S sequence was not anymore significant (60.8 ms and
63.9 ms during day 1 and day 2 respectively; p>0.05), and for patient CS, although
this difference was still significant, it was significantly lower than when including all
the contacts (58.3 ms and 61.7 ms during day 1 and day 2 respectively; p>0.05).
3.3. Intracranial event-related potentials
We report the ERP results for these two representative contacts, for the strips
that contributed significantly to the classification accuracy in the single-trial approach
(Figure 6).
For the S-sequence, a period of significant amplitude difference between day 1
and day 2 was observed, in both patients. For patient M.R., the amplitude modulation
started 293 ms after the visual cue onset (i.e. 211 ms before the average keypress
time) and lasted for 43 ms (Figure 6a). For patient C.S., this amplitude modulation
started 186 ms after the visual cue onset (i.e. 219 ms before the average keypress) and
lasted for 61 ms (Figure 6b). Overall, the iERPs in both patients shared some general
shape characteristics with an amplitude peak at around 400 ms followed by a steep
amplitude decrease up to 500-600 ms.
For the contacts of the frontal-caudate strip present in patient C.S. only, we found
a similar amplitude effect, which started just before the average keypress (72 ms before
66
keypress; 332 ms after the visual cue onset) and lasted until the end of the epoch. Figure
7 shows the ERP modulation for the contact in the medial part of the caudate
For both patients and for all electrodes of interest (e.g. the hippocampal contacts
in patients C.S. and M.R. and the frontal-caudate contact in patient C.S.), iERP
amplitude for the C-sequence did not differ between day 1 and day 2 at any time
during the whole epoch of interest (Figures 6 and 7).
Finally, we tested whether the amplitude modulation at the hippocampal sites
could be used for classifying single-trials as belonging to day 1 or day 2. This test
provides a lower bound for the classification accuracy that can be obtained based on
time-locked signal. We computed the average amplitude of the single-trials along the
temporal periods where the significant amplitude modulation was observed at
average-level. We extracted these averages for each day and each single trial from the
hippocampal sites. We then computed the ROC curve based on these average
amplitudes, separately for each patient (as we did for the 21 analysis based on the
reaction time in Figure 3). We found that the ROC underlying area was 0.53 for
subject CS and 0.54 for subject MR. This additional analysis confirms that timelocked
amplitude information does not allow accurate classification of single trials, even
when selecting the most likely informative portion of the ERP (which also does not
guarantee non-circularity), and thus cannot be used to generate statistically robust
inferences, unlike the GMM-based multivariate decoding of raw single-epochs.
4. DISCUSSION
The present study provides evidence of learning-related changes of intracranial
electrical activity corresponding to single events (individual keypresses) belonging to a
deterministic visuomotor sequence. Specifically, we show that a multivariate
classification algorithm can accurately assign single-trial iEEG responses to either the
initial training phase (day 1) or to the later consolidated phase (day 2), i.e. after
sequence-specific improvement occurred. We first demonstrate that the classification
results relied on neural features selectively associated to the learning of the regularity of
the 8-item sequence, as evidenced by the comparatively low classification accuracy
obtained for single-trials from a random sequence. We then found that the classification
failed when applied to the single-trial reaction-times, thus indicating that the neuralrelated changes captured by the decoding algorithm cannot be trivially explained in
67
terms of speeding of motor response. Finally, while the localization of these learningrelated changes involved a distributed pattern of neural activity, we identified the
hippocampus as contributing critically to the reliable classification of the trained
sequence – independently in each patient, with an additional contribution from
frontostriatal regions for the patient implanted with electrodes in these brain areas.
4.1. Learning-related changes in the stability
of momentary neural states
The multivariate decoding approach implemented here offers new insights into
spatio-temporal patterns of distributed activity and how they change with learning. In
particular, we show that each single trial can be represented as a sequence of
momentary states, whose corresponding durations may change as a function of
learning. This change in duration was observed only for trials belonging to the
S-sequence, and critically related to the presence of the hippocampus in both subjects.
The existence of short periods of stability of voltage configurations (often referred to
as functional microstates) is a well established empirical observation at the level of the
scalp electroencephalography (Lefevre and Baillet, 2009; Lehmann, 1987; Michel et
al., 2004; Michel et al., 2001). Indeed, their spatio-temporal properties have been
extensively used to analyze both event-related potentials (De Lucia et al., 2010;
Murray et al., 2006) as well as resting state EEG (Britz et al., 2010). Our finding
demonstrates that the statistical properties of distributed voltage configurations carry
the signature of functional-related changes also at the level of intracranial responses
and that it is detectable by a multivariate decoding approach at the single-trial level.
The significant change in average duration of these momentary states suggests that the
temporal properties rather than the actual amplitude values of these states are critically
related to learning stage. Indeed, this is also confirmed by the poor classification
accuracy obtained when considering amplitude modulation only. At the level of scalp
EEG, temporal duration of microstates has been reported to be linked to functional
deficits in specific clinical populations (Kindler et al., 2010; Lehmann et al., 2005;
Strik et al., 1997). Very recently, it has been shown that resting state in humans is
characterized by a multi-fractal organization of miscrostates sequence and that its
degree of complexity is related to the temporal duration of microstates but not to their
actual values (Van de Ville et al., 2010). These studies together with our current
68
findings suggest that temporal scale of momentary states carries crucial information
about the subject‟s cognitive states emphasizing the importance of investigating the
dynamics of brain activity together with its spatial distribution.
4.2. Role of the hippocampus in sequence learning
Using a multivariate classification approach, we found that for both patients the
hippocampus provided a critical contribution to the classification accuracy, which
relied on the temporal stability of underlying neural states (average state duration is
increased after learning for both patients). Increasing the stability of neural states after
learning may reflect a major function of the hippocampus in associative processes,
including the integration of temporally- or spatially-organized information (Agster et
al., 2002; Degonda et al., 2005; Ergorul and Eichenbaum, 2004; Fortin et al., 2002;
Moscovitch et al., 2006; Nadel et al., 2000; Ryan et al.; Wallenstein et al., 1998). By
implementing a multivariate decoding approach to singletrial iEEG data in patients
with intact hippocampal regions, our study suggests that the learning process
conferred some local „contextual‟ information to individual items (or trials) within the
trained sequence so that the decoding algorithm could accurately discriminate between
hippocampal responses generated at different stages of the learning. Our results
therefore highlight the implication of the MTL in the integration of single events into
structured representations (Moscovitch et al., 2006; Winocur et al., 2007), such as the
integration of constituent parts of motor sequences (Albouy et al., 2008; Eichenbaum,
2004; Levy, 1996; Schendan et al., 2003; Spencer et al., 2006; Wallenstein et al.,
1998). The formation of local contexts or predictions within the sequence supports the
„glueing‟ together of the successive elements in the sequence with repeated exposure
(i.e. after training on the sequence; see Kumaran and Maguire, 2006; Levy et al.,
2005), and may be thus consistent with enhanced temporal stability of neural states as
found here.
For one patient in whom we could measure neural activity along the lateral
prefrontal cortex (PFC) and caudate regions, the classification showed additional
contributions from these regions. Because of its dense interactions with both MTL
structures and the basal ganglia, it has been suggested that the PFC may participate to
the extraction of regularities based on internal representations, so as to improve
behavioral control (Robertson et al., 2001; Shima et al., 2007). In support of a role of
69
hippocampal-prefrontal interactions in memory processes, Peyrache and al. (2009)
recently demonstrated that following the acquisition of new rules, PCF activity
patterns reflected the neural patterns that occurred during the training phase (see also
Euston et al., 2007), and that this reactivation occurred predominantly when
hippocampal and cortico-hippocampal interaction was enhanced (see also, Clemens et
al., 2007; Diekelmann and Born, 2010; Frankland and Bontempi, 2005; Ji and Wilson,
2007; Kali and Dayan, 2004; Qin et al., 1997; Rasch and Born, 2007). While this
could only be observed in one patient, the present finding of a contribution of both the
hippocampus and PFC activity to the classification of learning stages would be
consistent with these findings.
4.3. Multivariate decoding approach to single-trial iEEG data
The present iEEG results do not only support a critical function of the
hippocampalneocortical circuitry in sequence learning, but they significantly extend
previous findings by showing that neural representation of single visuomotor events
within a learned sequence is selectively modulated by training only when embedded in
a structured sequence. Indeed, electrical activity of the single events was accurately
identified by a classification algorithm as belonging to an early versus more
consolidated learning phase, and classification accuracy relied on the contribution of
the hippocampus to increased enhanced temporal stability of voltage configuration.
The present study demonstrates that such multivariate decoding approach to singletrial datasets from individual patients offer an important methodological alternative to
group studies of averaged data. Indeed, grouping iEEG data from different patients
can be extremely challenging because the configuration of electrode placement varies
across individuals and pathological conditions (e.g. epilepsy, Parkinson disease), and
because iEEG data quality is often artefacted at different recording sites. Multivariate
decoding approaches can however capture task-related effects at the single-trial level,
that are not timelocked to the visual cue or motor response, and do not require a fixed
trial length. The restricted availability of such iEEG recordings is an additional factor
to motivate the development of methods to fully exploit individual patient datasets.
We therefore developed a procedure that classifies learning-related changes at the
single-trial level and establishes a causal link between specific electrode contribution
and discrimination power (Haynes and Rees, 2006; Norman et al., 2006; Tzovara et
70
al., submitted). This approach is essentially different from classical inferential analysis
in that it makes use of independent splits of the available data to validate differences
in the temporal and spatial configuration of the signal (Formisano et al., 2008; Haxby
et al., 2001; Staeren et al., 2009). Such pattern information analysis offers a general
solution to the issue of „circularity‟ in neuroimaging studies (Kriegeskorte et al.,
2009), which typically occurs when hypothesis and inferential analyses are not
independent (e.g. selection of few electrodes in ERP studies). Pattern information
analysis has predominantly been developed for fMRI studies while less effort has been
devoted to its application to electrophysiological data. As illustrated by the present
study, one main advantage of this methodological framework is that it does not require
any a priori selection of the relevant neural sites or time-window of activity
measurements. We complemented our approach by a localization procedure and
showed that we can identify the regions contributing most to the performance of the
classifier. Our results demonstrate that this methodological strategy is particularly
adapted to test the effect of learning-related changes in neural responses across
distinct brain regions and for datasets obtained from individual patients.
Finally, because we obtained highly consistent findings in both patients whose
data were analyzed independently and who had heterogeneous electrode
implantations, we think that our study may suggest general conclusions about the role
of the hippocampus in deterministic sequence processing. More specifically,
independently in both patients (i) we could classify single trials based on distributed
activity and only in relation to the structural sequences; (ii) the hippocampus provided
the highest contribution to the classification accuracy; (iii) the learning stage was
paralleled by a difference in the average temporal duration of momentary neural
states. Finally, the behavioral results from both patients concord with previous SRTT
studies in normal subjects (Maquet et al., 2000; Walker et al., 2002), by showing that
motor performance improved selectively for the S-sequence after a delay that included
one night of sleep. These data also demonstrate that sequence-selective motor learning
might occur even after a relatively short training session (when compared to other
studies of motor learning in healthy subjects), confirming that the use of such SRT
tasks is both valid and appropriate for the study of rapid and implicit motor learning in
clinical patients, who may have elevated levels of fatigability.
71
5. CONCLUSIONS
Intracranial electroencephalographic (iEEG) recordings in humans provide a
unique opportunity to assess neural activity at high temporal and spatial resolution, in
particular for deep regions that cannot be easily accessible to surface EEG recordings.
These data can thus provide an intermediate level of observation linking animal cellrecording data and human neuroimaging findings. However, much like cell-recordings
in animals, iEEG studies have a sparse distribution of recordings sites. Our study
provides a new multivariate decoding strategy to optimize the use of such distributed
neural signals by allowing the full analysis of datasets from individual patients at the
single-trial level and by offering an unbiased test for the contribution of individual
electrodes to the observed effects. We also show here that this approach may also
constitute an independent first step enabling an unbiased selection of recording sites of
interest for subsequent ERP analyses. Finally, the present study may also offer new
perspectives for the application of single-trial classification approaches to EEG-fMRI
whole-brain measurements.
Acknowledgments
We thank Peter Dayan (University College London, UK) for insightful theoretical
and methodological comments. We also thank Laurent Spinelli for technical support and
the patients for participating in the study.
This work was supported by Grants from the Swiss National Science Foundation (K33K1_122518/1 to M.D.L. and 310000-114008 to S.S.).
Reference list
Agster, K.L., Fortin, N.J., Eichenbaum, H., 2002. The hippocampus
disambiguation of overlapping sequences. J Neurosci 22, 5760-5768.
and
Albouy, G., Ruby, P., Phillips, C., Luxen, A., Peigneux, P., Maquet, P., 2006.
Implicit oculomotor sequence learning in humans: Time course of offline processing.
Brain Res 1090, 163-171.
Albouy, G., Sterpenich, V., Balteau, E., Vandewalle, G., Desseilles, M., Dang-Vu,
T., Darsaud, A., Ruby, P., Luppi, P.H., Degueldre, C., Peigneux, P., Luxen, A., Maquet,
P., 2008. Both the hippocampus and striatum are involved in consolidation of motor
sequence memory. Neuron 58, 261-272.
Baum, L.E., Petrie, T., Soules, G., Weiss, N., 1970. A maximization technique
occurring in the statistical analysis of probabilistic functions of Markov chains. Ann. Math.
Statist. 41, 164-171.
Bishop, C.M., 1995. Neural Networks for Pattern Recognition. Oxford University
Press, Oxford, UK.
72
Britz, J., Van De Ville, D., Michel, C.M., 2010. BOLD correlates of EEG topography
reveal rapid resting-state network dynamics. Neuroimage 52, 1162-1170.
Clemens, Z., Molle, M., Eross, L., Barsi, P., Halasz, P., Born, J., 2007. Temporal
coupling of parahippocampal ripples, sleep spindles and slow oscillations in humans.
Brain 130, 2868-2878.
Dave, A.S., Margoliash, D., 2000. Song replay during sleep and computational
rules for sensorimotor vocal learning. Science 290, 812-816.
De Lucia, M., Clarke, S., Murray, M.M., 2010. A temporal hierarchy for conspecific
vocalization discrimination in humans. J Neurosci 30, 11210-11221.
De Lucia, M., Michel, C.M., Clarke, S., Murray, M.M., 2007a. Single-trial
topographic analysis of human EEG: A new 'image' of event-related potentials.
Information Technology Applications in Biomedicine, pp. 95 - 98.
De Lucia, M., Michel, C.M., Clarke, S., Murray, M.M., 2007b. Single subject EEG
analysis based on topographic information. International Journal of Bioelectromagnetism,
168-117.
Degonda, N., Mondadori, C.R., Bosshardt, S., Schmidt, C.F., Boesiger, P., Nitsch,
R.M., Hock, C., Henke, K., 2005. Implicit associative learning engages the hippocampus
and interacts with explicit associative learning. Neuron 46, 505-520.
Dempster, A., Laird, N., Rubin, D., 1977. Maximum likelihood from incomplete data
via the EM algorithm. Journal of the Royal Statistical Society, Series B 39, 1-38.
Destrebecqz, A., Peigneux, P., Laureys, S., Degueldre, C., Del Fiore, G., Aerts, J.,
Luxen, A., Van Der Linden, M., Cleeremans, A., Maquet, P., 2005. The neural correlates
of implicit and explicit sequence learning: Interacting networks revealed by the process
dissociation procedure. Learn Mem 12, 480-490.
Diekelmann, S., Born, J., 2010. The memory function of sleep. Nat Rev Neurosci
11, 114-126.
Diekelmann, S., Wilhelm, I., Born, J., 2009. The whats and whens of sleepdependent memory consolidation. Sleep Med Rev 13, 309-321.
Doyon, J., Korman, M., Morin, A., Dostie, V., Hadj Tahar, A., Benali, H., Karni, A.,
Ungerleider, L.G., Carrier, J., 2009. Contribution of night and day sleep vs. simple
passage of time to the consolidation of motor sequence and visuomotor adaptation
learning. Exp Brain Res 195, 15-26.
Eichenbaum, H., 2004. Hippocampus: cognitive processes
representations that underlie declarative memory. Neuron 44, 109-120.
and
neural
Ergorul, C., Eichenbaum, H., 2004. The hippocampus and memory for "what,"
"where," and "when". Learn Mem 11, 397-405.
Euston, D.R., Tatsuno, M., McNaughton, B.L., 2007. Fast-forward playback of
recent memory sequences in prefrontal cortex during sleep. Science 318, 1147-1150.
Fischer, S., Hallschmid, M., Elsner, A.L., Born, J., 2002. Sleep forms memory for
finger skills. Proc Natl Acad Sci U S A 99, 11987-11991.
Fletcher, P.C., Zafiris, O., Frith, C.D., Honey, R.A., Corlett, P.R., Zilles, K., Fink,
G.R., 2005. On the benefits of not trying: brain activity and connectivity reflecting the
interactions of explicit and implicit sequence learning. Cereb Cortex 15, 1002-1015.
Formisano, E., De Martino, F., Bonte, M., Goebel, R., 2008. "Who" is saying
"what"? Brainbased decoding of human voice and speech. Science 322, 970-973.
Fortin, N.J., Agster, K.L., Eichenbaum, H.B., 2002. Critical role of the hippocampus
in memory for sequences of events. Nat Neurosci 5, 458-462.
Foster, D.J., Wilson, M.A., 2006. Reverse replay of behavioural sequences in
hippocampal place cells during the awake state. Nature 440, 680-683.
Frankland, P.W., Bontempi, B., 2005. The organization of recent and remote
memories. Nat Rev Neurosci 6, 119-130.
Guthrie, D., Buchwald, J.S., 1991. Significance testing of difference potentials.
Psychophysiology 28, 240-244.
73
Haxby, J.V., Gobbini, M.I., Furey, M.L., Ishai, A., Schouten, J.L., Pietrini, P., 2001.
Distributed and overlapping representations of faces and objects in ventral temporal
cortex. Science 293, 2425-2430.
Haynes, J.D., Rees, G., 2006. Decoding mental states from brain activity in
humans. Nat Rev Neurosci 7, 523-534.
Hazeltine, E., Grafton, S.T., Ivry, R., 1997. Attention and stimulus characteristics
determine the locus of motor-sequence encoding. A PET study. Brain 120 (Pt 1), 123-140.
Hotermans, C., Peigneux, P., Maertens de Noordhout, A., Moonen, G., Maquet, P.,
2006. Early boost and slow consolidation in motor skill learning. Learn Mem 13, 580-583.
Iber, C., Ancoli-Israel, S., Chesson, A., Quan, S.F., 2007. The AASM Manual for the
Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications.
American Academy of Sleep Medicine, Westchester, Illinois.
Ji, D., Wilson, M.A., 2007. Coordinated memory replay in the visual cortex and
hippocampus during sleep. Nat Neurosci 10, 100-107.
Kali, S., Dayan, P., 2004. Off-line replay maintains declarative memories in a
model of hippocampal-neocortical interactions. Nat Neurosci 7, 286-294.
Karni, A., Meyer, G., Jezzard, P., Adams, M.M., Turner, R., Ungerleider, L.G., 1995.
Functional MRI evidence for adult motor cortex plasticity during motor skill learning.
Nature 377, 155-158.
Kindler, J., Hubl, D., Strik, W.K., Dierks, T., Koenig, T., 2010. Resting-state EEG in
schizophrenia: Auditory verbal hallucinations are related to shortening of specific
microstates. Clin Neurophysiol.
Korman, M., Raz, N., Flash, T., Karni, A., 2003. Multiple shifts in the representation
of a motor sequence during the acquisition of skilled performance. Proc Natl Acad Sci U S
A 100, 12492-12497.
Kriegeskorte, N., Simmons, W.K., Bellgowan, P.S., Baker, C.I., 2009. Circular analysis
in systems neuroscience: the dangers of double dipping. Nat Neurosci 12, 535-540.
Kumaran, D., Maguire, E.A., 2006. The dynamics of hippocampal activation during
encoding of overlapping sequences. Neuron 49, 617-629.
Lefevre, J., Baillet, S., 2009. Optical flow approaches to the identification of brain
dynamics. Hum Brain Mapp 30, 1887-1897.
Lehmann, D., . Principles of spatial analysis. In: Gevins, A.S., Remond, A., (Eds.)
Handbook of electroencephalography and clinical neurophysiology, Vol.1: Methods of
analysis of brain electrical and magnetic signals (pp. 309-354). Amsterdam: Elsevier,
1987. Principles of spatial analysis. . In: Gevins, A.S., Remond, A. (Eds.), Handbook of
electroencephalography and clinical neurophysiology. Elsevier, Amsterdam, pp. 309-354.
Lehmann, D., Faber, P.L., Galderisi, S., Herrmann, W.M., Kinoshita, T., Koukkou,
M., Mucci, A., Pascual-Marqui, R.D., Saito, N., Wackermann, J., Winterer, G., Koenig, T.,
2005. EEG microstate duration and syntax in acute, medication-naive, first-episode
schizophrenia: a multi-center study. Psychiatry Res 138, 141-156.
Levy, W.B., 1996. A sequence predicting CA3 is a flexible associator that learns
and uses context to solve hippocampal-like tasks. Hippocampus 6, 579-590.
Levy, W.B., Hocking, A.B., Wu, X., 2005. Interpreting hippocampal function as
recoding and forecasting. Neural Netw 18, 1242-1264.
Louie, K., Wilson, M.A., 2001. Temporally structured replay of awake hippocampal
ensemble activity during rapid eye movement sleep. Neuron 29, 145-156.
Maquet, P., 2001. The role of sleep in learning and memory. Science 294, 1048-1052.
Maquet, P., Laureys, S., Peigneux, P., Fuchs, S., Petiau, C., Phillips, C., Aerts, J.,
Del Fiore, G., Degueldre, C., Meulemans, T., Luxen, A., Franck, G., Van Der Linden, M.,
Smith, C., Cleeremans, A., 2000. Experience-dependent changes in cerebral activation
during human REM sleep. Nat Neurosci 3, 831-836.
Michel, C.M., Murray, M.M., Lantz, G., Gonzalez, S., Spinelli, L., Grave de Peralta,
R., 2004. EEG source imaging. Clin Neurophysiol 115, 2195-2222.
74
Michel, C.M., Thut, G., Morand, S., Khateb, A., Pegna, A.J., Grave de Peralta, R.,
Gonzalez, S., Seeck, M., Landis, T., 2001. Electric source imaging of human brain
functions. Brain Res Brain Res Rev 36, 108-118.
Moscovitch, M., Nadel, L., Winocur, G., Gilboa, A., Rosenbaum, R.S., 2006. The
cognitive neuroscience of remote episodic, semantic and spatial memory. Curr Opin
Neurobiol 16, 179-190.
Murray, M.M., Camen, C., Gonzalez Andino, S.L., Bovet, P., Clarke, S., 2006. Rapid
brain discrimination of sounds of objects. J Neurosci 26, 1293-1302.
