SLEEP AND COGNITION

E D I T E D
BY
RICHARD R.
BOOTZIN,
J O H N F.
KIHLSTROM
AND
DANIEL 1.
SCHACTER

UNIVERSITY
OF
ARIZONA

Arne rican Psychologica I Association

Washington, DC
 CONTENTS

Contributors ix

Foreword X

Preface xii
PART ONE: GENERAL OVERVIEWS
1 The Neurocognition of Sleep Mentation: Rapid 1
Eye Movements, Visual Imagery, and Dreaming
John Antrobus

2 Activation, Input Source, and Modulation: A 25

Neurocognitive Model of the State of the
Brain-Mind
J. Allan Hobson
PART TWO: COGNITION DURING SLEEP
3 Event-Related Brain Potential (ERP) Studies of 43
Cognition During Sleep: Is It More Than a
Dream?
Marta Kutas

4 Stimulus Control and Sleep 58

John Harsh and Pietro Badia

5 Memories in Sleep: Old and New 67

Pietro Badia

vii
 6 Behavioral Responses During Sleep 77
Frederick J. Evans

7 Learning During Sleep 88

Eric Eich

8 Lucid Dreaming: Psychophysiological Studies of 109

Consciousness During REM Sleep
Stephen LaBerge

9 Cognitive Processing and General Anesthesia 127

Les Goldmann

PART THREE: COGNITION BEFORE AND AFTER SLEEP

10 Insomnia: The Patient and the Pill 139

WallaceB. Mendelson

1 1 The Perception of Sleep Onset in Insomniacs and 148

Normal Sleepers
Michael H. Bonnet

12 Are You Awake? Cognitive Performance and 159

Reverie During the Hypnopompic State
David F. Dinges

PART FOUR CLINICAL TOPICS

13 A Network Model of Dreams 179

Rosalind Cartwright

14 Nightmares (Dream Disturbances) in 190

Posttraumatic Stress Disorder: Implications for a
Theory of Dreaming
Milton Kramer

Index 203

viii
 CONTRIBUTORS

John Antrobus, The City College of the City Universityof New York
Pietro Badia, Bowling Green State University
Michael H . Bonnet, Long Beach VeteransAdministration Hospital and University of
California,Irvine
Rosalind Cartwright, Rush-Presbyterian-St. Luke's Hospital
David F. Dinges, University of Pennsylvania
Eric Eich, The Universityof British Columbia
Frederick J . Evans, Universityof Medicine and Dentistry of New Jersey
Les Goldmann, Reed College
John Harsh, Universityof Southern Mississippi
J . Allan Hobson, Harvard Medical School
Milton Kramer, Sleep Disorders Center of Greater Cincinnati
Marta Kutas, Universityof California,San Diego
Stephen LaBerge, Stanford University
Wallace B. Mendelson, State Universityof New York at Stony Brook

ix
 FOREWORD

Federal research agencies stopped regularly supporting investigator-initiated

“state-of-the-art” research conferencesin scientific psychology well over a de-
cade ago. Yet over that same period, scientific psychology has continued to
grow-as well as to diversify into many new areas. Thus there have been rela-
tively few opportunities for investigatorsin new and cutting-edge research ar-
eas to convene in special settingsto discuss their findings.
The American Psychological Association (APA), as part of its continu-
ing effortsto enhance the dissemination of scientific knowledge in psychology,
undertook a number of new initiatives designed to foster scientific research
and communication. In particular, the APA Science Directorate, in 1988, ini-
tiated the Scientific Conferences Program.
The APA ScientificConferences Program provides university-based psy-
chological researchers with seed monies essential to organizing specialty con-
ferences on critical issues in basic research, applied research, and methodologi-
cal issues in psychology. Deciding which conferences to support involves a
competitive process. An annual call for proposals is issued by the APA Science
Directorate to solicit conference ideas. Proposals from all areas of psychologi-
cal research are welcome. They are then reviewed by qualified psychologists,
who forward substantive suggestions and funding recommendations to the
Science Directorate. At each stage, the criteria used to determine which con-
ferences to support include relevance, timeliness, and comprehensivenessof
the topics, and qualifications of the presenters. In 1988, seven conferences
were funded under the APA Science Directorate program’s sponsorship, and
six conferences were funded in 1989. We expect to fund six more in 1990, at
an annual program expense of $90,000 to $100,000.
The APA ScientificConferences Program has two major goals. The first
is to provide, by means of the conferences, a broad view of specifictopics (and,
when appropriate, to provide for interdisciplinary participation). The second
goal is to assure timely dissemination of the findings presented by publishing a
seriesof carefully crafted scholarly volumes based, in part, on each conference.
Thus the information reaches the audiences at each conference as well as the
broader psychological and scientific communities. This enables psychology
and related fields to benefit from the most current research on a given topic.
X
 This volume presents findings reported at the January 1989 conference,
“Sleep and Cognition.” Researchers in this area are focusing on a number of
fascinating topics. They are applying the methods of contemporary cognitive
psychology to traditional areas of interest such as the purpose of dreams and
our potential to learn during sleep, as well as exploring newer dimensions like
lucid dreaming, cognitive processes under anesthesia, and effects of sleepin-
ducing medication. Researchers from the typically separate areas of sleep and
cognition came together at the conferenceto exchange ideas and focus on new
directions for research.
This volume is representative of what we at the American Psychological
Association believe will be a number of exceptional volumes that give readers
a broad sampling of the diverse and outstanding research now being done in
scientific psychology. We hope you will enjoy and be stimulated by this book