Murray, M.M., De Lucia, M., Brunet, D., Michel, C.M., 2009. Principles of
Topographic Analyses for Electrical Neuroimaging. In: Handy, T.C. (Ed.), Brain Signal
Analysis: Advances in Neuroelectric and Neuromagnetic Methods. MIT Press, Cambridge,
Mass, pp. 21-54.
Nadel, L., Samsonovich, A., Ryan, L., Moscovitch, M., 2000. Multiple trace theory
of human memory: computational, neuroimaging, and neuropsychological results.
Hippocampus 10, 352-368.
Norman, K.A., Polyn, S.M., Detre, G.J., Haxby, J.V., 2006. Beyond mind-reading:
multivoxel pattern analysis of fMRI data. Trends Cogn Sci 10, 424-430.
Pascual-Leone, A., Tarazona, F., Keenan, J., Tormos, J.M., Hamilton, R., Catala,
M.D., 1999. Transcranial magnetic stimulation and neuroplasticity. Neuropsychologia 37,
207-217.
Peigneux, P., Laureys, S., Fuchs, S., Destrebecqz, A., Collette, F., Delbeuck, X.,
Phillips, C., Aerts, J., Del Fiore, G., Degueldre, C., Luxen, A., Cleeremans, A., Maquet, P.,
2003. Learned material content and acquisition level modulate cerebral reactivation
during posttraining rapid-eye-movements sleep. Neuroimage 20, 125-134.
Peyrache, A., Khamassi, M., Benchenane, K., Wiener, S.I., Battaglia, F.P., 2009.
Replay of rule-learning related neural patterns in the prefrontal cortex during sleep. Nat
Neurosci 12, 919-926.
Pourtois, G., Peelen, M.V., Spinelli, L., Seeck, M., Vuilleumier, P., 2007. Direct
intracranial recording of body-selective responses in human extrastriate visual cortex.
Neuropsychologia 45, 2621-2625.
Pourtois, G., Rauss, K.S., Vuilleumier, P., Schwartz, S., 2008. Effects of perceptual
learning on primary visual cortex activity in humans. Vision Res 48, 55-62.
Pourtois, G., Spinelli, L., Seeck, M., Vuilleumier, P., 2010a. Modulation of face
processing by emotional expression and gaze direction during intracranial recordings in
right fusiform cortex. J Cogn Neurosci 22, 2086-2107.
Pourtois, G., Vocat, R., N'Diaye, K., Spinelli, L., Seeck, M., Vuilleumier, P., 2010b.
Errors recruit both cognitive and emotional monitoring systems: simultaneous
intracranial recordings in the dorsal anterior cingulate gyrus and amygdala combined
with fMRI. Neuropsychologia 48, 1144-1159.
Press, D.Z., Casement, M.D., Pascual-Leone, A., Robertson, E.M., 2005. The time
course of off-line motor sequence learning. Brain Res Cogn Brain Res 25, 375-378.
Qin, Y.L., McNaughton, B.L., Skaggs, W.E., Barnes, C.A., 1997. Memory
reprocessing in corticocortical and hippocampocortical neuronal ensembles. Philosophical
Transactions of the Royal Society of London 352, 1525-1533.
Rabiner, L.R., 1989. A tutorial on Hidden Markov Models and selected applications
in speech recognition. Proceedings of the IEEE 77, 257-286.
Rasch, B., Born, J., 2007. Maintaining memories by reactivation. Curr Opin
Neurobiol 17, 698-703.
Robertson, E.M., Tormos, J.M., Maeda, F., Pascual-Leone, A., 2001. The role of the
dorsolateral prefrontal cortex during sequence learning is specific for spatial information.
Cereb Cortex 11, 628-635.
Ryan, L., Lin, C.Y., Ketcham, K., Nadel, L., The role of medial temporal lobe in
retrieving spatial and nonspatial relations from episodic and semantic memory.
Hippocampus 20, 11-18.
75
Schendan, H.E., Searl, M.M., Melrose, R.J., Stern, C.E., 2003. An FMRI study of the
role of the medial temporal lobe in implicit and explicit sequence learning. Neuron 37,
1013-1025.
Seidler, R.D., Purushotham, A., Kim, S.G., Ugurbil, K., Willingham, D., Ashe, J.,
2005. Neural correlates of encoding and expression in implicit sequence learning. Exp
Brain Res 165, 114-124.
Shima, K., Isoda, M., Mushiake, H., Tanji, J., 2007. Categorization of behavioural
sequences in the prefrontal cortex. Nature 445, 315-318.
Spencer, R.M., Sunm, M., Ivry, R.B., 2006. Sleep-dependent consolidation of
contextual learning. Curr Biol 16, 1001-1005.
Staeren, N., Renvall, H., De Martino, F., Goebel, R., Formisano, E., 2009. Sound
categories are represented as distributed patterns in the human auditory cortex. Curr
Biol 19, 498-502.
Stickgold, R., Hobson, J.A., Fosse, R., Fosse, M., 2001. Sleep, learning, and
dreams: off-line memory reprocessing. Science 294, 1052-1057.
Strik, W.K., Chiaramonti, R., Muscas, G.C., Paganini, M., Mueller, T.J., Fallgatter,
A.J., Versari, A., Zappoli, R., 1997. Decreased EEG microstate duration and
anteriorisation of the brain electrical fields in mild and moderate dementia of the
Alzheimer type. Psychiatry Res 75, 183-191.
Swets, J.A., Pickett, R.M., Whitehead, S.F., Getty, D.J., Schnur, J.A., Swets, J.B.,
Freeman, B.A., 1979. Assessment of diagnostic technologies. Science 205, 753-759.
Tzovara, A., Murray, M.M., Plomp, G., Herzog, M.H., Michel, C.M., De Lucia, M.,
submitted. Decoding stimulus-related information from single-trial EEG responses based
on voltage topographies.
Van de Ville, D., Britz, J., Michel, C.M., 2010. EEG microstate sequences in healthy
humans at rest reveal scale-free dynamics. Proc Natl Acad Sci U S A 107, 18179-18184.
Walker, M.P., Brakefield, T., Morgan, A., Hobson, J.A., Stickgold, R., 2002. Practice
with sleep makes perfect: sleep-dependent motor skill learning. Neuron 35, 205-211.
Wallenstein, G.V., Eichenbaum, H., Hasselmo, M.E., 1998. The hippocampus as an
associator of discontiguous events. Trends Neurosci 21, 317-323.
Wilson, M.A., McNaughton, B.L., 1994. Reactivation of hippocampal ensemble
memories during sleep. Science 265, 676-679.
Winocur, G., Moscovitch, M., Sekeres, M., 2007. Memory consolidation or
transformation: context manipulation and hippocampal representations of memory. Nat
Neurosci 10, 555-557.
76
Figures
Figure 1
Illustration of the experimental procedure. (a) The experimental protocol included
one training session on day one, followed by one night of sleep, and a test session on
day 2. Each session lasted about 30 min and comprised 60 S-sequence (3 blocks) and 20
C-sequences (1 block). The task was performed with the non-dominant hand and the
trained 8-item sequence is shown on the bottom: finger 1 corresponds to the index and
finger 4 to the little finger. (b) On each trial in the sequence, one of the four grey
rectangles briefly flashed (yellow during 200 ms), indicating which button to press next.
One sequence in this serial reaction time task required 8 keypresses in a specific order.
Figure 2
Classification procedure. (a) Representation of the implanted areas in patient C.S,
lateral view of both hemispheres; (b) intracranial EEG (iEEG) recorded over the N
contacts at each timeframe form an N-dimensional vector; (c) iEEG measures are pooled
together within an Ndimensional space irrespective of their temporal order and a GMM is
used to estimate Q Gaussian parameters (here Q = 3); (d) a Hidden Markov Model with
Q hidden states, initialized by the GMM parameters, is used to update the Gaussian
estimations as well as the transition matrix between the hidden states.
77
4
a
b
1
M.R.
True positive rate (sensitivity)
1
0.8
0.6
0.4
iEEG S-sequence
iEEG C-sequence
0.2
RTs S-sequence
RTs C-sequence
0
C.S.
0.8
0.6
0.4
0.2
0
0
0.2
0.4
0.6
0.8
1
0
False positive rate (1-specificity)
False pos
b
1
iEEG S-sequence
iEEG C-sequence
RTs S-sequence
RTs C-sequence
C.S.
0.8
0.6
0.4
0.2
0
0.6
0.8
1
ve rate (1-specificity)
0
0.2
0.4
0.6
0.8
0.2
1
False positive rate (1-specificity)
Figure 3
Classification results. ROC curves for MR (a) and CS (b) representing the
classification results when taking into account all contacts, comparing the iEEG data from
S-sequence (plain red line) and those from the C-sequence (plain blue line). The same
model was used to classify reaction times from the S-sequence (dotted red line) and
from the C-sequence (dotted blue line). Accuracy for classifying single-trials belonging to
S-sequence was always significantly higher than that for the C-sequence and for the
reaction times, suggesting that the classification algorithm captured information in the
signal that was specific for the learning of a deterministic visuomotor sequence.
78
S-sequence
C-sequence
S-sequence one area out
C-sequence one area out
0.9
b
a
0.8
ROC area
M.R.
0.7
0.6
0.5
LPH1 RFO2 RA1 RAH2 RPH1 LFO2
LPH8 RFO8 RA7 RAH8 RPH8 LFO8
LPH1
LPH8
LA1
LA8
LAH1
LAH8
0.9
c
d
ROC area
0.8
C.S.
0.7
0.6
0.5
RAH1 RC1 RFO1 RA2
RAH8 RC8 RFO8 RA8
RAH1 RFC1
RAH8 RFC8
RO2
RO8
LA1
LA7
LO1
LO8
Figure 4
Localization results. ROC areas for MR (a) and CS (b), comparing classification
performance when taking into account all implanted areas versus after dropping in turn
one area. Red bars refer to the S-sequence, blue bars to the C-sequence. Significant
differences between keeping all the contacts and dropping in turn couples of electrodes
were observed for the S-sequence only for the string including the hippocampus and
right frontal caudate contact for CS. By contrast for the C-sequence, a significant drop in
the classification accuracy was evident for many areas. Importantly single-trial
classification remains always lower in the CS than in the S-sequence.
Abbreviations: LA = left amygdala; RA = right amygdala; LAH = left anterior
hippocampus; LPH = left posterior hippocampus; RAH = right anterior hippocampus;
RPH = right posterior hippocampus; LFO = left fronto-orbital area; RFO = right frontoorbital area; LO = left occipital; RO = right occipital; RFC = right frontal-caudate.
Standard errors are shown.
a
c
b
RFC1
LPH1
LPH8
M.R. Coronal section, left posterior
hippocampus
RAH8
RFC8
RAH1
C.S. Coronal section, right anterior
hippocampus
C.S. Axial section, right frontalcaudate
Figure 5
Visualization of the electrode arrays targeting the critical areas revealed by the
classification. (a) Left posterior hippocampus for patient M.R. (b) Right anterior
hippocampus and (c) right frontal-caudate region for patient C.S. Electrodes localized by
the CT scan were coregistered to the MRI T1-weighted brain volumes.
79
a
Epoch Origin
40
S-sequence day1
S-sequence day2
Keypress
Visual cue
C-sequence day1
C-sequence day2
0
1
-20
.95
-40
.90
M.R.
LPH1
-60
Confidence level
Amplitude ( V)
20
.85
.80
-200
-100
0
100
200
300
400
500
600
700
800
Time (ms)
b
15
Epoch Origin
Visual cue
Keypress
5
1
0
1
-5
.95
-10
.90
C.S.
RAH1
-15
.85
Confidence level
Amplitude ( V)
10
.80
-200
-100
0
100
200
300
400
600
500
700
800
Time (ms)
Figure 6
Illustration of stimulus-locked ERP results for the hippocampal contacts
contributing most to the classification results. (a) Patient M.R., left posterior
hippocampus showed a positive amplitude modulation starting 293 ms after the visual
cue onset (i.e. 211 ms before the average keypress time) and lasting for 43 ms. (b)
Patient C.S., right anterior hippocampus showed a positive amplitude modulation
starting 186 ms after visual cue onset (i.e. 219 ms before the keypress) and lasting for
61 ms. The time window of significant difference between day 1 and day2 (p<.05,
stability > 20ms) is indicated in yellow. Note a similar overall morphology of the
averaged potential modulation in the two different patients.
8
Epoch Origin
Visual cue
S-sequence day1
S-sequence day2
C-sequence day1
C-sequence day2
Keypress
6
4
2
1
1
0
.95
-2
.90
-4
-6
C.S.
RFC1
.85
Confidence level
Amplitude ( V)
a
.80
-200
-100
0
100
200
300
400
500
600
700
800
Time (ms)
Figure 7
Illustration of the stimulus-locked ERP modulation for the frontal-caudate strip.
Here, the positive amplitude modulation started 332 ms after visual cue onset (i.e.72 ms
before keypress) and lasted until the end of the epoch. The time window of significant
difference between day 1 and day2 (p<.05, stability > 20ms) is indicated in yellow.
80
Evidence of overt replay of a recently
learned motor sequence during human
sleep
Experiment
2.
CONTEXT
Sleep facilitates memory consolidation. According to this hypothesis, brain
structures engaged in the learning process may be reactivated during sleep possibly by
replaying newly acquired information. While temporally-organized sequences of
neural activity have been found to be replayed in animals, there is up to date no direct
evidence for such replay in humans.
Based on the observation that sleepwalkers and patients with rapid eye movement
(REM) sleep behavior disorder (RBD) may overtly act out their sleep mentations,
respectively during non-REM and REM sleep, we thought to test these patients for a
possible behavioral replay of recent waking events. These parasomnia models may
provide a new window into the cognitive processes during sleep and dreaming.
We trained subjects on a modified version of a serial reaction time task (SRTT)
sequence, consisting in complex and large arm movements, which could be easily
detected if overtly replayed during sleep.
Video-polysomnographic recordings were performed to assess sleep structure
and sleep-related behavior.
SUMMARY OF RESULTS
Both patients groups showed sleep-related performance improvement of the
trained sequence. Importantly, one (possibly two) sleepwalker overtly replayed part of
the trained motor sequence during slow-wave sleep.
CONCLUSION
This study provides, to our knowledge, the first demonstration for a realistic,
temporally-structured replay of a recently-learned behavior during sleep in humans.
81
Article submitted to PLoS ONE
Evidence for the re-enactment of a recently learned behavior during sleepwalking
Irina Constantinescu*4, MD, Delphine Oudiette*1,3, MSc, Laurène Leclair-Visonneau1,3,
MD, Marie Vidailhet2,3, MD, Sophie Schwartz4,5, PhD, and Isabelle Arnulf1,2,3, MD, PhD
*
These authors contributed equally to this work
1
Sleep Disorders Unit, Pitié-Salpêtrière Hospital, APHP, Paris, France
2
Inserm, U975, Paris, France
3
Université Pierre et Marie Curie-Paris6, Centre de Recherche de l'Institut du Cerveau
et de la Moelle epiniere, UMR-S975, Paris, France
4
Department of Neuroscience, and Geneva Neuroscience Center, University of
Geneva, Switzerland
5
Swiss Center for Affective Sciences (NCCR), University of Geneva, Switzerland
Correspondence to:
Dr Isabelle Arnulf
Unité des Pathologies du Sommeil, Hôpital Pitié-Salpêtrière
47-83 boulevard de l‟Hôpital 75651 Paris Cedex 13
Phone: 01 42 16 77 02 E mail: isabelle.arnulf@psl.aphp.fr
ABSTRACT
Animal studies have shown that sequenced patterns of neuronal activity may be
replayed during sleep. However, the existence of such replay in humans has not yet
been directly demonstrated. Here we studied patients who exhibit overt behaviors
during sleep to test whether sequences of movements trained during the day may be
spontaneously reenacted by the patients during sleep.
We recruited 19 sleepwalkers (who displayed complex and purposeful behaviors
emerging from non REM sleep), 20 patients with REM sleep behavior disorder (who
enacted their dreams in REM sleep) and 18 healthy controls. Continuous video sleep
recordings were performed during sleep following intensive training on a sequence of
large movements (learned during a variant of the serial reaction time task).
82
Both patient groups showed learning of the intensively trained motor sequence after
sleep. We report the re-enactement of a fragment of the recently trained motor
behavior during one sleepwalking episode.
This study provides, to our knowledge, the first evidence of a temporally-structured
replay of a learned behavior during sleep in humans. Our observation also suggests
that the study of such sleep disorders may provide unique and critical information
about cognitive functions operating during sleep.
Keywords: learning, replay, REM sleep behavior disorder, sleepwalking, parasomnia
INTRODUCTION
It is well established that sleep facilitates plastic changes that underlie the
consolidation of recently acquired knowledge [1,2,3,4]. The prevailing hypothesis
states that the neural traces coding for the newly acquired information are reactivated
during sleep, thus fostering memory consolidation. In rats and birds, specific patterns
of neural activity associated with recent waking behavior are spontaneously replayed
during subsequent sleep [5,6]. Similarly, functional neuroimaging studies in humans
have shown that brain regions involved in motor skill learning are reactivated during
post-training sleep [7]. These regional increases in cerebral blood flow could
correspond to the replay during sleep of patterns of neural activity coding for newly
acquired information, as observed in animals, or reflect other experience-dependent
brain processes, such as local homeostasis [4]. Replay during sleep is also suggested
by two recent studies in which sensory cues, previously associated with learned
material, were presented during sleep, yielding better memory performance [8,9].
Dreams also contain a high proportion of recent waking experiences [10,11,12].
However, direct evidence for a replay of temporally-structured information during
human sleep is still lacking.
Here we propose to address this question by studying patients who exhibit overt
behaviors while asleep, such as patients with REM sleep behavior disorder (RBD) or
sleepwalkers. RBD is a recently-discovered parasomnia characterized by a loss of the
physiological REM sleep-associated muscle atonia [13], which results in motor
activity reflecting the enactment of violent and vivid dreams [13,14]. The behaviors
during RBD are various, purposeful and complex [14,15,16,17]. Chronic RBD mainly
affects middle-aged men, as an isolated condition (idiopathic RBD) or associated with
83
narcolepsy and neurodegenerative diseases (mainly synucleinopathies) [18].
Sleepwalking affects mostly children and young adults and is characterized by overt
and complex motor behaviors initiated during slow-wave sleep [13]. Dreaming during
sleepwalking is sometimes reported in adults, which constitutes a form of dreamenacting behavior much like RBD [19].
In the present study, we trained both patients with RBD and sleepwalkers on a
modified version of a serial reaction time task (SRTT), which is known to robustly
benefit from sleep [7,20,21]. We tested whether, during sleep, the patients would
replay fragments of a recently trained sequence involving large arm movements.
METHODS
Ethic Statement
The institutional review board 06-03 gave its approval for the study, which was
considered as non invasive.
Subjects
A total of 20 patients with RBD (mean age 66.5 ± 6.5 years; 16 males, 4
females), 19 sleepwalking patients (mean age 34.4 ± 15.4 years; 6 males, 13 females),
and 18 healthy controls (mean age 57.9 ± 5.3 years; 14 males, 4 females) took part in
this study. As expected from these pathologies, the population of sleepwalkers was
younger than the population of RBD patients (p<0.001). We also tested a group of
healthy controls whose mean age would lie between those of the patient groups. It is
important to note that the inclusion of the control group mainly serves to confirm that
the sequence of ample movements could be learned equally by all these distinct
populations. The critical measures of learning involve within-subject comparisons,
while differences in mean reaction-times due to age are expected but not relevant here.
RBD was defined according to standard criteria [13], including i) a positive
clinical history (bed partners reporting most often violent, purposeful limb or body
movements, as if the patients were acting out their dreams), and ii) the presence of
abnormally enhanced chin or phasic muscle tone and/or complex and non-stereotyped
movements (such as gesturing, reaching, grabbing, punching, kicking, talking,
laughing, running, chewing, feeding, drinking) during REM sleep on video sleep
84
recording. In the RBD group, 13 patients had idiopathic RBD and 7 patients had
Parkinson disease, with a mild to moderate motor disability (Hoehn and Yahr score
between I and III) [22], which did not impact their daily activities, and no dementia
(Mini-Mental State Examination score higher than 23) [23].
Sleepwalking was defined according to clinical international criteria [13], plus at
least one arousal during slow-wave sleep associated with a motor episode suggestive
of surprise, confusion, or fear (startle response, sitting up in the bed, looking around
surprised), or numerous sudden arousals during slow-wave sleep and no epilepsy or
sleep-disordered breathing. We included in the study primary sleepwalking patients,
without any pre-existent psychiatric and/or neurological illness.
Subjects were informed that we tested their performance in the task execution,
but they were not aware of the regularity of the trained sequence, nor that we expected
a potential replay during sleep.
All participants gave informed consent to participate in this study, which was
approved by the local ethics committee. Healthy volunteers (but not the patients) were
paid for their participation.
Behavioral task and experimental procedure
The subjects were trained on a modified version of the SRTT [24], which was
previously shown to elicit sleep-related performance improvement [7,20,21]. Instead
of requiring finger movements, the present task involved large hand, forearm and arm
movements that would be clearly visible on the video recorded during sleep. Note that
twitches of finger muscles are extremely frequent during normal REM sleep, and even
more during RBD [25], so that a replay of learned fingers movements could not be
easily distinguished from multiple and non-specific twitches during the REM sleep.
The subjects sat comfortably facing the computer screen; they were asked to
react as quickly and accurately as possible when a large, colored rectangle appeared
on the screen by pressing the corresponding colored response button (e.g., when the
rectangle on the screen was green, the subject had to press the green button). The four
response buttons were attributed four different colors (red, yellow, blue, or green) and
were placed at different spatial localizations, as shown in Figure 1. The subjects were
asked to use their left hand for the blue and green buttons and their right hand for the
85
red and yellow buttons so that all participants carried out the same sequence of
movements, including movements crossing the midline (see Supplemental Video).
Each correct response was immediately followed by the presentation of the next
stimulus on the screen, thus eliciting a new response, and so on. If the subject failed to
press the appropriate button, visual (a white rectangle) and auditory (a simple beep)
feedback were simultaneously delivered followed by the presentation of the next
visual stimulus. We used E-Prime (Psychology Software Tools, Inc.) for stimulus
presentation and response recordings.
The subjects were intensively trained on a fixed 8-item sequence (Blue-YellowGreen-Red-Yellow-Blue-Red-Green) and were not informed about the regularity of
the sequence. To distinguish between structured sequence learning and improvement
in visuomotor responses independent of the sequence, a random sequence was also
presented. Like the structured sequence, the random sequence was composed of two
series of four colors, in which each of the four buttons was pressed once, but in a
completely random order. Because each color appeared twice in each eight-item
sequence of the random condition, only the order of the colors was altered between
the random and the structured conditions. Subject performance was quantified by
measuring reaction time and response accuracy (correct hand and correct button use).
We developed this version of the SRTT to optimize the detection of spontaneous
replay during a subsequent REM sleep behavior or a sleepwalking episode. In
particular, the task involved ample movements, with the response buttons placed
about 50 cm apart in the peripersonal space, and the movements executed during the
task were unusual as compared to the behaviors generally observed during parasomnia
episodes (our task included pronating forearm and flexing hand movements, as well as
arms crossing the midline). Supplemental Video shows the sequence performed by a
healthy control subject during a training session and, after acquisition, while lying in a
bed and performing the sequence from memory to facilitate later visual comparison
with any potential replay during sleep.
The experimental procedure consisted of a training session at 6 pm and an initial
test session at 8 pm, followed by a night of sleep. The training session consisted of
four consecutive blocks of structured sequences (10 sequences in each block, 80 trials
per block). The initial test session consisted of four such blocks and one additional
block of random sequences placed in the middle of the session (third block of the five86
block session). The following morning, the subjects were retested at 9 am with the
same series of blocks as during the initial test session. In this protocol, each subject
performed a total of 1,120 trials.
On the morning immediately after awakening from sleep and before the test
session, the participants were asked about their night dreams the previous night. A
summary of the experimental procedure is provided in Figure 2.
Sleep and nocturnal behavior monitoring
All subjects underwent video and sleep monitoring during the night immediately
following the training; 7 sleepwalkers and 7 patients with RBD were recorded over
two consecutive nights, with no additional training or retesting in between the two
nights. Scalp electroencephalography included three bipolar channels for the patients
with RBD and the controls (Fp1/Cz, O2/Cz, C3/A2) and eight bipolar channels for
sleepwalkers (FP1/C3, C3/O1, C3/T3, T3/O1, FP2/C4, C4/O2, C4/T4, T4/O2) to
exclude nocturnal frontal lobe epilepsy. Monitoring also included EEG-synchronized
infrared video-monitoring and sound recording in the room (Brainet®, Medatec Ltd,
France), a right and left electro-oculogram, submentalis and tibialis anterior muscle
electromyogram, nasal pressure through a cannula, tracheal sound recording through a
microphone placed at the surface of the trachea, thoracic and abdominal strain jauges
to assess respiratory efforts, electrocardiography, and pulse oximetry. Sleep stages,
EEG arousals, muscle activity, periodic leg movements and respiratory events were
scored by visual inspection of the multimodal recordings, using standard criteria [26],
by two independent and experienced scorers.
Statistical analyses of motor performance
Statistical analyses of behavioral measurements were performed using Statistical
Package for Social Sciences (SPSS Inc, Chicago, IL, USA). For each of the three
groups of subjects (sleepwalking group, RBD patients group and control group) we
performed a repeated-measures ANOVA with sessions (training, testing) and type of
sequence (structured, random) as factors. For the type of sequence factor, we selected
the random block and the structured block performed just afterwards, because they
were close in time (respectively block 3 and block 4), which was important given the
87
fatigability of some of the patients. The selected blocks were also contextually very
similar (both blocks implied a switch of sequence type, from structured to random and
from random to structured sequence). In these analyses, we considered each group
separately because the subjects in the different groups were not matched for age (e.g.,
RBD patients were significantly older and had significantly slower reaction times than
the sleepwalkers, as expected in these disorders) and because the goal of these
analyses was to test for sequence-specific learning by comparing change in
performance for the structured sequence and for the random sequence between the
pre-sleep and post-sleep testing sessions.
Assessment of sequence replay during sleep
To confirm the putative replay of a fragment of the sequence observed in one
sleepwalker (see Results), we asked 11 independent judges blind to the aim of the
experiment to assess resemblance of 113 video clips of sleepwalking episodes with a
video showing an awake subject performing the task from memory while lying in a
bed, in the same recording conditions as the patients (Supplemental Video). For each
video clip, the judges were asked six questions concerning the sleepwalking episode
(see Supplemental Material), including a final assessment of the resemblance of
sleepwalking movements with the movements performed by the awake subject on a
scale from 0 (no resemblance at all) to 10 (identical movement sequence).
Note that because we did not observe any sign of replay in the RBD, a similar
procedure was not used with RBD episodes.
RESULTS
Sleep and cognitive performances
Sleep recordings confirmed a normal sleep pattern in the night following training
in all the three tested groups. Expected differences were found for some sleep
parameters when comparing RBD patients with healthy controls (such as total sleep
time and sleep efficiency, see Table 1). Of note, the duration of slow-wave sleep was
longer in controls than in RBD patients (p<0.001) but remained within normal values.
Upon testing after one night of sleep, both groups performed better than during
the pre-sleep session (Figure 3). Improvement after sleep was greater for the trained,
88
structured sequence than for the random sequence. Furthermore, the results showed an
interaction between session (pre-sleep or post-sleep) and condition (structured
sequence or random sequence) (F(1,16)=9.456, p=0.007 in controls; F(1,18)=6.670,
p=0.019 in RBD patients; F(1,18)=20.247, p<0.001 in sleepwalkers), confirming
sequence-specific learning for our task, which was more complex than the classical
SRTT. It is already known that sleep may facilitate learning of sequences of
movements [7,20]. In the present work, our aim was not to re-validate pre-existent
knowledge about sleep-related memory consolidation, but to make use of a task that
may benefit from sleep to increase the probability to observe task-related behaviors
during sleep.
Dreams content after the post-training night of sleep
On the morning following each of the first night of recorded sleep, only seven
subjects out of the 57 reported some dream content. Most (40/57) tested subjects did
not recall any dreams, and 10 subjects reported a „blank dream‟ (feeling of having
dreamt but no memory of it). There was no difference in dream recall frequency
between the controls, the patients with RBD and the sleepwalkers. Interestingly, 2 out
of the 7 subjects who recalled a dream reported some content related to the recently
learned task. One sleepwalker, who spontaneously used the word “cross” to verbally
characterize part of the motor sequence required by the task, reported following the
night of sleep, a “cross” dream involving two diagonal paths. In his dream, the patient
was driving to school to register his children (this path corresponded to the first
diagonal). Suddenly, an unknown driver struck the patient‟s parked car and fled to the
airport located in the opposite direction (second diagonal). The second report of dream
content related to the learned task came from a healthy control subject who dreamed
that he was concerned about respecting the order of the colors in the sequential task
and tapped the colored response buttons in his dream.
Behaviors exhibited during post-training sleep
RBD patients
The patients with RBD exhibited several complex behaviors during REM sleep
of the two experimental nights (i.e., hand movements, defense posture, kicking,
89
punching, reaching, smiling, pointing, leaping out of bed, whispering and speaking).
No obvious motor replay of the task was identified among these REM sleepassociated behaviors.
Sleepwalkers
During the two recorded nights following training, the sleepwalking patients
exhibited several typical behaviors of sleepwalking during slow-wave sleep (startling,
looking around confused, speaking, raising the head and the chest, touching the
bedside table), the part of these enacted behaviors corresponding only to a mean of
one minute of non REM sleep per night. Furthermore, we observed one uncommon
behavior in a sleepwalker, suggestive of a replay of the recently trained motor task:
during slow-wave sleep of the second post-training night, at 00:58 am, Patient 1
initially startled, then opened her eyes, stood up, laid down again, whispered, and
engaged in a sequence of hand and arm movements. She sequentially raised her left
arm with her hand extended and pronated like when executing the recently learned
task, raised her right arm with her fingers outspread, waited in this raised position, and
slowly pressed down with the right hand, left hand and then right hand, as if pressing
down invisible pads (Figure 4 and Supplemental Video). The unusual behavior
described in this sleepwalking patient reflected obvious and accurate re-enactment of
a short fragment of the recently learned sequence of movements (as later confirmed by
the independent judges, see below). The next morning, the patient had the feeling of
having dreamt but could not remember any dream content.
The sleepwalking patient who described the cross dream (Patient 2) also
displayed an uncommon behavior for a sleepwalking episode (he raised and dropped
his right hand twice and his left hand once) when emerging from slow-wave sleep,
before being fully awake. While the observed movements in this patient presented
some analogy with the gestures of the task, they cannot be considered as a replay of
the sequence itself.
Evidence for behavioral replay during sleep
Only the putative replay of the first sleepwalker was judged resembling with the
wake task movements, with an average score of 7.2±2.6 (score/10); the others
90
movements displayed by sleepwalkers during the experimental nights were rated
below 5/10 by the judges (Supplemental Figure 1). Because the judges differed in the
way they used the scoring scale (some being more conservative than others), the 0-10
score was converted to a 0-1 scale (1 corresponding to the maximal rate given over 10
by each judge), with a resulting average resemblance score of 0.93/1. Note that the
other behavioral episode possibly related to task re-enactement (Patient 2; sleepwalker
who experienced the “cross” dream) was evaluated below 5/10 (0.6/1).
We also computed the average resemblance score (with the trained motor
sequence) separately for sleepwalking episodes from patients who were trained on the
task (47 episodes corresponding to 14 patients) and for episodes from patients not
trained on the task (68 episodes corresponding to 30 patients). The average
resemblance was higher in the trained compared to the untrained group (p<0.00001)
(Figure 3).
A summary of the movements‟ description of the sleepwalking videos by the
eleven judges is provided in the Supplementary data.
DISCUSSION
To our knowledge, the present findings represent the first direct and
unambiguous demonstration of overt behavioral replay of a recently learned skill
during human sleep.
Several lines of evidence support our claim that the behavior observed during an
episode of sleepwalking (Patient 1) is a true re-enactment of the trained serial reaction
time task. First, there is a striking similarity between the hand movements displayed
by Patient 1 and those of the waking subject performing the sequence from memory:
11 judges blind to the experimental procedure scored this episode as closely similar to
the trained sequence, after careful observation of 113 video sleepwalking episodes
(Supplemental Figure1). Second, the complex motor patterns performed by Patient 1
during slow-wave sleep are consistent with activation of large-scale, distributed brain
networks during slow-wave sleep-associated replay, as previously suggested by
animal studies [27]. Finally, our finding that the behavioral replay concerned only a
part of the sequence is in line with the fragmentary replay observed in animal studies
and in human dream studies [28].
91
Actually, the probability of observing overt behaviors in patients with RBD and
in sleepwalkers is low, making our finding of overt replay highly remarkable. Indeed,
patients with RBD exhibit complex, purposeful behaviors during only 0.1% to 20% of
the total time spent in REM sleep [29]. In sleepwalkers, overt behavior is even rarer,
corresponding to a mean of 0.7% (or ~one minute) of non REM sleep per night. These
estimations suggest that the observed behavior in the first sleepwalking patient may
emerge from “a strong pressure”, probably arising from active neural processes of reshaping and strengthening of the intensively trained information during post-training
sleep, related to off-line recapitulation of the newly acquired cognitive schema.
Furthermore, because elements from recent memory are frequent in dreams [30] and
may correlate with learning [31], it is likely that mental replay did occur in other
trained patients in the absence of overt behavioral expression. Further support for a
replay of waking motor behavior during subsequent episodes of sleepwalking is also
provided by the fact that sleepwalkers who performed the task were more likely to
show movements that were on averaged rated as more similar to the trained motor task
compared to sleepwalkers who were not trained on the task (Figure 3).
Such complex motor patterns performed by the sleepwalker showing evidence of
experience-dependent behavior during slow-wave sleep is consistent with data in
animal studies, showing activation of large-scale, distributed brain networks during
slow-wave sleep-associated replay [27]. The evidence
for a behavioral replay
reflecting only a part of the sequence is in line with the fragmentary replay observed
in animal studies and in human dream studies [28].
We would like to suggest that the study of motor behaviors in sleepwalkers
provides highly valuable information about cognitive and motor processes occurring
during sleep. Indeed, sleep macro- and micro-structure (e.g. rapid eye movement
density, EEG) is normal in patients with RBD and sleepwalkers [29,32]. Sleepwalking
results from concomitant local sleep in the frontoparietal associative cortices and local
arousal in motor and cingular cortices, as shown by recent functional brain imaging
and deep brain monitoring in humans during sleepwalking [33,34]. Therefore, except
for their motor aspects, the mechanisms regulating sleep in sleepwalkers are mostly
intact. Furthermore, like healthy controls, both sleepwalkers and the patients with
RBD improved their performance, thus showing spared overnight memory
consolidation processes.
92
Our work therefore demonstrates that parasomnias such as sleepwalking and
RBD are useful neurological models for studying cognitive functions during sleep,
and may for example motivate further investigations on the respective contribution of
non REM and REM sleep on learning and brain plasticity.
Acknowledgements
We thank Cédric Oudiette, Emmanuel Roze, and Lionel Naccache for their helpful
comments.
References
1. Diekelmann S, Born J (2010) The memory function of sleep. Nat Rev Neurosci
11: 114-26
2. Maquet P (2001) The role of sleep in learning and memory. Science 294: 10481052.
3. Stickgold R (2005) Sleep-dependent memory consolidation. Nature 437: 12721278.
4. Huber R, Ghilardi MF, Massimini M, Tononi G (2004) Local sleep and learning.
Nature 430: 78-81.
5. Wilson MA, McNaughton BL (1994) Reactivation of hippocampal ensemble
memories during sleep. Science 265: 676-679.
6. Dave AS, Margoliash D (2000) Song replay during sleep and computational rules
for sensorimotor vocal learning. Science 290: 812-816.
7. Maquet P, Laureys S, Peigneux P, Fuchs S, Petiau C, et al. (2000) Experiencedependent changes in cerebral activation during human REM sleep. Nat Neurosci 3:
831-836.
8. Rasch B, Buchel C, Gais S, Born J (2007) Odor cues during slow-wave sleep
prompt declarative memory consolidation. Science 315: 1426-1429.
9. Rudoy JD, Voss JL, Westerberg CE, Paller KA (2009) Strengthening individual
memories by reactivating them during sleep. Science 326: 1079.
10. Domhoff GW (1996) Finding Meaning in Dreams: A Quantitative Approach Book
Description. New York: Plenum press. 372 p.
11. Fosse MJ, Fosse R, Hobson JA, Stickgold RJ (2003) Dreaming and episodic
memory: a functional dissociation? J Cogn Neurosci 15: 1-9.
12. Schwartz S (2003) Are life episodes replayed during dreaming? Trends Cogn
Sci 7: 325-327.
13. AASM (2005) The International Classification of Sleep Disorders-Revised.
Chicago, IL: American Academy of Sleep Medicine.
14. Schenck CH, Bundlie SR, Ettinger MG, Mahowald MW (1986) Chronic behavioral
disorders of human REM sleep: a new category of parasomnia. Sleep 9: 293-308.
15. Sforza E, Krieger J, Petiau C (1997) REM sleep behavior disorder: clinical and
physiopathological findings. Sleep Med Rev 1: 57-69.
16. Olson EJ, Boeve BF, Silber MH (2000) Rapid eye movement sleep behaviour
disorder: demographic, clinical and laboratory findings in 93 cases. Brain 123: 331-339.
17. Oudiette D, De Cock VC, Lavault S, Leu S, Vidailhet M, et al. (2009) Nonviolent
elaborate behaviors may also occur in REM sleep behavior disorder. Neurology 72:
551-557.
93
18. Boeve BF, Silber MH, Saper CB, Ferman TJ, Dickson DW, et al. (2007)
Pathophysiology of REM sleep behaviour disorder and relevance to neurodegenerative
disease. Brain 130: 2770-2788.
19. Oudiette D, Leu S, Pottier M, Buzare MA, Brion A, et al. (2009) Dreamlike
mentations during sleepwalking and sleep terrors in adults. Sleep 32: 1621-1627.
20. Walker MP, Brakefield T, Morgan A, Hobson JA, Stickgold R (2002) Practice
with sleep makes perfect: sleep-dependent motor skill learning. Neuron 35: 205-211.
21. Fischer S, Hallschmid M, Elsner AL, Born J (2002) Sleep forms memory for
finger skills. Proc Natl Acad Sci U S A 99: 11987-11991.
22. Hoehn MM, Yahr MD (1967) Parkinsonism: onset, progression and mortality.
Neurology 17: 427-442.
23. Folstein MF, Folstein SE, McHugh PR (1975) "Mini-mental state". A practical
method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12:
189-198.
24. Cleeremans A, McClelland JL (1991) Learning the structure of event sequences.
J Exp Psychol Gen 120: 235-253.
25. Frauscher B, Iranzo A, Hogl B, Casanova-Molla J, Salamero M, et al. (2008)
Quantification of electromyographic activity during REM sleep in multiple muscles in REM
sleep behavior disorder. Sleep 31: 724-731.
26. De Cock VC, Vidailhet M, S. L, A. T, E. A, et al. (2007) Restoration of normal
control in Parkinson's disease during REM sleep. Brain: 450-456.
27. Peyrache A, Khamassi M, Benchenane K, Wiener SI, Battaglia FP (2009) Replay
of rule-learning related neural patterns in the prefrontal cortex during sleep. Nat
Neurosci 12: 919-926.
28. Lee AK, Wilson MA (2002) Memory of sequential
hippocampus during slow wave sleep. Neuron 36: 1183-1194.
experience in the
29. Leclair-Visonneau L, Oudiette D, Gaymard B, Leu-Semenescu S, Arnulf I (2010)
Do the eyes scan dream images during rapid eye movement sleep? Evidence from the
rapid eye movement sleep behaviour disorder model. Brain 133: 1737-1746.
30. Epstein AW (1985) The waking event-dream interval. Am J Psychiatry 142:
123-124.
31. Wamsley EJ, Tucker M, Payne JD, Benavides JA, Stickgold R (2010) Dreaming
of a learning task is associated with enhanced sleep-dependent memory consolidation.
Curr Biol 20: 850-855.
32. Oudiette D, Leclair-Visonneau L, Arnulf I (2010) Snoring, penile erection and
loss of reflexive consciousness during REM sleep behavior disorder. Sleep Medicine.
33. Bassetti C, Vella S, Donati F, Wielepp P, Weder B (2000) SPECT during
sleepwalking. Lancet 356: 484-485.
34. Terzaghi M, Sartori I, Tassi L, Didato G, Rustioni V, et al. (2009) Evidence of
dissociated arousal states during NREM parasomnia from an intracerebral
neurophysiological study. Sleep 32: 409-412.
94
Figures
Figure 1
Experimental design. The subjects were trained on a modified version of a serial
reaction time task. Four colored response buttons were positioned at distinct spatial
locations and the subjects had to press on the response button matching the color shown
on a computer screen. The subjects were intensively trained on a fixed eight-item
sequence (Blue-Yellow-Green-Red-Yellow-Blue-Red-Green). They were asked to use
their left hand for the blue and green buttons and the right hand for the red and yellow
buttons. The task involved a sequence of ample and uncommon movements (e.g.,
crossing arms, pronated forearm, flexed and extended hand) that would be
unambiguously recognized if replayed during sleep.
Dream
reports
S
S Structured block
R
R Random block
S
S S
S S
S S
S
Training
S
R S
S S
S R
S S
S
Video
Video--PSG
PSG
S
R S
S S
S R
S S
S
Pre-sleep testing
Sleep
Post-sleep testing
Figure 2
Experimental procedure. The experimental procedure included one training session
immediately followed by an initial testing session in the evening, and one testing session
the following morning after one night of sleep. The training session consisted of four
consecutive blocks of ten repetitions of the structured sequence. The testing sessions
before and after sleep consisted each of four blocks of ten repetitions of the structured
sequence, with one block of ten random sequences.
95
Reaction time
(ms)
1300
1200
1100
1000
900
800
700
600
500
400
300
Controls
RBD patients
Sleepwalkers
R
Training
R
Pre-sleep testing
R
SLEEP
Post-sleep testing
Figure 3
Improvement in tapping speed for all groups. Mean reaction times (RT) showing a
reduction during training session and a further decrease between pre and post-sleep
testing. Persistence of increased RT for the random blocks relative to decreased RT for
the structured blocks within the same session of testing (before-sleep and post-sleep
sessions) confirmed sequence-specific learning with training. Faster RT in sleepwalkers
may be explained by their lower mean age, as compared to RBD patients (sleepwalkers:
34.42 ± 15.35; RBD: 66.45 ±6.46 yrs). R = random sequence
A
B
1
2
3
4
Figure 4
Snapshots extracted from video recordings. (A) Execution of the sequence by a
wake control subject lying in a bed. The subject reaches toward the imaginary location of
the yellow response button (left arrow) while preparing to press the next (green) button
(right arrow). (B) Overt replay of the structured sequence by a sleepwalker during slowwave sleep. After a sudden arousal, the patient raised both arms, with pronated
forearm, and waited two seconds as if preparing to perform the task (1). Then, she
appeared to press imaginary response buttons sequentially with the right, left and right
hand (2, 3, and 4, respectively). Note the striking similarity between the posture
displayed in A and in the fourth panel in B.
96
EOG1
EOG2
Fp1-C3
C3-O1
C3-T3
T3-O1
Fp2-C4
C4-O2
C4-T4
A
B
C
D
T4-O2
EMG1
EKG
5s
Figure 5
Sleep-monitoring during the overt replay of the sequence in a sleepwalker. During
slow-wave sleep (non-REM stage 4), the patient first startled (A), raised her left arm (B)
and then sequentially pressed invisible response buttons with the right and left hands
(between C and D). During the episode, the associated EEG showed persistent slow
background activity with mixed delta, theta and alpha frequencies. A mild heart rate
increase was also observed on the electrocardiogram. After the replay, a spindle was
observed on the EEG recording.
EOG1 and EOG2, electrooculograms, FP1-C3, C3-O1, C3-T3, T3-O1, FP2-C4, C4O2, C4-T4; T4-O2, 8 channels of electroencephalograms; EMG1, electromyogram of the
submentalis muscle; EKG, electrocardiogram.
Table 1. Sleep measures in sleepwalkers, patients with REM sleep behavior
disorders (RBD), and healthy middle-aged controls
Sleepwalkers
Number of subjects
19
Age (y)
34.4 ± 15.4
Night-time sleep
Total sleep time (min)
488 ± 66
Sleep efficiency (%)
88.5 ± 10.8
Latency to sleep onset (min)
25.5 ± 15.5
Number of REM sleep episodes
4.6 ± 0.9
Sleep duration (% total sleep time)
Stage N1
4.7 ± 2.5
Stage N2
49.6 ± 8.0
Stage N3
26.1 ± 6.5
(former stage 3-4, slow-wave sleep)
REM sleep
19.5 ± 4.1
Sleep fragmentation (No events/h)
Arousals
13.8 ± 7.0
Apnea/hypopnea
1.1 ± 1.6
Periodic leg movements
4.6 ± 6.7
*p<0.05 comparison patients with RBD and controls.
Patients with
RBD
20
66.5 ± 6.5*
Healthy
controls
18
57.9 ± 5.3
376 ± 75*
78.6 ± 11.4*
40.5 ± 33.8*
3.1 ± 1.6*
428 ± 73
86.4 ± 6.6
22.3 ± 16.0
4.5 ± 1.1
8.5 ± 6.9
55.5 ± 9.6*
6.0 ± 6.9
48.3 ± 8.6
19.1 ± 6.4*
25.9 ± 6.8
16.8 ± 7.9
19.6 ± 5.2
19.2 ± 9.2
8.0 ± 10.4
33.9 ± 37.4
20.1 ± 9.5
12.1 ± 10.2
14.4 ± 29.9
97
SUPPLEMENTAL DATA
Video recordings of 113 sleepwalking episodes were edited into short clips and
unique numerical identifiers were inserted before each clip. The order of the episodes
was randomized and the clips were then copied one after the other into a single movie
file that was shown to 11 judges, blind to the experimental hypotheses and conditions.
The judges were asked to answer five yes/no questions for each of the video clip, as
well as final evaluation of the global resemblance with the clip of learned sequence
performed from memory during wakefulness (recorded with the same infrared
camera). This procedure provided an independent assessment of the similarity of
several motor aspects in each sleepwalking episode and the learned motor sequence.
Questions to judges
Question 1. During this episode, did the subject perform any movements
involving the superior limbs (e.g. arm, hand movements)?
If the answer to Question 1 was no, then the judge could skip the remaining
questions and watch the next clip. If the answer was yes, then the judge had to answer
the next 5 questions.
Question 2. During this episode, did the observed behavior include
movements of both superior limbs (left AND right hand, arm movements)?
Question 3. During this episode, were the movements of superior limbs ample
(i.e., more than 10 cm amplitude)?
Question 4. During this episode, did the movements of superior limbs include
the use of the environment (e.g. touch the electrodes, grab an object on the bed table).
Question 5. During this episode, did the movements of superior limbs include
a crossing of the body midline?
After the 5 yes/no questions, the judges were asked to evaluate the similarity of
the current episode with the learned motor sequence on a scale ranging from 0 (no
resemblance at all) to 10 (very high resemblance).
Based on the answers provided by the 11 judges, our database of 113
sleepwalking episodes contained: 92 clips with at least one movement of superior
limbs, 60 clips with movements of both superior limbs, 60 clips with ample
98
movements of superior limbs, 12 clips with alternating movements of the superior
limbs and 5 clips with a crossing of the midline.
Only 10 clips (9%) showed, at the same time, ample and alternating movements
of both right and left superior limb, confirming that a behavior resembling to the
learned sequence is not typical during sleepwalking episodes.
Only one sleepwalking episode was rated with a resemblance score above 5
(7.2±2.6, or 0.93 when converting the range of responses for each judge to a 0-1
scale), thus pointing to one of the episodes as a potential replay of the learned motor
sequence (Supplemental Figure 1).
Supplemental Figure 1
Evaluation of the resemblance between the sleepwalking episodes and the
sequence performed during wakefulness. Each point corresponds to one of the 113 video
clips. Most sleepwalking episodes were rated below 2/10 by the 11 judges. Only one clip
(number 93) obtained a mean rate greater than 5/10: this clip is the putative replay
performed by Patient 1.
Video Caption
Execution of the structured sequence in the training setting by a wake control (Part
1); execution of the sequence from memory by a wake control lying in a bed (Part 2);
overt replay of a part of the structured sequence during slow-wave sleep in a
sleepwalker (Part 3);
99
Experience-dependent induced
reactivations during post-training
wakefulness
Experiment
3.
CONTEXT
The same neural patterns may be specifically induced by presentation of paired
associated stimulus (Toita 1991, Rasch 2007, Rudoy 2009). Peigneux et al (2006)
have showed that recently learned information is spontaneously replayed during the
first hours of post-training wakefulness, thus allowing memory trace reorganization
immediately post acquisition.
Face processing includes distinct steps, reflected by distinct potential modulation
within millisecond scale. The N170 negative component, peaking around 170 ms post
stimulus, reflects early encoding of structural face features. Specific maps of stable
potential configuration characterize at scalp level the perceptual learning of faces
(Ganis and Schendan 2008). In the present study, by training human subjects on
sound-visual associations, we aimed at showing that a specifically trained sound
elicits the same potential configuration as the actual presentation of visual stimuli.
We used a combination of signal processing methods: event-related potential
(ERP) and functional microstate approach. Unlike electrode-wise approach that takes
into account only one electrode or a limited set of electrode (c.f. ERPs), the
microstates model provides information from large-scale neural networks, as it allows
the exploitation of distribution of voltage over the whole scalp. We therefore
determined the specific configuration maps reflecting particular steps of information
processing (Michel et al 2009) and statistically evaluate the presence of specific maps
across experimental conditions.
SUMMARY OF RESULTS
Our results show that sounds previously associated with a visual stimulation
generate a potential modulation corresponding to N170, which is not the case for the
sounds associated with no visual stimulation. Furthermore, by analyzing the sequence
100
of stable map configurations, we found that the same specific voltage map
corresponding to the presentation of the face is reactivated during the presentation of
the sound previously associated with faces, and that this is more important than when
presenting the sounds previously associated with scrambled faces or no stimuli.
CONCLUSIONS
Our results show that a specifically trained sound elicits the same potential
configuration as the actual presentation of visual stimuli. The associated sound may
trigger recently acquired memory traces during immediate post training period. We
believe that our results may have new implications for the understanding the plasticity
of the neural representations reflected by the N170, especially those concerning
feature-binding processes.
Article in preparation
Title: Reactivation of scalp EEG microstates after audio-visual training
Authors: I.Constantinescu, L. Ceravolo, G. Pourtois, S.Schwartz
INTRODUCTION
Important insights into the brain mechanisms of learning and memory have been
provided by the study of learned associations between initially naive and uncoupled
stimuli (Suzuki 2007).
It has been shown that experience-dependent reactivations of neural patterns can
be induced by the presentation of an associated cue (e.g. odor as a context, Rasch,
Buchel et al. 2007, sound specifically reactivating individual visual locations, Rudoy,
Voss et al. 2009). During sleep, it has been shown in animals that the same patterns of
neural activity can be reinstated following intensive training (Wilson 1994); in
humans, it has been shown that the same brain areas previously engaged in performing
a task are active during post-training sleep (Peigneux 2004; Maquet 2000).
Face recognition represents one of the most important acquired perceptual skills
in humans (Haxby 2000). Face recognition has been intensively studied by using
different neuroimaging (Haxby 1993; Puce 1995) and electrophysiological approaches
(Bentin 1996; Allison 1994; Seeck 1992). Event-related studies have identified few
101
components corresponding to different steps of face processing, at millisecond scale.
The early perceptual categorization of faces is reflected by a negativity peaking
between 140-200 ms (N170). This component is larger over the right occipitotemporal regions (Bentin 1996; Eimer 2000) and it seems to correspond to structural
encoding of faces.
It has been shown that relatively early visual areas are involved not only in lowelevel visual processing, but also in higher-level cognitive processes, such as featurebinding across associative processes (Ganis and Schendan 2008). Functional brain
imaging studies show that perception of faces and visual mental imagery of faces
activates the same temporal and occipito-temporal regions (Ishai et al 2002; O‟Craven
and Kanwisher 2000). Visual mental imagery may recruit the same neural machinery
as for visual processing, yet the pathways that influence these processes seem to be
different (bottom-up mechanisms for early visual processing and top-down
mechanisms for visualized stimuli). These effects seem to apply for other visual
categories too, like test objects (Ganis, Schendan 2008).
Moreover, visual mental imagery may involve the reactivation of previously
acquired memory traces, as shown in non-human primate work (Tomita 1999):
specific neuronal long-term memory reactivations may be elicited by the presentation
of a paired associates stimulus, in the absence of the corresponding external stimulus.
Cognitive processes involve a succession of functionally distinct brain
microstates variable duration and strength (Michel et al 2009; Lehman, 1987).
According to the microstate model, different steps of information processing imply a
sequence of stable and distinct potential map configurations. The transition from one
map to another reflects the change in signal stability.
The aim of the present study was show that a sound that had previously been
associated to a face can elicit similar topographic maps when presented alone than
when presented with the face.
EXPERIMENTAL PROCEDURE
Subjects
A total of 15 healthy right-handed participants, free of any medication, without
neurological or psychiatric history, with normal or corrected-to-normal vision and
102
normal hearing were included in the experiment (8 females, age range 25-33 yr, mean
28.45 (±2.81), 6-13 years university education). All participants gave informed
consent according to the local ethic committee regulations.
Two subjects could not complete the entire experimental procedure, due to
technical problems and data for two subjects had to be removed from further analysis
because they had too much artifacts. Therefore, data from 11 subjects were included
in the final results.
Stimuli and task
Auditory stimuli consisted in three different sounds and of a simple tone. The
sounds duration was of 500 ms each and the simple tone lasted for 100 ms. Triggers
marked the sound onset and respectively the tone onset. All acoustic stimuli had an
audio sampling rate of 44 kHz and were presented to the subjects binaurally via
headphones. The sounds and the tone were presented at an unobtrusive intensity.
The visual stimuli were photos of human faces with either a happy or a neutral
expression, and scrambled versions of these face stimuli forming either vertical or
horizontal patterns. The faces were either of neutral expression or happy and they
were taken from a series of photographs from the Ekman and Friesen (1976),
representing different emotional expressions in human subjects. Scrambled images
contained the same luminance as the faces images and were created by first
partitioning each human face image into a predetermined number of rectangular
segments (either horizontal or vertical), and then randomly repositioning these
segments. The images resolution was of 302 x 460 pixels. Visual stimuli subtending
100 at 50 cm viewing distance were presented in grayscale on the center of a
computer screen.
Behavioral task and experimental procedure
Subjects sat comfortably in front of a computer screen, in an electrical and sound
shielded room. They were asked not to move their eyes and to avoid blinking as much
as possible, throughout the whole experiment.
103
Figure 1. Experimental design with three distinct phases:
perceptual, training and testing
The experimental protocol consisted in three phases (see Figure 1): an initial
phase, where subjects were first familiarized with the three distinct sounds. A visual
block followed, in which faces and scramble images were presented repetitively. A
simple tone accompanied the beginning of each visual stimulus presentation. Trials
with only the tone alone and no visual stimulus were added. Therefore, for the visual
condition, we had three possibilities: tone coupled with a face (either happy or
neutral), tone coupled with scramble (either horizontal or vertical) and tone only.
There were 60 trials for each type of tone-visual stimulus association. We will further
refer to the visual block as the “perceptual” phase. Subjects were simply asked to
pay attention to the different sounds during the presentation of the sounds and to the
different types of images during the visual block. Throughout this initial phase,
subjects kept their eyes opened (as monitored by a video camera) and they were
fixating a cross in the center of the computer screen.
The second phase consisted in “training” on associations among sounds and
visual stimuli. For each subject, one sound was specifically associated with either
faces (happy and neutral), scrambled images (horizontal and vertical), or with no
visual stimulus (e.g. sound A was specifically associated with faces, sound B was
specifically associated with scrambled and sound C was always alone). The specific
associations between sounds and visual stimuli were counterbalanced across subjects.
Figure 2 presents the timeline of one experimental trial. During the audio-visual
association condition, the sound presentation preceded with a variable interval (from
500 to 750 ms) the image onset (and the coincident tone). Stimuli included 3 different
short sounds that were presented either before a face stimulus, a scrambled version of
a face stimulus, or no visual stimulus.
104
Figure 2. Time line of one training trial
During the learning phase, subjects watched these audio-visual sequences and
had to discriminate the gender of the faces for face stimuli, and discriminate the
predominant orientation -horizontal or vertical- for the scrambled stimuli. The trials
with no visual stimulation did not require any response from the subjects. Thus, three
distinct auditory-visual associations were created, by using three different sounds each
coupled with one specific visual condition and task. Subjects were instructed to
promptly react when seeing the image on the screen by respecting the following
configuration: when the image represented a face, they had to press with the right
hand the “x” keypress for the “happy” faces and the “c” keypress for the “neutral”
faces on the computer keyboard. For the scrambled stimuli, they had to press with the
left hand, the “n” keypress for horizontally scrambled images and the “m” keypress
for the vertically scrambled images. The hands response assigned for each visual
category was counterbalanced across subjects. When no image was presented on the
screen, the subjects were asked to refrain from pressing any response button. In the
absence of a visual stimulus, the fixation cross remained on the screen. There was a
reaction time limit, pre-determined at 600 ms after the delivery of the visual stimulus.
The time allocated for a motor answer was therefore very short, so that subjects had to
answer very quickly; therefore, in order to get a good performance on the task, the
subjects had to prepare to provide a response either to a face or to a scrambled or no
answer at all (when it was no visual stimulus) during the presentation of the sound.
Feedback was provided on the screen, when the response was not fast enough, or
when the subject pressed a button when no visual stimulus was presented on the
screen, or when the wrong hand or the wrong button press were used for an answer.
The feedback duration was of 200 milliseconds. No feedback was provided for the
105
correct answers. The order of the trials was randomized within a block. At the end of
each block a global performance score was given to the subjects, consisting in mean
reaction time and total number of errors. This was meant to motivate subjects to
continuously improve their performance. There were two blocks of training on the
audio-visual associations.
One training block of trials consisted in 90 sound-visual sensory associations,
with 30 trials associating a sound to a face, 30 trials associating a sound to scrambled
stimuli, 30 trials where a sound was not associated to any visual stimulus.
A third phase of the experiment, immediately following the previous phase
consisted in “testing” the learned associations. By presenting only the sounds and the
accompanying tone (identically organized as in the first phase) and asking subjects to
visually project in front of their eyes kept opened, the visual “category” when a
specific sound is delivered (e.g. if the subjects heard the sound A, which has been
previously associated with faces, the subject had to imagine a face). Like in the first
phase, the trials were equally distributed across the three sound types (30 trials per
distinct sound) and a tone followed the sound presentation. No motor response was
required for this experimental part.
A questionnaire assessing the vividness of mental imagery (Marks, 1973) was
filled in after the tests, by asking to rate some visual images from 1 to 5 according to
the vividness of the visual representation (see Supplementary data).
Electrophysiological data acquisition
EEG data were recorded using the 64 active-electrode Biosemi system (32
electrodes for each hemisphere). Sampling rate was 2048Hz, bandwidth 417Hz and no
external reference.
The electroencephalogram (EEG) data was collected using Biosemi system. Data
was first pre-processed using BrainVision, by performing the following steps:
downsampling at 512 Hz (sinc interpolation), average reference (the electrode Fp1 was
excluded from the average because of eye movements contamination). A high-pass
filter was set at 0.05 Hz, with a 50 Hz notch filter added. Electrode impedance was kept
below 5 kΏ. Ocular artefacts were manually detected at frontal electrodes and than
automatic ocular correction was performed. A low-pass filter of 30 Hz was set.
106
Event-related potential analysis
Event-related potentials (ERPs), locked to the tone for the “perceptive” and
“imagery” condition were averaged off-line for an epoch of 1300 ms (500 ms prestimulus baseline, 800 ms post-stimulus period). Trials contaminated by muscle
artifact, eye movements and blinks were thoroughly eliminated. Peak analyses was
performed for the N170 (negativity peaking around 170 ms for faces and 165 ms for
scrambled), measured at temporo-occipital sites (P8, P7, Po8, Po7 contacts) which
have been previously shown to provide the best representation of this component (Itier
2004). Peaks were measured within a ± 40 ms window centered on the maximum of
the grand-average (group average, N=11 subjects) means for faces and scrambled
faces and for both “perceptive” and “imagery” conditions. Component latencies were
measured at their maximum over each hemisphere and the amplitudes at each of the
chosen electrodes were considered at this latency (Picton et al 2000). Peak latencies
and amplitudes were analyzed with repeated measures analyses of variance
(ANOVA). First, we determined the contact which showed the best representation
among the selected contacts. For the best contact, we than preformed peak and latency
comparisons across conditions.
Segmentation analyses
Functional microstates (Michel et al. 1992, Lehman, 1987) refer to time
segments of stable potential map configuration supposed to reflect different steps of
information processing. The transition from one map to another reflects the change in
signal stability. According to the microstate model, the brain activity can be seen as a
sequence of non-overlapping microstates of variable duration and strength. We used a
statistical approach in order to determine the temporal sequence of microstates within
the spontaneous EEG map series. The statistical analysis is based on the global map
dissimilarity (i.e. the spatial correlation between maps) (Pasqual-Marqui et al 1995).
This parameters is taken into consideration while performing a cluster-analysis which
allows the grouping of maps which are highly correlated. The algorithm determines
for each number of clusters, the most dominant topographies in a given ERP map
series, that best explain the variance of the data. The microstate segmentation of the
ERPs therefore proposes to define ERP components in terms of the sequentially
107
appearing map topographies. We assessed for a defined component, the spatial
configuration of the scalp potential that characterizes it.
Since different potential configurations are caused by different intracranial
generators, the characterization of the ERP‟s as a sequence of spatially distinct maps aims
to define the different large-scale neuronal networks that are activated by the performed
mental operation. As for the spontaneous EEG, the microstate analysis of the ERPs
proposes that each global network is activated during a certain period of time and then it
is replaced by a new network that remains stable for a certain time period, and so on.
The second step within the segmentation analyses is the fitting procedure, which
allows to check statistically the presence of the most representatiive segmentation
maps in the individual ERPs of each subject. The fitting uses the spatial correlation to
establish the time when these topographies are present in the data (Pegna et al., 1997).
Thus, at each time point in the individual ERPs, the scalp topography was compared
to the segmentation maps of grand averages and labeled according to the one with
which is the best correlated. This allowed us to quantify how much of the global
explained variance (GEV) of one condition is explained by a given segmentation map.
Repetitive measures ANOVA were performed on the GEV using conditions (face,
scramble and tone only) and percetive-imagery as within-subject factors (2x3
ANOVA).
All electrophysiological analysis (ERP‟s, topography, segmentation) were
performed by using Cartool software (Denis Brunet, Brain Mapping Group, Geneva).
RESULTS
Behavioral results
Subjects show learning on the discrimination tasks with training of the three
different categories of associations through an improvement in RT with training
(p=0.05, t=2.16, paired t-test) and also an increase in the number of correct answers
(p=10-5, t=-6.21, paired t-test), reflecting learning of the precise associations (Figure 3).
The mean reaction time for the correct answers was 436.62 (standard deviation
44.16) for the first block and 422.62 (standard deviation 49.76) for the second block.
The mean number of correct answers was 39 in the first block as compared to 51.50 in
the second block.
108
a.
b.
Figure 3. Behavioral results. Increased performance with training
of the audio-visual associations a. decreased reaction times with training;
b. increased accuracy of response with training
ELECTROPHYSIOLOGICAL RESULTS
ERPs
Perception condition
N170 amplitude was larger for faces and scrambled stimuli in the right hemisphere
(contacts PO8 and P8) than in the left hemisphere (contacts PO7 and P7; paired t-test,
t=2.94, p=0.05) which is in agreement with literature data (Rossion 2008).
Between the two right hemisphere contacts (P8 and PO8), N170 amplitude for
faces was stronger for the PO8 contact (paired t-test, t=3.02, p=0.01). No significant
difference was found for the scrambled faces between these two contacts. For the next
analysis steps, we will refer to the PO8 contact.
At the chosen contact, faces elicited a larger N170 than scrambled stimuli (mean
amplitude for faces: -12.68 μV; mean amplitude for scrambled stimuli: - 4.90 μV, paired
t-test, t=7.48, p=10-5), consistent with reports in the literature (Bentin et al 1996).
N170 peak latency was similar for the faces and scrambled stimuli (mean latency
for faces: 157.18 ms; scrambled: 153.09 ms).
109
a.
left
b.
right
P7
P8
uv
F ace
uv
6
6
4
2
4
2
-2
-4
-2
-4
-6
1.75
PO7
-6
P7
0
500
4
2
-2
-4
-2
-4
-6
-6
500
0
6
4
2
0
P8
ms
ms
6
PO8
500
S cramble
-1.75
ms
0
P O7
ms
500
PO7
P O8
PO8
P7
P8
Figure 4. a. ERPs for faces (black) and scrambled stimuli (red) for each of the two
contacts from the left and right hemisphere; b. Overview of the contacts distribution
over the scalp; the chosen temporo-occipital contacts are highlighted.
Mental imagery condition
At the chosen contact (PO8), imagined faces elicited a N170-like negative
deflection peaking at 190.09 ms after the tone onset with a mean amplitude of -3.71 μV
(Figure 5). Imagined scrambled stimuli also triggered a negative potential peaking at
189.55 ms post tone onset with a peak amplitude of -3.97 μV.
+
PO8
100
200
300
400
500 ms
PO8
100
200
300
400
500
ms
Figure 5. ERP’s for faces (up) ad scramble (bottom) in the imaginary condition
and the corresponding map configuration at maximum of amplitude
A 2x2 ANOVA was performed for the amplitude and latency of the components,
with condition (perception, imagery) and stimulus (face, scrambled faces) as within
subject factors. The results for the amplitude, showed a main effect of condition
110
(F(1.10)= 12.41, p=0.06) and stimulus (F(1, 10)= 41.89, p=10-5), and an interaction
condition by stimulus, F(1,10) = 52.34, p = 10-5). The modulation for scrambled stimuli
was similar in the perception and imagery condition (two-tailed, paired t-test, t=-0.74,
p=0.48) while it differed for faces between perception and imagery condition (twotailed, paired t-test, t=5.23, p=0.0003).
For the latencies of this component, a 2x2 ANOVA showed a phase effect, F(1,
10)=42.40, p=10-5, which was due to triggered modulations appearing later than the
perceptual representations; post-hoc comparisons were significant for both faces (t=
6.91, p=10-5) and for scramble (t=4.65, p=0.01).
faces
uv
6
4
2
-2
-4
-6
ms
-200
0
200
400
scramble
uv
6
4
2
-2
-4
ms
-6
-200
0
200
400
tone alone
uv
6
4
2
-2
-4
ms
-6
-200
0
200
400
Figure 6. Grand-average ERPs for the three visual stimulations
(faces, scrambled, none) and for the two conditions
(perception in black, imagery in red) at the PO8 contact
Scalp topography analysis
A segmentation was performed on the grand averages for the three stimulus
categories during the perception condition (faces, scrambled, none) for the whole
post-origin period (0-800 ms). Figure 7 shows the temporal distribution of the maps,
111
as well as the voltage topography for the 4 maps that dominated during periods with
highest global field power (GFP). According to the timing and to the topography of
the voltage distribution, map number 4 corresponds to the P100 (temporo-occipital
positivity peaking starting around 70 ms post stimulus and ending around 130 ms post
stimulus); map number 5 corresponds to N170 (temporo-occipital negativity, starting
around 130 ms post faces presentation and ending around 193 ms post face
presentation, with a maximum around 157 ms for faces, which corresponds to the
found peak latency; for the scrambled stimuli, the map 5 starts at 115 ms and it ends at
183 ms). Map 5 is not found in the no-stimuli condition (tone alone). Map 7 is
characterized by a temporo-occipital positivity, starting around 181 ms post stimulus
and ending around 368 ms post stimulus, with a maximum around 250 ms. Map 6
ends starts around 138 ms post stimulus and end around 267 ms. Map 6 is found only
in the tone alone condition. Map 4 (P100-like), map 5 (N170-like) and map 7 (P250like) appeared in both face and scrambled conditions.
Figure 7. a.The results of the segmentation analysis performed on grand averages
of the three categories of stimuli (faces, scrambled stimuli, none) in the perception
condition. b. Display of the voltage configuration for the most representative maps
To test whether the topography of the N170 could also be found during the
imagery condition, we assessed how much of the variance was explained by map 5
(corresponding to the N170) in each subject for each stimulus condition. Thus, map 5
found for the average ERP was „back-fitted‟ at the single individual level. We used a
112
repeated measure ANOVA of global explained variance (GEV) to evaluate the
contribution of this map across conditions (perception and imagery) and for the
different stimuli (face, scrambled, none). Figure 8a presents the results of the statistical
back fitting across subjects for the map 5. As expected, we found a main effect of
condition (F(1,10)=22.87, p=0.0002), suggesting that the map corresponding to N170 is
differentially present between the two conditions, namely it is more present in the
condition perception as compared to imagery; we also found a main effect of stimulus
(F(1,10)= 5.67, p=0.03), showing that the map was more present in the face condition
than for scrambled and tone alone (t-tests comparing faces to scramble and tone alone
respectively for each condition: t=3.17, p=0.01 when comparing faces to scramble in
the perceptual condition, t=-4.72, p=0.001 for tone in the perceptual condition; t=1.97,
p= 0.03 when comparing face and tone alone in the imagery condition and no statistical
difference when comparing face and scramble stimuli in the imagery condition). The
lack of interaction (F(1,10)= 1.43, p=0.28) shows that the map corresponding to the
N170 presents the same profile (face > scrambled > none) in both perception and
imagery phases (Figure 8b).
Figure 8. a. Global variance explained by map 5 corresponding to N170 across all
subjects in each condition: condition 1, 2, 3 represent the perceptual phase for face,
scramble and tone alone respectively; condition 4, 5, 6 represent the imagery phase for
face, scramble and tone alone respectively. b. Map 5 shows the same trend in both
perception (series 1, in blue) and in imagery condition (series 2, in red)
113
DISCUSSION
We show that the presentation of the three sounds previously associated with
visual stimuli shows a differential modulation of EEG from 150-200 ms post-stimulus,
corresponding to an N170 topography for both faces and scrambled faces.
Specifically, our results also show that the stable neural configuration (functional
microstate) corresponding to the presentation of the face is reactivated during the
presentation of the sound previously associated with the face, and this significantly
more than during the sounds previously associated with the scrambled stimuli or no
stimuli.
Unlike electrode-wise approach that takes into account only one electrode or a
limited set of electrode, the microstate approach that we used here allows exploiting
the distribution of voltage over the whole scalp (Michel et al 2009). Fitting of the map
on the individual ERPs showed that the map was more present in the „face‟ trials
during both perceptual and imagery conditions, thus demonstrating a category-specific
re-activation of large-scale neural configurations corresponding to a very restricted
step (i.e. N170) in the processing of visual stimuli.
We believe that our results may have new implications for the understanding of
the neural processes reflected by the N170. Building a perceptual representation can
be done by feature-binding mechanisms: Bentin et al (2002) have shown that stimuli
that are not normally perceived as face components may trigger face-specific activity
(N170) when primed by a face context. Here, we show that sound stimuli, which do
not normally activate face processing may prompt an N170 after associative learning.
Thus, context (as in Bentin, 2002) or associative learning (as in the present study) may
early steps during visual processing, most probably via top-down modulation (Rauss
2010 and 2009, Pourtois 2008). These results thus highlight the flexibility of early
perceptual mechanisms (here for face processing).
Peigneux et al. (2006) have showed that recent learned information is
spontaneously replayed during the first hours of post-training wakefulness. The posttraining reinstatement of a specific neural configuration as found in our data may also
be interpreted in the context of the offline processing and early reorganization of the
recently acquired memories during wakefulness.
114
SUPPLEMENTARY DATA
VIVIDNESS OF VISUAL IMAGERY QUESTIONNAIRE (VVIQ)
adapted from Marks, 1973
Instructions
Visual imagery refers to the ability to visualize, that is, the ability to form mental
pictures, or to „see in the mind‟s eye‟. Marked individual differences have been found
in the strength and clarity of reported visual imagery and these differences are of
considerable psychological interest.
The aim of this test is to determine the vividness of your visual imagery. The
items of the test will possibly bring certain images to your mind. You are asked to rate
the vividness of each image by reference to the 5-point scale given below. For
example, if your image is "vague and dim" then give it a rating of 4. After each item
write the appropriate number in the box provided. The first box is for an image
obtained with your eyes open and the second box is for an image obtained with your
eyes closed. Before you turn to the items on the next page, familiarize yourself with
the different categories on the rating scale. Throughout the test, refer to the rating
scale when judging the vividness of each image. Try to do each item separately,
independent of how you may have done other items.
Complete all items for images obtained with the eyes open and then return
to the beginning of the questionnaire and rate the image obtained for each item
with your eyes closed.
Try and give your ‘eyes closed’ rating independently of the ‘eyes open’ rating.
The two ratings for a given item may not in all cases be the same.
Rating Scale
The image aroused by an item might be:
 Perfectly clear and as vivid as normal vision rating 1
 Clear and reasonably vivid rating 2
 Moderately clear and vivid rating 3
 Vague and dim rating 4
 No image at all, you only "know" that you are thinking of an object rating 5
115
Eyes Opened
In answering items 1 to 4, think of some relative or friend whom you frequently
see (but who is not with you at present) and consider carefully the picture that comes
before your mind’s eye.
1.
2.
3.
4.
The exact contour of face, head, shoulders and body.
Characteristic poses of head, attitudes of body etc.
The precise carriage, length of step, etc. in walking.
The different colors worn in some familiar clothes.
Visualize the rising sun. Consider carefully the picture that comes before your
mind’s eye.
5.
6.
7.
8.
The sun is rising above the horizon into a hazy sky.
The sky clears and surrounds the sun with blueness.
Clouds. A storm blows up, with flashes of lightening.
A rainbow appears.
Think of the front of a shop which you often go to. Consider the picture that
comes before your mind’s eye.
9. The overall appearance of the shop from the opposite side of the road.
10. A window display including colors, shape and details of individual items
for sale.
11. You are near the entrance. The color, shape and details of the door.
12. You enter the shop and go to the counter.
The counter assistant serves you.
Money changes hands.
Cui et al – Supplementary Material –3
Finally, think of a country scene which involves trees, mountains and a lake.
Consider the picture that comes before your mind’s eye.
13.
14.
15.
16.
The contours of the landscape.
The color and shape of the trees.
The color and shape of the lake.
A strong wind blows on the tree and on the lake causing waves.
116
Eyes Closed
In answering items 1 to 4, think of some relative or friend whom you frequently
see (but who is not with you at present) and consider carefully the picture that comes
before your mind’s eye.
1.
2.
3.
4.
The exact contour of face, head, shoulders and body.
Characteristic poses of head, attitudes of body etc.
The precise carriage, length of step, etc. in walking.
The different colors worn in some familiar clothes.
Visualize the rising sun. Consider carefully the picture that comes before your
mind’s eye.
5.
6.
7.
8.
The sun is rising above the horizon into a hazy sky.
The sky clears and surrounds the sun with blueness.
Clouds. A storm blows up, with flashes of lightening.
A rainbow appears.
Think of the front of a shop which you often go to. Consider the picture that
comes before your mind’s eye.
9. The overall appearance of the shop from the opposite side of the road.
10. A window display including colors, shape and details of individual items
for sale.
11. You are near the entrance. The color, shape and details of the door.
12. You enter the shop and go to the counter.
The counter assistant serves you.
Money changes hands.
Cui et al – Supplementary Material –3
Finally, think of a country scene which involves trees, mountains and a lake.
Consider the picture that comes before your mind’s eye.
13.
14.
15.
16.
The contours of the landscape.
The color and shape of the trees.
The color and shape of the lake.
A strong wind blows on the tree and on the lake causing waves.
117
REFERENCES
Allison, T., H. Ginter, et al. (1994). "Face recognition in human extrastriate
cortex." J Neurophysiol 71(2): 821-5.
Allison, T., G. McCarthy, et al. (1994). "Human extrastriate visual cortex and the
perception of faces, words, numbers, and colors." Cereb Cortex 4(5): 544-54.
Allison, T., A. Puce, et al. (1999). "Electrophysiological studies of human face
perception. I: Potentials generated in occipitotemporal cortex by face and non-face
stimuli." Cereb Cortex 9(5): 415-30.
Bentin, S., T. Allison, et al. (1996). "Electrophysiological Studies of Face Perception
in Humans." J Cogn Neurosci 8(6): 551-565.
Dierks, T., V. Jelic, et al. (1997). "EEG-microstates in mild memory impairment
and Alzheimer's disease: possible association with disturbed information processing." J
Neural Transm 104(4-5): 483-95.
Eimer, M. (2000). "Event-related brain potentials distinguish processing stages
involved in face perception and recognition." Clin Neurophysiol 111(4): 694-705.
Eimer, M. (2000). "The face-specific N170 component reflects late stages in the
structural encoding of faces." Neuroreport 11(10): 2319-24.
Haxby, J. V., E. A. Hoffman, et al. (2000). "The distributed human neural system
for face perception." Trends Cogn Sci 4(6): 223-233.
Ishai, A., L. G. Ungerleider, et al. (2000). "Distributed neural systems for the
generation of visual images." Neuron 28(3): 979-90.
Ishai, A., L. G. Ungerleider, et al. (2000). "The representation of objects in the
human occipital and temporal cortex." J Cogn Neurosci 12 Suppl 2: 35-51.
Lehman,
D.
(1987).
Principles
of
spatial
analysis.
Handbook
of
electroencephalography and clinical neurophysiology, revised series. R. A. Gevins A.
Amsterdam, Elsevier: 309-354.
Maquet, P., S. Laureys, et al. (2000). "Experience-dependent changes in cerebral
activation during human REM sleep." Nat Neurosci 3(8): 831-6.
McCarthy, G., A. Puce, et al. (1999). "Electrophysiological studies of human face
perception. II: Response properties of face-specific potentials generated in
occipitotemporal cortex." Cereb Cortex 9(5): 431-44.
Michel, M. C., Koenig,T, Brandeis, D, Gianotti, L, Wakermann, J (2009). Electrical
Neuroimaging, Cambridge University Press.
Nobre, A. C., T. Allison, et al. (1994). "Word recognition in the human inferior
temporal lobe." Nature 372(6503): 260-3.
O'Craven, K. M. and N. Kanwisher (2000). "Mental imagery of faces and places
activates corresponding stiimulus-specific brain regions." J Cogn Neurosci 12(6): 1013-23.
Pascual-Marqui, R. D., C. M. Michel, et al. (1995). "Segmentation of brain electrical
activity into microstates: model estimation and validation." IEEE Trans Biomed Eng
42(7): 658-65.
Pegna, A. J., A. Khateb, et al. (1997). "Unraveling the cerebral dynamics of mental
imagery." Hum Brain Mapp 5(6): 410-21.
Peigneux, P., S. Laureys, et al. (2004). "Are spatial memories strengthened in the
human hippocampus during slow wave sleep?" Neuron 44(3): 535-45.
Peigneux, P., P. Orban, et al. (2006). "Offline persistence of memory-related
cerebral activity during active wakefulness." PLoS Biol 4(4): e100.
Picton, T. W., S. Bentin, et al. (2000). "Guidelines for using human event-related
potentials to study cognition: recording standards and publication criteria."
Psychophysiology 37(2): 127-52.
Pourtois, G., K. S. Rauss, et al. (2008). "Effects of perceptual learning on primary
visual cortex activity in humans." Vision Res 48(1): 55-62.
118
Puce, A., T. Allison, et al. (1995). "Face-sensitive regions in human extrastriate
cortex studied by functional MRI." J Neurophysiol 74(3): 1192-9.
Puce, A., T. Allison, et al. (1999). "Electrophysiological studies of human face
perception. III: Effects of top-down processing on face-specific potentials." Cereb Cortex
9(5): 445-58.
Rasch, B., C. Buchel, et al. (2007). "Odor cues during slow-wave sleep prompt
declarative memory consolidation." Science 315(5817): 1426-9.
Rauss, K., S. Schwartz, et al. "Top-down effects on early visual processing in
humans: A predictive coding framework." Neurosci Biobehav Rev.
Rauss, K. S., G. Pourtois, et al. (2009). "Attentional load modifies early activity in
human primary visual cortex." Hum Brain Mapp 30(5): 1723-33.
Rossion, B. and C. Jacques (2008). "Does physical interstimulus variance account
for early electrophysiological face sensitive responses in the human brain? Ten lessons
on the N170." Neuroimage 39(4): 1959-79.
Rudoy, J. D., J. L. Voss, et al. (2009). "Strengthening individual memories by
reactivating them during sleep." Science 326(5956): 1079.
Seeck, M. and O. J. Grusser (1992). "Category-related components in visual
evoked potentials: photographs of faces, persons, flowers and tools as stimuli." Exp
Brain Res 92(2): 338-49.
Suzuki, W. A. (2007). "Integrating associative learning signals across the brain."
Hippocampus 17(9): 842-50.
Tomita, H., M. Ohbayashi, et al. (1999). "Top-down signal from prefrontal cortex in
executive control of memory retrieval." Nature 401(6754): 699-703.
Wilson, M. A. and B. L. McNaughton (1994). "Reactivation of hippocampal
ensemble memories during sleep." Science 265(5172): 676-9.
119
Instrumental modulation of
electrophysiological features
of sleep
Experiment
4.
CONTEXT
It has been shown that sleep physiological features may be externally
manipulated by applying rhythmic sensory inputs. One explanation would be that the
monotonous external information induces brain synchrony, which may modulate
sleep/arousal systems (Sato 2007; Gao 2009).
One of the most common and archaic way to facilitate sleep is rocking: we
irresistibly fall asleep in a rocking-chair and, since immemorial times, we cradle our
babies to sleep. Yet, no clear physiological explanation has been provided to this
observed effect.
In our study, we aimed at answering to this question, by studying if a low and
repeated stimulation mimicking rocking could modify sleep architecture in humans,
using a swinging-bed during afternoon naps.
We used standard scalp EEG recordings in healthy young people, while they
were performing a nap in a swinging bed. The bed (Figure1) was specifically designed
to allow slow and repetitive horizontal movements (the frequency which provided the
best sleep conditions was of 0.25 Hz). Subjects were recorded in two conditions: bed
moving and bed stationary. Sleep physiological parameters were recorded and
quantified in the two conditions.
SUMMARY OF RESULTS
We observed a faster transition to sleep in each and every subject in the
swinging condition, a result which supports the intuitive notion of facilitation of sleep
associated with this procedure.
120
Figure 1. a. Experimental «hammock», motorized bed producing
a linear accelerated lateral movement; b. Detail of the rotor
Surprisingly however, we also found a dramatic boosting of certain types of
sleep-related oscillations (sleep spindles, occurring during sleep stage 2).
Importantly, we found that some of these EEG changes lasted throughout the
entire duration of the swinging condition.
Importantly, we found that some of these EEG changes lasted throughout the
entire duration of the swinging condition.
CONCLUSION
From these findings, we conclude that rocking beneficial impact on sleep is
related to a synchronizing effect on brain circuits implicated in sleep, leading not only
to a faster transition from waking to sleep but also to a consolidation into more
profound sleep. This is in agreement with very recent data, proving that spindle rate
may predict sleep stability (Dang-Vu et al, 2010).
121
Furthermore, we advance the hypothesis that the influence of rocking on sleep
rhythms may be similar to that observed using transcranial direct current stimulation:
Marshall, Helgadottir et al. 2006 and Marshall 2007 have shown by applying an
external electrical stimulation over the frontal regions that inducing cortical slow
oscillations facilitates the retention of hippocampus-dependent declarative memories.
We believe that our study will motivate the development of new nonpharmacological strategies to cope with sleep initiation (or sleep maintenance)
difficulties, which may be important in a sleep deprived society.
Article submitted in Current Biology
Rocking to sleep. Rocking synchronizes sleep brain waves
Irina Constantinescu1*, Laurence Bayer1*, Stephen Perrig2, Julie Vienne3, Pierre-Paul
Vidal4, Michel Mühlethaler1*, Sophie Schwartz1,5*
*
These authors contributed equally to the work
1
Department of Neuroscience, University of Geneva, Switzerland
2
Sleep Laboratory, Geneva University Hospital, Switzerland
3
Center for Integrative Genomics, University of Lausanne, Switzerland
4
Centre National de la Recherche Scientifique, Unité Mixte de Recherche 8194-
Université Paris Descartes, France
5
Swiss Center for Affective Sciences, University of Geneva, Switzerland
Corresponding authors:
Sophie Schwartz and Michel Muhlethaler,
University Medical Center, Dept. of Neuroscience,
Michel-Servet 1, 1211 Geneva 4, Switzerland
Email: sophie.schwartz@unige.ch and michel.muhlethaler@unige.ch
Acknowledgments: This work was supported by the Swiss National Science
Foundation. We thank A. Borbély, P. Franken, C. Frith, B. E. Jones, C. Leonard, and
M. Tafti for their comments on previous versions of the manuscript.
122
Why do we cradle babies or irresistibly fall asleep in a hammock? Although such
simple behaviors are common across cultures and generations, the nature of the link
between rocking and sleep is poorly understood [1, 2]. Here we aimed at
demonstrating that swinging can modulate physiological parameters of human sleep.
To this end, we chose to study sleep during an afternoon nap using polysomnography
and EEG spectral analyses. We show that lying on a slowly rocking bed (0.25 Hz)
facilitates the transition from waking to sleep, and increases the duration of stage N2
sleep. Rocking also induces a sustained boosting of slow oscillations and spindle
activity. It is proposed that sensory stimulation associated with a swinging motion
exerts a synchronizing action in the brain that reinforces endogenous sleep rhythms.
These results thus provide a scientific support to the traditional belief that rocking can
sooth our sleep.
In the present study, we asked twelve healthy male volunteers (22-38 year old)
to nap on a custom-made bed or “experimental hammock” that could either remain
stationary or rock gently (0.25 Hz, Figure 1A). All the participants were good
sleepers, non-habitual nappers with no excessive daytime sleepiness; they had low
anxiety level. Sleep quality and quantity were assessed by questionnaires and
actimetry recordings. The experimental procedure involved spending two 45-minute
afternoon naps (2:30 to 3:15 PM): one with the bed stationary, one with the bed put in
motion (condition order randomized). The motion parameters were set to stimulate
vestibular and proprioceptive sensory systems, without causing nausea or any
entrainment of cardiac rhythm. In both conditions the naps were spent in complete
darkness in a controlled room temperature (21° ± 1° C) and the level of auditory
stimulation was around 37 dB. During both sessions, polysomnography data were
recorded continuously. Sleep stages and sleep spindles were visually identified by two
experienced scorers, blind to the experimental conditions. We also performed spectral
analysis (FFT routine) using the frontocentral derivation (Fz). The data from two
participants were excluded from the final analyses; all statistical tests were performed
using paired 2-tailed t-tests (Supplemental Information).
All participants had a good quality of sleep and regular hours of night sleep (mean
± s.e.m.; 7.32 ± 0.78 h), as assessed by self-rated sleep questionnaires over 3
consecutive nights preceding each experimental day, with no difference for these
measurements between stationary and swinging conditions. Wrist actimetry did not
123
show any difference in sleep efficiency between conditions (mean ± s.e.m.; swinging:
86.63 ± 1.95%; stationary: 86.71 ± 1.23%). Participants were more alert (on visual
analogue scale) after napping than before [ANOVA, 2 conditions (swinging, stationary)
by 2 time-points (before, after nap); main effect of time-point: F(1,9) = 8.4, P = 0.018;
no effect of condition, no interaction]. Eight participants rated the swinging condition as
“more pleasant” than the stationary condition; for one participant both sessions were
equally pleasant and for one participant the stationary condition was more pleasant.
We found that rocking accelerated sleep onset, as evidenced by a shorter
duration of stage N1 sleep and a reduction of stage N2 latency, compared to the
stationary condition (Table S1). Rocking also affected deeper sleep stages by
increasing the duration of stage N2 sleep and the mean spindle density per 30-s epoch
(Table S1, Fig. 1B). The latter effect persisted throughout the entire duration of stage
N2 (Fig. 1C). All these modifications were observed in each and every participant (all
P<0.009; Table S1). In the only previous study investigating the effect of rocking on
sleep, Woodward et al. [1] found no consistent modulation for the percentage of stage
1 sleep and an overall reduction of the percentage of stage 2 sleep during the motion
condition. However, these data were computed over whole nights of sleep recordings,
and did not address the question of whether vestibular/somatosensory inputs influence
the transition from wakefulness to sleep (i.e., stage 1 and 2 sleep early in the night
after sleep onset), in contrast to the present study.
Rocking also increased EEG power within the slow oscillation band (<1 Hz), as
well as in adjacent delta 1 and delta 2 bands (Fig. 1D), as revealed by EEG spectral
analyses of frontocentral electrode (Fz) during stage N2 [3]. A significant increase of
EEG power within spindle frequency bands was also observed. Together these results
show that rocking induces a speeded transition to an unambiguous sleep state, and
may enhance sleep by boosting slow oscillations and spindle activity.
How can we explain that rocking may accelerate wake-sleep transition and
promote sleep consolidation? Below we propose three main plausible neural
mechanisms. First, because vestibular/somatosensory pathways have anatomical links
with structures implicated in emotions such as the amygdala [4] and because the
amygdala affects the regulation of sleep-wake states [5], faster sleep onset could be
due to a “relaxing” feeling associated with the rocking condition, which most of our
participants (8 out of 10) found pleasant. Second, rhythmic vestibular/somatosensory
124
inputs associated with rocking may modulate sleep-wake centres via direct or indirect
connections between sensory systems and hypothalamic [6] or brainstem areas [7].
Third, sensory inputs could affect the synchrony of neural activity within thalamocortical networks because both somatosensory and vestibular inputs send direct
projections to thalamic nuclei [8]. Consistent with this view, slow rhythmic cortical
stimulation was recently found to increase EEG slow oscillations and spindles [3, 9],
which are both hallmarks of deep sleep. The latter hypothesis of an influence on
neural synchrony fits best the present observation that rocking does not only facilitate
sleep onset but has a persistent effect on brain oscillations and spindles. Recent
evidence that increased spindle activity protects sleep against disruptive stimuli is in
agreement with this interpretation [10].
We propose that rhythmic rocking enhances synchronous activity within
thalamo-cortical networks, which in turn could promote the onset of sleep and its
maintenance. The use of rocking to sooth sleep thus belongs to our repertoire of
adaptive behaviours in which a natural mechanism of sleep (i.e. thalamo-cortical
synchronization) has been harnessed in the simplest manner since immemorial times.
Figure 1 Legend. (A) Schema of bed rocking. (B) Decreased stage N1 and increased
stage N2 during rocking compared to stationary condition. ** P < 0.001.
(C) Evolution of spindle density (mean of spindle per 30-s epoch) during stage N2,
n: number of subjects. ** P < 0.005, * P < 0.05.
(D) Spectral profile during stage 2. * P< 0.05, ** P< 0.005.
125
SUPPLEMENTAL DATA
Table S1. Sleep parameters in each experimental condition (n=10, mean ± s.e.m)
Latency to stage N1 (min)
Latency to stage N2 (min)*
Stage N1 (min)*
Stage N2 (min)*
Stage N3 (min)
Total sleep time (min)
Total sleep period (min)
Awakenings nb
Sleep efficiency
Total spindles number*
Bed stationary
Bed swinging
8.75 ± 2.45
7.75 ± 1.48
8.85± 2.05
12.4 ± 1.61
12.2 ± 1.94
0.3 ± 0.3
24.9 ± 3.07
34.05 ± 2.61
5.45 ± 1.08
73.1 ± 5.74
29.15 ± 6.27
5.35± 1.56
8.05 ± 1.15
17.8 ± 1.38
1.1 ± 0.52
26.95 ± 1.6
36.85 ± 8.24
4.85 ± 0.9
73.13 ± 2.79
60.15 ± 10.7
Note:
mean alpha frequency of the population was within normal ranges: 9.68 +/- 0.77 Hz
*significant difference between bed stationary and swinging, t-student, d.f.=9,
all P < 0.001 except for latency to stage N2 P < 0.01
SUPPLEMENTAL EXPERIMENTAL PROCEDURES
Participants
Twelve healthy male volunteers gave informed consent to participate in this study
according to the ethical regulations of the Geneva University Hospitals. Female
participants were not included because of the effects of menstrual cycle on EEG and
sleep [1] All participants included in the study were good sleepers, with normal and
regular sleep-wake habits, and were non-habitual nappers (taking a short nap less than
once per week in the last two years). None of them suffered from excessive daytime
sleepiness (as assessed by Epworth sleepiness scale [2]. They had no psychiatric or
neurological history, had never suffered from any vestibular disorder, and did not take
any medication during the whole experimental period. The data from two participants
had to be excluded from the analyses: one because of elevated anxiety before one
experimental session, and the other one because of technical problems with the EEG
recording. The remaining 10 participants had a mean age of 30.1 (range of 22-38 years)
and had low anxiety level, as assessed by State Trait Anxiety Inventory [3] (mean score
± s.d.; 32.56 ± 3.68). Sleep quality and quantity were assessed by self-rated sleep
questionnaires over 3 consecutive nights before each experimental session. Wrist
actimetry was also recorded during the last night preceding each experimental session.
126
Protocol
The experimental procedure consisted in a 45-minute afternoon nap spent in a
custom-made bed that could either remain stationary or rock gently (Figure 1A in
main text). This bed was suspended by four metal rods to a metallic frame and
connected to an electrical motor that produced sinusoidally-modulated horizontal
accelerations. The electrical engine and the experimental bed were built to be silent
(adding only 2.5 dB to background noise). During pilot testing, we selected a set of
motion parameters that generated stimulation while minimizing physical discomfort.
The optimal parameters were obtained for total lateral excursion of 10.5 cm amplitude
at the level of the bed and a swinging frequency of 0.25 Hz. Data recorded with an
accelerometer (sampling frequency: 500 Hz; 3D motion tracker; Xsens MTx,
Netherlands) on the bed and on the participants‟ head measured a peak horizontal
acceleration of 0.1 m/s2 (g load = 0.01) and confirmed a negligible (non-detectable)
vertical acceleration.
The nap protocol consisted of two sessions: one with the bed stationary, one with
the bed put in motion. The order of the experimental conditions was randomized
across participants and the two sessions were at least one week apart. Time in bed
from lights off (2:30 PM) to lights on (3:15 PM) was controlled by the experimenters.
The naps were spent in complete darkness and the temperature of the sleep room was
controlled (21° ± 1° C). During the stationary condition, the motor was turned on, but
disconnected from the bed, so that both conditions involved the same level of auditory
stimulation (37 dB in each condition). In these conditions, auditory input can be
excluded. Visual input can indeed discounted since in both controls and experimental
conditions subjects hold their eyes closed and lights were off. During both sessions,
polysomnography data were recorded continuously with a sampling rate of 1024 Hz
(Vitaport3, TEMEC, Netherlands). The montage included 10 scalp electrodes placed
according to International 10-20 system (Fz, Cz, C3, C4, Pz, Oz, O1, O2, A1, A2),
plus electrooculogram and electromyogram contacts.
EEG data analyses
Sleep polysomnography was scored over 30s epochs interval, according to
standard criteria [4], by two experienced scorers blind to the experimental conditions.
127
Several sleep parameters during the naps were determined: latencies to stage N1 and
N2 (from lights-off), time and percentage of each sleep stage, total sleep time (TST;
sum of the time spent in different sleep stages), total sleep period (TSP; total time
from sleep onset to final awakening, including intra-sleep wake intervals), sleep
efficiency (defined as TST/TSP x 100) and number of intra-sleep awakenings (Table
S1). sleep spindles were visually quantified at Cz contact referenced against mastoid
channels, based on their typical fuiform morphology, frequency of 11-16 Hz,
[4] (Table S1). EEG spectral
analysis was applied at frontocentral site (fz) on stage N2. Fast Fourier transform was
performed on average, 25% overlapping, 10 s windows, with a Hanning window, free
of artefacts resulting in a frequency resolution of 0.1 Hz. Values below 0.6 Hz and
above 15 Hz were omitted. Analyzes were performed using the Cartool software by
Denis Brunet (http://brainmapping.unige.ch/Cartool.htm). Mean power in the
following bands was calculated: slow oscillations (0.6-1 Hz), delta 1 (1-2 Hz), delta 2
(2-4 Hz), theta (4-8 Hz), slow spindles (8-12 Hz) and fast spindles (12-15 Hz).
Frequency bands were chosen as previously reported [5]. Comparisons between
rocking and stationary conditions were performed using paired 2-tailed t-tests.
References and Notes
1. Woodward, S., Tauber, E.S., Spielmann, A.J., and Thorpy, M.J. (1990). Effects
of otolithic vestibular stimulation on sleep. Sleep 13, 533-537.
2. Krystal, A.D., Zammit, G.K., Wyatt, J.K., Quan, S.F., Edinger, J.D., White, D.P.,
Chiacchierini, R.P., and Malhotra, A. (2010). The effect of vestibular stimulation in a
four-hour sleep phase advance model of transient insomnia. J Clin Sleep Med 6, 315321.
3. Marshall, L., Helgadottir, H., Molle, M., and Born, J. (2006). Boosting slow
oscillations during sleep potentiates memory. Nature 444, 610-613.
4. Carmona, J.E., Holland, A.K., and Harrison, D.W. (2009). Extending the
functional cerebral systems theory of emotion to the vestibular modality: a systematic
and integrative approach. Psychol Bull 135, 286-302.
5. Chou, T.C., Bjorkum, A.A., Gaus, S.E., Lu, J., Scammell, T.E., and Saper, C.B.
(2002). Afferents to the ventrolateral preoptic nucleus. J Neurosci 22, 977-990.
6. Horowitz, S.S., Blanchard, J., and Morin, L.P. (2005). Medial vestibular
connections with the hypocretin (orexin) system. J Comp Neurol 487, 127-146.
7. Jones, B.E. (2003). Arousal systems. Front Biosci 8, s438-451.
8. Moruzzi, G. (1972). The sleep-waking cycle. In Reviews of Physiology :
Biochemistry and experimental pharmacology, R.H. Adrian, E. Helmreich, H. Holzer, R.
Young, K. Kramer, O. Kreayer, F. Lynen, P.A. Miescher, H. Rasmussen, A.E. Renold, et
al., eds. (Berlin, Heidelberg, New York: Springer-Verlag), pp. 1-165.
128
9. Massimini, M., Ferrarelli, F., Esser, S.K., Riedner, B.A., Huber, R., Murphy, M.,
Peterson, M.J., and Tononi, G. (2007). Triggering sleep slow waves by transcranial
magnetic stimulation. Proc Natl Acad Sci U S A 104, 8496-8501.
10. Dang-Vu, T.T., McKinney, S.M., Buxton, O.M., Solet, J.M., and Ellenbogen, J.M.
(2010). Spontaneous brain rhythms predict sleep stability in the face of noise. Curr Biol
20, R626-627.
Supplemental References
1. Baker, F.C., and Driver, H.S. (2007). Circadian rhythms, sleep, and the
menstrual cycle. Sleep Med 8, 613-622.
2. Johns, M.W. (1991). A new method for measuring daytime sleepiness: the
Epworth sleepiness scale. Sleep 14, 540-545.
3. Spielberger, C.D. (1983). Manual for the State-Trait Anxiety Inventory, (Palo
Alto, California: Consulting Psychologist Press Inc.).
4. Iber, C., Ancoli-Israel, S., Chesson, A., and Quan, S.F. (2007). The AASM
Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and
Technical Specifications, (Westchester, Illinois: American Academy of Sleep Medicine).
5. Marshall, L., Helgadottir, H., Molle, M., and Born, J. (2006). Boosting slow
oscillations during sleep potentiates memory. Nature 444, 610-613.
129
GENERAL DISCUSSION
The aim of the present work was to study the influence of sleep-wake states on
human memory and underlying neural plasticity by using EEG recordings and
neurological models as parasomnia.
MULTIMODAL APPROACH TO MEMORY CONSOLIDATION
We investigated the physiological and cognitive processes related to the
consolidation of recently acquired knowledge using converging neuroscience and
clinical approaches. We also looked at memory (re)processing across different
vigilance states. Here I show how the integration of these multimodal data (behavior,
brain networks, local field potentials) contributes to our understanding of experiencedependent changes in the human brain.
At the behavioral level, we showed in Experiment 2 that trained sequences of
large movements may be spontaneous replayed during post-training sleep in
parasomnia patients. This study brings the first evidence to our knowledge of an
experience-related behavioral replay in humans. Indeed, overt and spontaneous replay
of waking behavior during sleep had never been previously shown in humans. These
data also support the interpretation, proposed by human neuroimaging studies, that
observed local neural reactivations during sleep in regions previously involved in
some trained waking behavior reflect a replay of the behavior (Maquet 2000). Of
course, during normal sleep, overt motor behavior is inhibited at the level of the
medulla, thus preventing dreams to be enacted. By demonstrating the influence of
recent waking learning experiences on mentation during sleep, this study also provides
some support to recent dream studies that suggested a causal link between dream
content and the consolidation of a recent task (Wamsley 2009).
This study opens up a new, promising perspective for the use of neurological
sleep disorders as models to investigate cognitive processes in humans.
At the local field potential (LFP) level, in Experiment 1, we studied
experience-related neural dynamics in pharmaco-resistant epileptic patients while they
learned to perform sequences of finger movements. We demonstrate that the learning
visuomotor
events
within
sequentially-organized
knowledge
relies
on
the
hippocampus and on distributed pattern of neural activity. We show that hippocampus
130
contributes importantly to assembling and processing sequences of motor episodic
events. These results are in line with very recent views on hippocampal functioning,
and provide further support for a role of the medial temporal lobe in procedural
memory. Our study also suggests that sleep may influence the re-shaping of neural
representations related to learning. Moreover, the single-trial processing approach
(classification method) that we developed allows the study of individual patients
(important for clinical purposes). We believe that the present study may motivate
further translational perspectives on brain dynamics related to learning, from the
individual studies in animals (place cells) to humans. Furthermore, an exciting
extension of the present study would be to apply the same method to study
experience-related replay of sequences of activity during sleep.
In Experiment 3, we studied induced reactivation and reorganization of neural
traces during post-training wakefulness at the scalp level by using high-density
recordings in humans. Specifically, we investigated the learning of audio-visual
associations and related temporal dynamics of high-density EEG topographies. By
using the momentary configuration of functional potential maps and their transition
with time, we show that induced-visual imagery (triggered by sounds previously
associated with a visual image) may reactivate the same neural representations as
when actually presenting the stimulus (visual perception). We also show that the
induced reactivation is specific to the category of audio-visual learned association. To
our knowledge, this is the first study that investigates induced reactivation of maps
specific to visual categories at the milliseconds scale. Our results open important new
perspective for the study of the spontaneous re-instatement of neural patterns during
post-training sleep in humans
Finally,
in
Experiment
4,
we
sought
to
experimentally
induce
neurophysiological changes during sleep. To this end, we built a motorized bed that
produced a slow swinging motion. By using this special bed during afternoon naps we
found that sleep dynamics can be successfully manipulated instrumentally.
Specifically, rocking motion led to a persistent increase in stage 2 sleep and to
increased spindle density. The present results related to sleep physiology motivate a
second study where the impact of sleep modulation on different cognitive aspects (e.g.
memory consolidation) in humans may be explored. Indeed, previous work suggests
that increased spindle density and slow wave activity may enhance memory
131
consolidation (Marshall 2006 and 2007). If we can show that rocking enhances
memory consolidation, this could ultimately have practical implications for education
or rehabilitation purposes.
NOVELTY OF THE TECHNIQUES
What makes brain so special is its organized action in time. The temporal
domain is fundamental when talking about the brain mechanisms (Buzsaki 2006). The
EEG reflects the immediate mass action of neural populations, and thus provides a
particularly direct and accurate window onto human brain functions (Michel
2009). In the present thesis, we used different EEG techniques, i.e., intracranial
recordings, high-density scalp recordings, sleep rythms recordings, to better
understand the neural mechanisms underlying memory consolidation and the putative
role of sleep in this process. This led us to apply or develop different methods to
analyse these electrophysiological data.
While cellular studies in animals provide invaluable data on brain functioning, in
humans, the use of a similar approach using intracranial recordings in populations of
patients, during cognitive studies, is still very limited. A first reason is the small
population of patients requiring presurgical implantation of electrodes, and second
reason is the lack of standard for the analysis of the data recorded. In the world, there
are still only few centers in which this unique approach can be performed. Therefore,
in the present thesis we highlight the value of the intracranial EEG for human
brain mapping. Due to the spatio-temporal resolution (almost “millimetermillisecond” resolution), human intracranial studies approach the “gold standard” of
human brain study (Lachaux, Rudrauf et al. 2003). Another major advantage of
intracranial techniques is that they allow the recording of deep brain structures that are
not normally accessible via scalp recordings.
In the first project (Experiment 1), we developed a multivariate decoding
algorithm that can capture learning-dependent changes in intracranial EEG signal at
the single trial level. By using this multivariate approach, we could establish a causal
link between specific electrode contribution and discrimination power classification.
The developed method is particularly suitable when analyzing unique sets of data
from unique patients, which may be an extremely valuable tool in clinical settings.
132
Based on the present results, we are strongly convinced that the present work may
motivate further development of new ways of exploring the human brain functioning,
The development of standard and robust signal processing methods to analyze
intracranial recordings may also be extremely important for the detection of
pathological features and confer therefore a clinical significance to the obtained results.
The biological rule that information processing in the brain depends on
successions of millisecond-scale and transient functional states, also holds at the
global brain level, analyzed via high-density scalp recordings. For example, while
specific microstates do characterize cerebral dynamics of mental imagery (Pegna
1997), here we show that this approach can reveal the re-presentation of learningrelated associations (Experiment 3).
All in all, from the methodological perspective, we believe that our work
contributes to the understanding of state-dependency of brain information
processing in the millisecond time range.
CONCLUSIONS AND PERSPECTIVES
Experience-related plasticity phenomena undergo different stages during
wakefulness and sleep. Thus, research in the domains of sleep and memory may
reveal new and fundamental cues for our understanding of human brain functioning
and plasticity.
In the present thesis, we used the concept of experience-related neural
reactivations to study brain dynamics subtending memory. We assessed re-shaping of
neural representations during post-training wakefulness (project 3) and sleep (project 1),
either produced spontaneously or induced by presentation of learning-related external
cues. We also assessed experience-related behavioral re-enactment during sleep states
in parasomnia patients. The replay phenomenon is a recently described mechanism in
animals and in humans (McNaughton 1994, Maquet 2000). Since the end of 20th
century, despite growing number of studies in the field, the precise functional relevance
of replay mechanisms in learning and memory processing is not completely understood.
It seems that the replay phenomena reflect more complex processes than believed
(Gupta, van der Meer et al. 2010). We believe that our work may contribute at further
understanding learning-related reorganization of brain activity. We also hope that the
results of our work may prompt new studies in this field.
133
Memory systems may be characterized according to the information input (e.g.
sequential learning), to the type of operation (e.g. forming of new associations) and to
the subtending brain structures (e.g. the hippocampus) (Schacter and Tulving, 1994).
In this thesis, we proposed a multimodal approach to memory functions. This was
done by performing a combination of techniques allowing for temporal and spatial
fine observation of neural activity. This could promote animal-human translational
studies, to assess different aspects of brain plasticity in a unitary way. We believe that
this multimodal approach (c.f. Figure 1 – in the beginning of the experimental part) is
innovative and useful to integrate data from animal studies with data from human
studies in neuroscience.
One of the most enigmatic natural processes is falling asleep. Yet, there are too
few scientifical demonstrations on this topic. We approached this challenging subject,
by analyzing one of the simplest and most well maintained behaviors- rocking to sleep
in human subjects. We showed that a rhythmic rocking enhances synchronous activity
within thalamo-cortical networks, which in turn could promote the onset of sleep and
its maintenance.
One of the nowadays challenges in science is the growing complexity of the
questions and the multitude of ways to explore them. One of the highlights of this PhD
work is the multidisciplinary approach (physics, biology, medicine) used, thus
promoting team-work and joint views, which make science evolve in an integrative
way. We like to believe that the present work opened constructive dialogues and new
collaborations among scientifical teams, into a multidisciplinary network.
The present work represents the finality of a first step of research, as new
projects emerged as a natural continuation of the previous ones throughout the thesis.
The relationship sleep-cognition is pushed further (sleep in a hammock and impact on
cognition, cellular intracranial recordings during sleep in humans and experiencerelated replay of sequences of activity), because once the adventure has started, it goes
on endlessnessly enthusiastic…
***
“When you have completed 95 percent of your journey,
you are only halfway there.”
Japanese Proverb
134
BIBLIOGRAPHY
Addis, D. R., A. T. Wong, et al. (2007). "Remembering the past and imagining the
future: common and distinct neural substrates during event construction and
elaboration." Neuropsychologia 45(7): 1363-77.
Agster, K. L., N. J. Fortin, et al. (2002). "The hippocampus and disambiguation of
overlapping sequences." J Neurosci 22(13): 5760-8.
Ahissar, M. and S. Hochstein (2004). "The reverse hierarchy theory of visual
perceptual learning." Trends Cogn Sci 8(10): 457-64.
Alberini, C. M. (2005). "Mechanisms of memory stabilization: are consolidation and
reconsolidation similar or distinct processes?" Trends Neurosci 28(1): 51-6.
Albouy, G., V. Sterpenich, et al. (2008). "Both the hippocampus and striatum are
involved in consolidation of motor sequence memory." Neuron 58(2): 261-72.
Albouy, G. and P. Maquet (2008). "[Hippocampus, striatum and sequences]." Med
Sci (Paris) 24(11): 921-2.
Alvarez, P. and L. R. Squire (1994). "Memory consolidation and the medial
temporal lobe: a simple network model." Proc Natl Acad Sci U S A 91(15): 7041-5.
Amaral, D. G., N. Ishizuka, et al. (1990). "Neurons, numbers and the hippocampal
network." Prog Brain Res 83: 1-11.
Andersen, P., R. Morris, et al. (2007). The Hippocampus Book. New York, Oxford
University Press, Inc.
Aslam, N., Y. Kubota, et al. (2009). "Translational
maintenance of synaptic plasticity." Mol Syst Biol 5: 284.
switch
for
long-term
Atkinson, R. C., Shiffrin, R.M (1968). Human memory: A proposed system and its
control processes. The psychology of learning and motivation K. W. Spence, Spence, J.T,
New York: Academic Press. 2: 89-195.
Aton, S. J., J. Seibt, et al. (2009). "Mechanisms of sleep-dependent consolidation
of cortical plasticity." Neuron 61(3): 454-66.
Axmacher, N., F. Mormann, et al. (2006). "Memory formation by neuronal
synchronization." Brain Res Rev 52(1): 170-82.
Bailey, D. J., J. Wade, et al. (2009). "Hippocampal lesions impair spatial memory
performance, but not song--a developmental study of independent memory systems in
the zebra finch." Dev Neurobiol 69(8): 491-504.
Backhaus, J. and K. Junghanns (2006). "Daytime naps improve procedural motor
memory." Sleep Med 7(6): 508-12.
Bar, M. (2009). "Predictions: a universal principle in the operation of the human
brain. Introduction." Philos Trans R Soc Lond B Biol Sci 364(1521): 1181-2.
Bar, M. (2009). "The proactive brain: memory for predictions." Philos Trans R Soc
Lond B Biol Sci 364(1521): 1235-43.
Beauregard, M., D. Gold, et al. (1998). "A role for the hippocampal formation in
implicit memory: a 3-D PET study." Neuroreport 9(8): 1867-73.
Bellgowan, P. S., E. A. Buffalo, et al. (2009). "Lateralized spatial and object
memory encoding in entorhinal and perirhinal cortices." Learn Mem 16(7): 433-8.
Benington, J. H. and H. C. Heller (1995). "Restoration of brain energy metabolism
as the function of sleep." Prog Neurobiol 45(4): 347-60.
Bergson, H. L. (1910). Matter and Memory (Authorized translation by Nancy M.
Paul and W. Scott Palmer). London, Allen.
135
Bernard, F. A., E. T. Bullmore, et al. (2004). "The hippocampal region is involved in
successful recognition of both remote and recent famous faces." Neuroimage 22(4):
1704-14.
Borbely, A. A. (1982). "A two process model of sleep regulation." Hum Neurobiol
1(3): 195-204.
Borbely, A. A. (2001). "From slow waves to sleep homeostasis: new perspectives."
Arch Ital Biol 139(1-2): 53-61.
Bozon, B., S. Davis, et al. (2003). "A requirement for the immediate early gene
zif268 in reconsolidation of recognition memory after retrieval." Neuron 40(4): 695-701.
Brasted, P. J. and S. P. Wise (2004). "Comparison of learning-related neuronal
activity in the dorsal premotor cortex and striatum." Eur J Neurosci 19(3): 721-40.
Brawn, T. P., K. M. Fenn, et al. (2008). "Consolidation of sensorimotor learning
during sleep." Learn Mem 15(11): 815-9.
Bruel-Jungerman, E., S. Davis, et al. (2007). "Brain plasticity mechanisms and
memory: a party of four." Neuroscientist 13(5): 492-505.
Buckner, R. L., J. Logan, et al. (2000). "Cognitive neuroscience of episodic memory
encoding." Acta Psychol (Amst) 105(2-3): 127-39.
Buckner, R. L. (2000). "Neural origins of 'I remember'." Nat Neurosci 3(11): 1068-9.
Burgess, N., E. A. Maguire, et al. (2002). "The human hippocampus and spatial and
episodic memory." Neuron 35(4): 625-41.
Burrell, B. D. and C. L. Sahley (2001). "Learning in simple systems." Curr Opin
Neurobiol 11(6): 757-64.
Buzsaki, G. (1989). "Two-stage model of memory trace formation: a role for
"noisy" brain states." Neuroscience 31(3): 551-70.
Buzsaki, G. (1996). "The hippocampo-neocortical dialogue." Cereb Cortex 6(2):
81-92.
Buzsaki, G. (2006). Rythms of the brain. New York, Oxford University Press.
Buzsaki, G. and J. J. Chrobak (2005). "Synaptic plasticity and self-organization in
the hippocampus." Nat Neurosci 8(11): 1418-20.
Cai, D. J. and T. C. Rickard (2009). "Reconsidering the role of sleep for motor
memory." Behav Neurosci 123(6): 1153-7.
Cai, D. J., T. Shuman, et al. (2009). "Sleep selectively enhances hippocampusdependent memory in mice." Behav Neurosci 123(4): 713-9.
Campbell, I. G., M. J. Guinan, et al. (2002). "Sleep deprivation impairs long-term
potentiation in rat hippocampal slices." J Neurophysiol 88(2): 1073-6.
Cappe, C., G. Thut, et al. (2009). "Selective integration of auditory-visual looming
cues by humans." Neuropsychologia 47(4): 1045-52.
Chadwick, M. J., D. Hassabis, et al. "Decoding individual episodic memory traces in
the human hippocampus." Curr Biol 20(6): 544-7.
Chang, Q. and P. E. Gold (2003). "Switching memory systems during learning:
changes in patterns of brain acetylcholine release in the hippocampus and striatum in
rats." J Neurosci 23(7): 3001-5.
Chklovskii, D. B., B. W. Mel, et al. (2004). "Cortical rewiring and information
storage." Nature 431(7010): 782-8.
Cipolotti, L., T. Shallice, et al. (2001). "Long-term retrograde amnesia...the crucial
role of the hippocampus." Neuropsychologia 39(2): 151-72.
Cirelli, C. and G. Tononi (2004). "Locus ceruleus control of state-dependent gene
expression." J Neurosci 24(23): 5410-9.
136
Clayton, N. S., T. J. Bussey, et al. (2003). "Can animals recall the past and plan for
the future?" Nat Rev Neurosci 4(8): 685-91.
Crist, R. E., W. Li, et al. (2001). "Learning to see: experience and attention in
primary visual cortex." Nat Neurosci 4(5): 519-25.
Cubelli, R. and S. Della Sala (2009). "Memories as re-presentations." Cortex
45(5): 688-92.
Dang-Vu, T. T., M. Desseilles, et al. (2006). "A role for sleep in brain plasticity."
Pediatr Rehabil 9(2): 98-118.
Dave, A. S. and D. Margoliash (2000). "Song replay during sleep
computational rules for sensorimotor vocal learning." Science 290(5492): 812-6.
and
Davis, C. J., S. G. Bohnet, et al. (2007). "Sleep loss changes microRNA levels in
the brain: a possible mechanism for state-dependent translational regulation." Neurosci
Lett 422(1): 68-73.
Davis, S., B. Bozon, et al. (2003). "How necessary is the activation of the
immediate early gene zif268 in synaptic plasticity and learning?" Behav Brain Res
142(1-2): 17-30.
Diekelmann, S., I. Wilhelm, et al. (2009). "The whats and whens of sleepdependent memory consolidation." Sleep Med Rev 13(5): 309-21.
Dierks, T., V. Jelic, et al. (1997). "EEG-microstates in mild memory impairment
and Alzheimer's disease: possible association with disturbed information processing." J
Neural Transm 104(4-5): 483-95.
Doyon, J., P. Bellec, et al. (2009). "Contributions of the basal ganglia and
functionally related brain structures to motor learning." Behav Brain Res 199(1): 61-75.
Doyon, J. and H. Benali (2005). "Reorganization and plasticity in the adult brain
during learning of motor skills." Curr Opin Neurobiol 15(2): 161-7.
Doyon, J., D. Gaudreau, et al. (1997). "Role of the striatum, cerebellum, and
frontal lobes in the learning of a visuomotor sequence." Brain Cogn 34(2): 218-45.
Doyon, J., V. Penhune, et al. (2003). "Distinct contribution of the cortico-striatal
and cortico-cerebellar systems to motor skill learning." Neuropsychologia 41(3): 252-62.
Doyon, J., A. W. Song, et al. (2002). "Experience-dependent changes in cerebellar
contributions to motor sequence learning." Proc Natl Acad Sci U S A 99(2): 1017-22.
Doyon, U. (2002). Functional anatomy of motor skill learning. Neuropsychology of
memory. S. D. Squire LR, The Guilford Press: 225-238.
Dudai, Y. (1989). The Neurobiology of Memory: Concepts, Findings, Trends,
Oxford: Oxford University Press.
Dudai, Y. (2004). "The neurobiology of consolidations, or, how stable is the
engram?" Annu Rev Psychol 55: 51-86.
Dudai, Y. (2009). "Predicting not to predict too much: how the cellular machinery
of memory anticipates the uncertain future." Philos Trans R Soc Lond B Biol Sci
364(1521): 1255-62.
Dudai, Y. and M. Eisenberg (2004). "Rites of passage of the engram:
reconsolidation and the lingering consolidation hypothesis." Neuron 44(1): 93-100.
Dupret, D., J. M. Revest, et al. (2008). "Spatial relational memory requires
hippocampal adult neurogenesis." PLoS One 3(4): e1959.
Durrant, S. and P. A. Lewis (2009). "Memory consolidation: tracking transfer with
functional connectivity." Curr Biol 19(18): R860-2.
Edin, F., T. Klingberg, et al. (2009). "Mechanism for top-down control of working
memory capacity." Proc Natl Acad Sci U S A 106(16): 6802-7.
Eichenbaum, H. (2004). "Hippocampus: cognitive processes and
representations that underlie declarative memory." Neuron 44(1): 109-20.
neural
137
Eichenbaum, H. and P. A. Lipton (2008). "Towards a functional organization of the
medial temporal lobe memory system: role of the parahippocampal and medial
entorhinal cortical areas." Hippocampus 18(12): 1314-24.
Eichenbaum, H. (2006). "Remembering: functional organization of the declarative
memory system." Curr Biol 16(16): R643-5.
Eichenbaum, H. and N. J. Fortin (2009). "The neurobiology of memory based
predictions." Philos Trans R Soc Lond B Biol Sci 364(1521): 1183-91.
Eichenbaum, H. a. C. N. J. (2001). From Conditioning to Conscious Recollection.
New York, Oxford University Press.
Ekstrom, A. D., M. J. Kahana, et al. (2003). "Cellular networks underlying human
spatial navigation." Nature 425(6954): 184-8.
Epp, J. R., M. D. Spritzer, et al. (2007). "Hippocampus-dependent learning
promotes survival of new neurons in the dentate gyrus at a specific time during cell
maturation." Neuroscience 149(2): 273-85.
Ergorul, C. and H. Eichenbaum (2004). "The hippocampus and memory for "what,"
"where," and "when"." Learn Mem 11(4): 397-405.
Eustache, F. and B. Desgranges (2008). "MNESIS: towards the integration of
current multisystem models of memory." Neuropsychol Rev 18(1): 53-69.
Euston, D. R., M. Tatsuno, et al. (2007). "Fast-forward playback of recent memory
sequences in prefrontal cortex during sleep." Science 318(5853): 1147-50.
Everson, C. A., C. D. Laatsch, et al. (2005). "Antioxidant defense responses to sleep
loss and sleep recovery." Am J Physiol Regul Integr Comp Physiol 288(2): R374-83.
Fell, J., E. Ludowig, et al. (2008). "Phase-locking within human mediotemporal lobe
predicts memory formation." Neuroimage 43(2): 410-9.
Finelli, L. A., H. Baumann, et al. (2000). "Dual electroencephalogram markers of
human sleep homeostasis: correlation between theta activity in waking and slow-wave
activity in sleep." Neuroscience 101(3): 523-9.
Fischer, S., M. F. Nitschke, et al. (2005). "Motor memory consolidation in sleep
shapes more effective neuronal representations." J Neurosci 25(49): 11248-55.
Fogel, S. M., C. T. Smith, et al. (2007). "Dissociable learning-dependent changes in
REM and non-REM sleep in declarative and procedural memory systems." Behav Brain
Res 180(1): 48-61.
Fortin, N. J., K. L. Agster, et al. (2002). "Critical role of the hippocampus in
memory for sequences of events." Nat Neurosci 5(5): 458-62.
Foster, D. J. and M. A. Wilson (2006). "Reverse replay of behavioural sequences in
hippocampal place cells during the awake state." Nature 440(7084): 680-3.
Frank, M. G., N. P. Issa, et al. (2001). "Sleep enhances plasticity in the developing
visual cortex." Neuron 30(1): 275-87.
Franken, P. and D. J. Dijk (2009). "Circadian clock genes and sleep homeostasis."
Eur J Neurosci 29(9): 1820-9.
Frankland, P. W. and B. Bontempi (2005). "The organization of recent and remote
memories." Nat Rev Neurosci 6(2): 119-30.
Frankland, P. W. and B. Bontempi (2006). "Fast track to the medial prefrontal
cortex." Proc Natl Acad Sci U S A 103(3): 509-10.
Friedman, D. and R. Johnson, Jr. (2000). "Event-related potential (ERP) studies of
memory encoding and retrieval: a selective review." Microsc Res Tech 51(1): 6-28.
Furmanski, C. S., D. Schluppeck, et al. (2004). "Learning strengthens the response
of primary visual cortex to simple patterns." Curr Biol 14(7): 573-8.
Gabrieli, J. D. (1998). "Cognitive neuroscience of human memory." Annu Rev
Psychol 49: 87-115.
138
Gabrieli, J. D., J. B. Brewer, et al. (1997). "Separate neural bases of two
fundamental memory processes in the human medial temporal lobe." Science
276(5310): 264-6.
Gais, S. and J. Born (2004). "Low acetylcholine during slow-wave sleep is critical
for declarative memory consolidation." Proc Natl Acad Sci U S A 101(7): 2140-4.
Gais, S., M. Molle, et al. (2002). "Learning-dependent increases in sleep spindle
density." J Neurosci 22(15): 6830-4.
Gao, J., K. Wong-Lin, et al. (2009). "Sequential effects in two-choice reaction time
tasks: decomposition and synthesis of mechanisms." Neural Comput 21(9): 2407-36.
Gentner, T. Q. and D. Margoliash (2003). "Neuronal populations and single cells
representing learned auditory objects." Nature 424(6949): 669-74.
Gilbert, C. D., M. Sigman, et al. (2001). "The neural basis of perceptual learning."
Neuron 31(5): 681-97.
Gilboa, A., G. Winocur, et al. (2004). "Remembering our past: functional
neuroanatomy of recollection of recent and very remote personal events." Cereb Cortex
14(11): 1214-25.
Girardeau, G., K. Benchenane, et al. (2009). "Selective suppression of hippocampal
ripples impairs spatial memory." Nat Neurosci 12(10): 1222-3.
Giuditta, A., M. V. Ambrosini, et al. (1995). "The sequential hypothesis of the
function of sleep." Behav Brain Res 69(1-2): 157-66.
Goda, Y. (2008). "Neuroscience: Along memory lane." Nature 456(7222): 590-1.
Gold, P. E. (2008). "Protein synthesis inhibition and memory: formation vs
amnesia." Neurobiol Learn Mem 89(3): 201-11.
Grafton, S. T., J. C. Mazziotta, et al. (1992). "Functional anatomy of human
procedural learning determined with regional cerebral blood flow and PET." J Neurosci
12(7): 2542-8.
Halsband, U. and H. J. Freund (1993). "Motor learning." Curr Opin Neurobiol 3(6):
940-9.
Halsband, U. and R. K. Lange (2006). "Motor learning in man: a review of
functional and clinical studies." J Physiol Paris 99(4-6): 414-24.
Hampstead, B. M. and S. P. Koffler (2009). "Thalamic contributions to anterograde,
retrograde, and implicit memory: a case study." Clin Neuropsychol 23(7): 1232-49.
Hassabis, D. and E. A. Maguire (2009). "The construction system of the brain."
Philos Trans R Soc Lond B Biol Sci 364(1521): 1263-71.
Hasselmo, M. E. (2008). "Temporally structured replay of neural activity in a model
of entorhinal cortex, hippocampus and postsubiculum." Eur J Neurosci 28(7): 1301-15.
Henke, K. "A model for memory systems based on processing modes rather than
consciousness." Nat Rev Neurosci 11(7): 523-532.
Herzog, M. H. and M. Fahle (1997). "The role of feedback in learning a vernier
discrimination task." Vision Res 37(15): 2133-41.
Hikosaka, O., K. Nakamura, et al. (2002). "Central mechanisms of motor skill
learning." Curr Opin Neurobiol 12(2): 217-22.
Hobson, J. A. and E. F. Pace-Schott (2002). "The cognitive neuroscience of sleep:
neuronal systems, consciousness and learning." Nature Reviews Neuroscience 3(9):
679-93.
Hoffman, K. L. and B. L. McNaughton (2002). "Coordinated reactivation of
distributed memory traces in primate neocortex." Science 297(5589): 2070-3.
Hoffman, K. L. and B. L. McNaughton (2002). "Sleep on it: cortical reorganization
after-the-fact." Trends Neurosci 25(1): 1-2.
139
Hoscheidt, S. M., L. Nadel, et al. "Hippocampal activation during retrieval of spatial
context from episodic and semantic memory." Behav Brain Res 212(2): 121-32.
Hotermans, C., P. Peigneux, et al. (2006). "Early boost and slow consolidation in
motor skill learning." Learn Mem 13(5): 580-3.
Huber, R., M. F. Ghilardi, et al. (2006). "Arm immobilization causes cortical plastic
changes and locally decreases sleep slow wave activity." Nat Neurosci 9(9): 1169-76.
Huber, R., M. F. Ghilardi, et al. (2004). "Local sleep and learning." Nature
430(6995): 78-81.
Inoue, S., K. Honda, et al. (1995). "Sleep as neuronal detoxification and
restitution." Behav Brain Res 69(1-2): 91-6.
James, W. (1890). Principles of psychology. New York, Holt
Ji, D. and M. A. Wilson (2007). "Coordinated memory replay in the visual cortex
and hippocampus during sleep." Nat Neurosci 10(1): 100-7.
Jouvet, M. (1998). "Paradoxical sleep as a programming system." J Sleep Res 7
Suppl 1: 1-5.
Jueptner, M., C. D. Frith, et al. (1997). "Anatomy of motor learning. II. Subcortical
structures and learning by trial and error." J Neurophysiol 77(3): 1325-37.
Kali, S. and P. Dayan (2004). "Off-line replay maintains declarative memories in a
model of hippocampal-neocortical interactions." Nat Neurosci 7(3): 286-94.
Kandel, E. R. (2009). "The biology of memory: a forty-year perspective." J
Neurosci 29(41): 12748-56.
Karlsson, M. P. and L. M. Frank (2009). "Awake replay of remote experiences in the
hippocampus." Nat Neurosci 12(7): 913-8.
Karmarkar, U. R. and Y. Dan (2006). "Experience-dependent plasticity in adult
visual cortex." Neuron 52(4): 577-85.
Karni, A., G. Meyer, et al. (1995). "Functional MRI evidence for adult motor cortex
plasticity during motor skill learning." Nature 377(6545): 155-8.
Karni, A., G. Meyer, et al. (1998). "The acquisition of skilled motor performance:
fast and slow experience-driven changes in primary motor cortex." Proc Natl Acad Sci U
S A 95(3): 861-8.
Karni, A. and D. Sagi (1991). "Where practice makes perfect in texture
discrimination: evidence for primary visual cortex plasticity." Proc Natl Acad Sci U S A
88(11): 4966-70.
Kavanau, J. L. (1997). "Memory, sleep and the evolution of mechanisms of
synaptic efficacy maintenance." Neuroscience 79(1): 7-44.
Kemp, A. and D. Manahan-Vaughan (2007). "Hippocampal long-term depression:
master or minion in declarative memory processes?" Trends Neurosci 30(3): 111-8.
Kim, J. J. and M. G. Baxter (2001). "Multiple brain-memory systems: the whole
does not equal the sum of its parts." Trends Neurosci 24(6): 324-30.
Kimble, G. A. (1961). Hilgard and Marquis’ Conditioning and Learning. New York,
Appleton-Century-Crofts.
Kirk, I. J. and J. C. Mackay (2003). "The role of theta-range oscillations in
synchronising and integrating activity in distributed mnemonic networks." Cortex 39(45): 993-1008.
Klemen, J., C. Buchel, et al. (2009). "Perceptual load interacts with stimulus
processing across sensory modalities." Eur J Neurosci 29(12): 2426-34.
Konishi, S., M. E. Wheeler, et al. (2000). "Neural correlates of episodic retrieval
success." Neuroimage 12(3): 276-86.
140
Korman, M., J. Doyon, et al. (2007). "Daytime sleep condenses the time course of
motor memory consolidation." Nat Neurosci 10(9): 1206-13.
Korte, M. (2009). "Neuroscience. Bridging the gap and staying local." Science
324(5934): 1527-8.
Krings, T., R. Topper, et al. (2000). "Cortical activation patterns during complex
motor tasks in piano players and control subjects. A functional magnetic resonance
imaging study." Neurosci Lett 278(3): 189-93.
Krueger, J. M., F. Obal, Jr., et al. (1999). "Why we sleep: a theoretical view of
sleep function." Sleep Med Rev 3(2): 119-29.
Kudrimoti, H. S., C. A. Barnes, et al. (1999). "Reactivation of hippocampal cell
assemblies: effects of behavioral state, experience, and EEG dynamics." J Neurosci
19(10): 4090-101.
Lamme, V. A. and P. R. Roelfsema (2000). "The distinct modes of vision offered by
feedforward and recurrent processing." Trends Neurosci 23(11): 571-9.
Langston, R. F., C. H. Stevenson, et al. "The role of hippocampal subregions in
memory for stimulus associations." Behav Brain Res 215(2): 275-91.
Laureys, S., P. Peigneux, et al. (2001). "Experience-dependent changes in cerebral
functional connectivity during human rapid eye movement sleep." Neuroscience 105(3):
521-5.
Lavenex, P. and D. G. Amaral (2000). "Hippocampal-neocortical interaction: a
hierarchy of associativity." Hippocampus 10(4): 420-30.
Lavie, N. (1995). "Perceptual load as a necessary condition for selective attention."
J Exp Psychol Hum Percept Perform 21(3): 451-68.
LeDoux, J. (2003). "The emotional brain, fear, and the amygdala." Cell Mol
Neurobiol 23(4-5): 727-38.
Lee, A. K. and M. A. Wilson (2002). "Memory of sequential experience in the
hippocampus during slow wave sleep." Neuron 36(6): 1183-94.
Lee, J. L. (2009). "Reconsolidation: maintaining memory relevance." Trends
Neurosci 32(8): 413-20.
Lehmann, D. and C. M. Michel "EEG-defined functional microstates as basic building
blocks of mental processes." Clin Neurophysiol.
Leuner, B. and E. Gould "Structural plasticity and hippocampal function." Annu Rev
Psychol 61: 111-40, C1-3.
Leuner, B., S. Mendolia-Loffredo, et al. (2004). "Learning enhances the survival of
new neurons beyond the time when the hippocampus is required for memory." J
Neurosci 24(34): 7477-81.
Li, W., V. Piech, et al. (2004). "Perceptual learning and top-down influences in
primary visual cortex." Nat Neurosci 7(6): 651-7.
Lisman, J. and A. D. Redish (2009). "Prediction, sequences and the hippocampus."
Philos Trans R Soc Lond B Biol Sci 364(1521): 1193-201.
Loebrich, S. and E. Nedivi (2009). "The function of activity-regulated genes in the
nervous system." Physiol Rev 89(4): 1079-103.
Lorincz, A. and G. Buzsaki (2000). "Two-phase computational model training longterm memories in the entorhinal-hippocampal region." Ann N Y Acad Sci 911: 83-111.
Lorincz, A. and G. Szirtes (2009). "Here and now: how time segments may become
events in the hippocampus." Neural Netw 22(5-6): 738-47.
Louie, K. and M. A. Wilson (2001). "Temporally structured replay of awake
hippocampal ensemble activity during rapid eye movement sleep." Neuron 29(1): 145-56.
Malenka, R. C. and R. A. Nicoll (1999). "Long-term potentiation--a decade of
progress?" Science 285(5435): 1870-4.
141
Manns, J. R. and H. Eichenbaum (2006). "Evolution of declarative memory."
Manns, J. R., R. O. Hopkins, et al. (2003). "Recognition memory and the human
hippocampus." Neuron 37(1): 171-80.
Maquet, P. (2001). "The role of sleep in learning and memory." Science
294(5544): 1048-52.
Maquet, P., S. Laureys, et al. (2000). "Experience-dependent changes in cerebral
activation during human REM sleep." Nat Neurosci 3(8): 831-6.
Maquet, P., S. Laureys, et al. (2003). "Festina lente: evidences for fast and slow
learning processes and a role for sleep in human motor skill learning." Learn Mem 10(4):
237-9.
Maquet, P., S. Schwartz, et al. (2003). "Sleep-related consolidation of a visuomotor
skill: brain mechanisms as assessed by functional magnetic resonance imaging." J
Neurosci 23(4): 1432-40.
Maret, S., S. Dorsaz, et al. (2007). "Homer1a is a core brain molecular correlate of
sleep loss." Proc Natl Acad Sci U S A 104(50): 20090-5.
Markham, J. A. and W. T. Greenough (2004). "Experience-driven brain plasticity:
beyond the synapse." Neuron Glia Biol 1(4): 351-363.
Marrone, D. F. (2007). "Ultrastructural plasticity associated with hippocampaldependent learning: a meta-analysis." Neurobiol Learn Mem 87(3): 361-71.
Marshall, L., H. Helgadottir, et al. (2006). "Boosting slow oscillations during sleep
potentiates memory." Nature 444(7119): 610-3.
Martin, A. and L. L. Chao (2001). "Semantic memory and the brain: structure and
processes." Curr Opin Neurobiol 11(2): 194-201.
Massimini, M., R. Huber, et al. (2004). "The sleep slow oscillation as a traveling
wave." J Neurosci 24(31): 6862-70.
Matynia, A., S. G. Anagnostaras, et al. (2001). "Weaving the molecular and
cognitive strands of memory." Neuron 32(4): 557-9.
McClelland, J. L., B. L. McNaughton, et al. (1995). "Why there are complementary
learning systems in the hippocampus and neocortex: insights from the successes and
failures of connectionist models of learning and memory." Psychol Rev 102(3): 419-57.
McDermott, E. C. a. D. (1985). Introduction to Artificial Intelligence, Reading,
Mass.: Addison-Wesley.
McDonald, R. J., N. S. Hong, et al. (2004). "The challenges of understanding
mammalian cognition and memory-based behaviours: an interactive learning and
memory systems approach." Neurosci Biobehav Rev 28(7): 719-45.
McGaugh, J. L. (2000). "Memory – a century of consolidation." Science 287(5451):
248-51.
McGaugh, J. L. (2004). "The amygdala modulates the consolidation of memories of
emotionally arousing experiences." Annu Rev Neurosci 27: 1-28.
McGinty, D., M. N. Alam, et al. (2001). "Hypothalamic sleep-promoting
mechanisms: coupling to thermoregulation." Arch Ital Biol 139(1-2): 63-75.
McNaughton, B. L., F. P. Battaglia, et al. (2006). "Path integration and the neural
basis of the 'cognitive map'." Nat Rev Neurosci 7(8): 663-78.
Mecklinger, A. (2009). "The control of long-term memory: Brain systems and
cognitive processes." Neurosci Biobehav Rev.
Mednick, S., K. Nakayama, et al. (2003). "Sleep-dependent learning: a nap is as
good as a night." Nat Neurosci 6(7): 697-8.
Mednick, S. C., D. J. Cai, et al. (2008). "Comparing the benefits of caffeine, naps
and placebo on verbal, motor and perceptual memory." Behav Brain Res 193(1): 79-86.
142
Mehta, M. R., A. K. Lee, et al. (2002). "Role of experience and oscillations in
transforming a rate code into a temporal code." Nature 417(6890): 741-6.
Milner, C. E., S. M. Fogel, et al. (2006). "Habitual napping moderates motor
performance improvements following a short daytime nap." Biol Psychol 73(2): 141-56.
Mishkin, M. (1978). "Memory in monkeys severely impaired by combined but not
by separate removal of amygdala and hippocampus." Nature 273(5660): 297-8.
Mitchell, C. J. (2009). "Human and animal perceptual learning: some common and
some unique features." Learn Behav 37(2): 154-60.
Mizumori, S. J., O. Yeshenko, et al. (2004). "Parallel processing across neural
systems: implications for a multiple memory system hypothesis." Neurobiol Learn Mem
82(3): 278-98.
Monchi, O., M. Petrides, et al. (2006). "Functional role of the basal ganglia in the
planning and execution of actions." Ann Neurol 59(2): 257-64.
Mukai, I., D. Kim, et al. (2007). "Activations in visual and attention-related areas
predict and correlate with the degree of perceptual learning." J Neurosci 27(42): 11401-11.
Nadel, L. and M. Moscovitch (1997). "Memory consolidation, retrograde amnesia
and the hippocampal complex." Curr Opin Neurobiol 7(2): 217-27.
Nader, K., G. E. Schafe, et al. (2000). "The labile nature of consolidation theory."
Nat Rev Neurosci 1(3): 216-9.
Oakley, D. A. (1981). "Brain mechanisms of mammalian memory." Br Med Bull
37(2): 175-80.
O'Keefe, J. and J. Dostrovsky (1971). "The hippocampus as a spatial map.
Preliminary evidence from unit activity in the freely-moving rat." Brain Res 34(1): 171-5.
Opitz, B. (2009). "Neural binding mechanisms in learning and memory." Neurosci
Biobehav Rev.
Parkosadze, K., T. U. Otto, et al. (2008). "Perceptual learning of bisection stimuli
under roving: slow and largely specific." J Vis 8(1): 5 1-8.
Pascual-Marqui, R. D., C. M. Michel, et al. (1995). "Segmentation of brain electrical
activity into microstates: model estimation and validation." IEEE Trans Biomed Eng
42(7): 658-65.
Pare, D., D. R. Collins, et al. (2002). "Amygdala oscillations and the consolidation
of emotional memories." Trends in Cognitive Sciences 6(7): 306-314.
Pastalkova, E., P. Serrano, et al. (2006). "Storage of spatial information by the
maintenance mechanism of LTP." Science 313(5790): 1141-4.
Pavlides, C. and J. Winson (1989). "Influences of hippocampal place cell firing in
the awake state on the activity of these cells during subsequent sleep episodes." J
Neurosci 9(8): 2907-18.
Pavlov, I. P. (1927). Conditioned Reflexes: An Investigation of the Physiological
Activity of the Cerebral Cortex. London.
Pennartz, C. M., J. D. Berke, et al. (2009). "Corticostriatal Interactions during
Learning, Memory Processing, and Decision Making." J Neurosci 29(41): 12831-8.
Pennartz, C. M., H. B. Uylings, et al. (2002). "Memory reactivation and
consolidation during sleep: from cellular mechanisms to human performance." Prog Brain
Res 138: 143-66.
Peigneux, P., S. Laureys, et al. (2001). "Sleeping brain, learning brain. The role of
sleep for memory systems." Neuroreport 12(18): A111-24.
Peigneux, P., S. Laureys, et al. (2004). "Are spatial memories strengthened in the
human hippocampus during slow wave sleep?" Neuron 44(3): 535-45.
Peigneux, P., P. Orban, et al. (2006). "Offline persistence of memory-related
cerebral activity during active wakefulness." PLoS Biol 4(4): e100.
143
Pennartz, C. M., J. D. Berke, et al. (2009). "Corticostriatal Interactions during
Learning, Memory Processing, and Decision Making." J Neurosci 29(41): 12831-8.
Pennartz, C. M., H. B. Uylings, et al. (2002). "Memory reactivation and
consolidation during sleep: from cellular mechanisms to human performance." Prog Brain
Res 138: 143-66.
Petrov, A. A., B. A. Dosher, et al. (2006). "Perceptual learning without feedback in
non-stationary contexts: data and model." Vision Res 46(19): 3177-97.
Peyrache, A., M. Khamassi, et al. (2009). "Replay of rule-learning related neural
patterns in the prefrontal cortex during sleep." Nat Neurosci 12(7): 919-26.
Plihal, W. and J. Born (1999). "Effects of early and late nocturnal sleep on priming
and spatial memory." Psychophysiology 36(5): 571-82.
Poe, G. R., D. A. Nitz, et al. (2000). "Experience-dependent phase-reversal of
hippocampal neuron firing during REM sleep." Brain Res 855(1): 176-80.
Poldrack, R. A., F. W. Sabb, et al. (2005). "The neural correlates of motor skill
automaticity." J Neurosci 25(22): 5356-64.
Portas, C. M., B. A. Strange, et al. (2000). "How does the brain sustain a visual
percept?" Proc Biol Sci 267(1446): 845-50.
Pourtois, G., K. S. Rauss, et al. (2008). "Effects of perceptual learning on primary
visual cortex activity in humans." Vision Res 48(1): 55-62.
Press, D. Z., M. D. Casement, et al. (2005). "The time course of off-line motor
sequence learning." Brain Res Cogn Brain Res 25(1): 375-8.
Qin, Y. L., B. L. McNaughton, et al. (1997). "Memory reprocessing in corticocortical
and hippocampocortical neuronal ensembles." Philos Trans R Soc Lond B Biol Sci
352(1360): 1525-33.
Quirk, M. C., K. I. Blum, et al. (2001). "Experience-dependent changes in
extracellular spike amplitude may reflect regulation of dendritic action potential backpropagation in rat hippocampal pyramidal cells." J Neurosci 21(1): 240-8.
Ranganath, C. and M. D'Esposito (2001). "Medial temporal lobe activity associated
with active maintenance of novel information." Neuron 31(5): 865-73.
Rasch, B. and J. Born (2007). "Maintaining memories by reactivation." Curr Opin
Neurobiol 17(6): 698-703.
Rasch, B., C. Buchel, et al. (2007). "Odor cues during slow-wave sleep prompt
declarative memory consolidation." Science 315(5817): 1426-9.
Reber, P. J. and L. R. Squire (1994). "Parallel brain systems for learning with and
without awareness." Learn Mem 1(4): 217-29.
Recanzone, G. H., M. M. Merzenich, et al. (1992). "Topographic reorganization of
the hand representation in cortical area 3b owl monkeys trained in a frequencydiscrimination task." J Neurophysiol 67(5): 1031-56.
Recanzone, G. H., C. E. Schreiner, et al. (1993). "Plasticity in the frequency
representation of primary auditory cortex following discrimination training in adult owl
monkeys." J Neurosci 13(1): 87-103.
Ribeiro, S., C. V. Mello, et al. (2002). "Induction of hippocampal long-term
potentiation during waking leads to increased extrahippocampal zif-268 expression
during ensuing rapid-eye-movement sleep." J Neurosci 22(24): 10914-23.
Ribeiro, S. and M. A. Nicolelis (2004). "Reverberation, storage, and postsynaptic
propagation of memories during sleep." Learn Mem 11(6): 686-96.
Robertson, E. M. (2007). "The serial reaction time task: implicit motor skill
learning?" J Neurosci 27(38): 10073-5.
Roediger, H. L., Dudai, Y., & Fitzpatrick, S. M. (2007). Science of memory:
Concepts. Oxford, UK, Oxford University Press
144
Rolls, E. T. (2000). "Memory systems in the brain." Annu Rev Psychol 51: 599-630.
Rolls, E. T. (2000). "Hippocampo-cortical and cortico-cortical backprojections."
Rudoy, J. D., J. L. Voss, et al. (2009). "Strengthening individual memories by
reactivating them during sleep." Science 326(5956): 1079.
Russeler, J. and F. Rosler (2000). "Implicit and explicit learning of event
sequences: evidence for distinct coding of perceptual and motor representations." Acta
Psychol (Amst) 104(1): 45-67.
Sagi, D. and D. Tanne (1994). "Perceptual learning: learning to see." Curr Opin
Neurobiol 4(2): 195-9.
Sasaki, Y., J. E. Nanez, et al. "Advances in visual perceptual learning and
plasticity." Nat Rev Neurosci 11(1): 53-60.
Schacter, D. L. and D. R. Addis (2007). "Constructive memory: the ghosts of past
and future." Nature 445(7123): 27.
Schacter, D. L., D. R. Addis, et al. (2008). "Episodic simulation of future events:
concepts, data, and applications." Ann N Y Acad Sci 1124: 39-60.
Schacter, D. L. a. T., E. (1994). Memory systems 1994, Massachusetts Institute of
Technology.
Scharf, M. T., N. Naidoo, et al. (2008). "The energy hypothesis of sleep revisited."
Prog Neurobiol 86(3): 264-80.
Schendan, H. E., M. M. Searl, et al. (2003). "An FMRI study of the role of the
medial temporal lobe in implicit and explicit sequence learning." Neuron 37(6): 1013-25.
Schendan, H. E., M. M. Searl, et al. (2003). "Sequence? What Sequence?: the
human medial temporal lobe and sequence learning." Mol Psychiatry 8(11): 896-7.
Schmidt, C., P. Peigneux, et al. (2006). "Encoding difficulty promotes postlearning
changes in sleep spindle activity during napping." J Neurosci 26(35): 8976-82.
Schmidt, C., F. Collette, et al. (2007). "A time to think: circadian rhythms in
human cognition." Cogn Neuropsychol 24(7): 755-89.
Schoups, A., R. Vogels, et al. (2001). "Practising orientation identification improves
orientation coding in V1 neurons." Nature 412(6846): 549-53.
Schultz, W., P. Dayan, et al. (1997). "A neural substrate of prediction and reward."
Science 275(5306): 1593-9.
Schwartz, S., P. Maquet, et al. (2002). "Neural correlates of perceptual learning: a
functional MRI study of visual texture discrimination." Proc Natl Acad Sci U S A 99(26):
17137-42.
Schwartz, S., P. Vuilleumier, et al. (2005). "Attentional load and sensory
competition in human vision: modulation of fMRI responses by load at fixation during
task-irrelevant stimulation in the peripheral visual field." Cereb Cortex 15(6): 770-86.
Scoville, W. B. and B. Milner (1957). "Loss of recent memory after bilateral
hippocampal lesions." J Neurol Neurosurg Psychiatry 20(1): 11-21.
Scoville, W. B. and B. Milner (2000). "Loss of recent memory after bilateral
hippocampal lesions. 1957." J Neuropsychiatry Clin Neurosci 12(1): 103-13.
Seeck, M., N. Mainwaring, et al. (1997). "Neurophysiologic correlates of implicit
face memory in intracranial visual evoked potentials." Neurology 49(5): 1312-6.
Seitz, A. and T. Watanabe (2005). "A unified model for perceptual learning."
Trends Cogn Sci 9(7): 329-34.
Seitz, A. R. and T. Watanabe (2003). "Psychophysics: Is subliminal learning really
passive?" Nature 422(6927): 36.
145
Shaban, H., Y. Humeau, et al. (2006). "Generalization of amygdala LTP and
conditioned fear in the absence of presynaptic inhibition." Nat Neurosci 9(8): 1028-35.
Shadmehr, R. and H. H. Holcomb (1997). "Neural correlates of motor memory
consolidation." Science 277(5327): 821-5.
Shank, S. S. and D. Margoliash (2009). "Sleep and sensorimotor integration during
early vocal learning in a songbird." Nature 458(7234): 73-7.
Shima, K., M. Isoda, et al. (2007). "Categorization of behavioural sequences in the
prefrontal cortex." Nature 445(7125): 315-8.
Shors, T. J., G. Miesegaes, et al. (2001). "Neurogenesis in the adult is involved in
the formation of trace memories." Nature 410(6826): 372-6.
Siapas, A. G. and M. A. Wilson (1998). "Coordinated interactions between
hippocampal ripples and cortical spindles during slow-wave sleep." Neuron 21(5): 1123-8.
Sigala, N. (2004). "Visual categorization and the inferior temporal cortex." Behav
Brain Res 149(1): 1-7.
Sirota, A., J. Csicsvari, et al. (2003). "Communication between neocortex and
hippocampus during sleep in rodents." Proc Natl Acad Sci U S A 100(4): 2065-9.
Skaggs, W. E. and B. L. McNaughton (1996). "Replay of neuronal firing sequences
in rat hippocampus during sleep following spatial experience." Science 271(5257):
1870-3.
Skrandies, W. and M. Fahle (1994). "Neurophysiological correlates of perceptual
learning in the human brain." Brain Topogr 7(2): 163-8.
Sperling, R., E. Chua, et al. (2003). "Putting names to faces: successful encoding
of associative memories activates the anterior hippocampal formation." Neuroimage
20(2): 1400-10.
Spiegel, K., R. Leproult, et al. (1999). "Impact of sleep debt on metabolic and
endocrine function." Lancet 354(9188): 1435-9.
Squire, L. R. (1987). "The organization and neural substrates of human memory."
Int J Neurol 21-22: 218-22.
Squire, L. R. (1992). "Memory and the hippocampus: a synthesis from findings
with rats, monkeys, and humans." Psychol Rev 99(2): 195-231.
Squire, L. R. (2004). "Memory systems of the brain: a brief history and current
perspective." Neurobiol Learn Mem 82(3): 171-7.
Squire, L. R. (2009). "The legacy of patient H.M. for neuroscience." Neuron 61(1):
6-9.
Squire, L. R. (2009). "Memory and brain systems: 1969-2009." J Neurosci 29(41):
12711-6.
Squire, L. R. and P. Alvarez (1995). "Retrograde amnesia and memory
consolidation: a neurobiological perspective." Curr Opin Neurobiol 5(2): 169-77.
Squire, L. R., C. E. Stark, et al. (2004). "The medial temporal lobe." Annu Rev
Neurosci 27: 279-306.
Squire, L. R. and S. Zola-Morgan (1991). "The medial temporal lobe memory
system." Science 253(5026): 1380-6.
Squire, L. R. and S. M. Zola (1996). "Structure and function of declarative and
nondeclarative memory systems." Proc Natl Acad Sci U S A 93(24): 13515-22.
Staresina, B. P. and L. Davachi (2009). "Mind the gap: binding experiences across
space and time in the human hippocampus." Neuron 63(2): 267-76.
Steriade, M. (2006). "Grouping of brain rhythms in corticothalamic systems."
Neuroscience 137(4): 1087-106.
146
Sterpenich, V., G. Albouy, et al. (2009). "Sleep promotes the neural reorganization
of remote emotional memory." J Neurosci 29(16): 5143-52.
Stickgold, R. (2005). "Sleep-dependent memory consolidation." Nature 437(7063):
1272-8.
Stickgold, R., J. A. Hobson, et al. (2001). "Sleep, learning, and dreams: off-line
memory reprocessing." Science 294(5544): 1052-7.
Stickgold, R., L. James, et al. (2000). "Visual discrimination learning requires sleep
after training." Nat Neurosci 3(12): 1237-8.
Stickgold, R. and M. P. Walker (2005). "Memory consolidation and reconsolidation:
what is the role of sleep?" Trends Neurosci 28(8): 408-15.
Stickgold, R., D. Whidbee, et al. (2000). "Visual discrimination task improvement:
A multi-step process occurring during sleep." J Cogn Neurosci 12(2): 246-54.
Sutherland, G. R. and B. McNaughton (2000). "Memory trace reactivation in
hippocampal and neocortical neuronal ensembles." Curr Opin Neurobiol 10(2): 180-6.
Suzuki, W. A. (2007). "Integrating associative learning signals across the brain."
Suzuki, W. A. (2008). "Associative learning signals in the brain." Prog Brain Res
169: 305-20.
Takahashi, S. and Y. Sakurai (2009). "Information in small neuronal ensemble
activity in the hippocampal CA1 during delayed non-matching to sample performance in
rats." BMC Neurosci 10: 115.
Takashima, A., I. L. Nieuwenhuis, et al. (2009). "Shift from hippocampal to
neocortical centered retrieval network with consolidation." J Neurosci 29(32): 10087-93.
Tononi, G. and C. Cirelli (2003). "Sleep and synaptic homeostasis: a hypothesis."
Brain Res Bull 62(2): 143-50.
Tsodyks, M. and C. Gilbert (2004). "Neural networks and perceptual learning."
Nature 431(7010): 775-81.
Tulving, E. (1985). "How many memory systems are there?" American Psychologist
40(4).
Tulving, E. (2001). "Episodic memory and common sense: how far apart?" Philos
Trans R Soc Lond B Biol Sci 356(1413): 1505-15.
Tulving, E. (2002). "Episodic memory: from mind to brain." Annu Rev Psychol 53:
1-25.
Ungerleider, L. G., J. Doyon, et al. (2002). "Imaging brain plasticity during motor
skill learning." Neurobiol Learn Mem 78(3): 553-64.
Van Cauter, E., K. Spiegel, et al. (2008). "Metabolic consequences of sleep and
sleep loss." Sleep Med 9 Suppl 1: S23-8.
Vassalli, A. and D. J. Dijk (2009). "Sleep function: current questions and new
approaches." Eur J Neurosci 29(9): 1830-41.
Van de Ville, D., J. Britz, et al. "EEG microstate sequences in healthy humans at
rest reveal scale-free dynamics." Proc Natl Acad Sci U S A 107(42): 18179-84.
Voss, J. L. and K. A. Paller (2008). "Brain substrates of implicit and explicit
memory: the importance of concurrently acquired neural signals of both memory types."
Neuropsychologia 46(13): 3021-9.
Vuilleumier, P. and J. Driver (2007). "Modulation of visual processing by attention
and emotion: windows on causal interactions between human brain regions." Philos
Trans R Soc Lond B Biol Sci 362(1481): 837-55.
Wachs, K. (1994). "Intermanual transfer of implicit knowledge of sequential finger
movements." Neurology 44 (Suppl.2).
147
Wagner, U., S. Gais, et al. (2001). "Emotional memory formation is enhanced
across sleep intervals with high amounts of rapid eye movement sleep." Learn Mem
8(2): 112-9.
Walker, M. P. (2005). "A refined model of sleep and the time course of memory
formation." Behav Brain Sci 28(1): 51-64; discussion 64-104.
Walker, M. P. (2008). "Sleep-dependent memory processing." Harv Rev Psychiatry
16(5): 287-98.
Walker, M. P. (2009). "The role of sleep in cognition and emotion." Ann N Y Acad
Sci 1156: 168-97.
Walker, M. P. and R. Stickgold (2006). "Sleep, memory, and plasticity." Annu Rev
Psychol 57: 139-66.
Walker, M. P., R. Stickgold, et al. (2005). "Sleep-dependent motor memory
plasticity in the human brain." Neuroscience 133(4): 911-7.
Wallenstein, G. V., H. Eichenbaum, et al. (1998). "The hippocampus as an
associator of discontiguous events." Trends Neurosci 21(8): 317-23.
Wamsley, E. J., K. Perry, et al. "Cognitive replay of visuomotor learning at sleep
onset: temporal dynamics and relationship to task performance." Sleep 33(1): 59-68.
Wamsley, E. J. and R. Stickgold "Dreaming and offline memory processing." Curr
Biol 20(23): R1010-3.
Watanabe, T., J. E. Nanez, Sr., et al. (2002). "Greater plasticity in lower-level than
higher-level visual motion processing in a passive perceptual learning task." Nat
Neurosci 5(10): 1003-9.
Watanabe, T., J. E. Nanez, et al. (2001). "Perceptual learning without perception."
Nature 413(6858): 844-8.
Wieser, S. and H. G. Wieser (2003). "Event-related brain potentials in memory:
correlates of episodic, semantic and implicit memory." Clin Neurophysiol 114(6): 1144-52.
White, N. M. (2009). "Some highlights of research on the effects of caudate
nucleus lesions over the past 200 years." Behav Brain Res 199(1): 3-23.
Willingham, D. B. (1998). "A neuropsychological theory of motor skill learning."
Psychol Rev 105(3): 558-84.
Wilson, M. A. and B. L. McNaughton (1994). "Reactivation of hippocampal
ensemble memories during sleep." Science 265(5172): 676-9.
Wolpert, D. M., Z. Ghahramani, et al. (1995). "An internal model for sensorimotor
integration." Science 269(5232): 1880-2.
Yotsumoto, Y., T. Watanabe, et al. (2008). "Different dynamics of performance and
brain activation in the time course of perceptual learning." Neuron 57(6): 827-33.
Zola-Morgan, S., L. R. Squire, et al. (1982). "The neuroanatomy of amnesia:
amygdala-hippocampus versus temporal stem." Science 218(4579): 1337-9.
148

Thesis Reference - Archive ouverte UNIGE

Transcription

Similar documents

the university of notre dame graduate school banner book

duck-billed platypus echidna

House of Mouse Innards - Knives Forks and Spoons Press

example pages - Tierschutz macht Schule

From THe Desk Of Renee Shepherd Owens, CEo

information about their activities may be viewed

Marlton Office – New Patient Forms

SUMMERHOUSE OF SHOREVIEW

Healthy News IE - A.Vogel Herbal Remedies

Effective Treatment for Chronic Fatigue Syndrome and Fibromyalgia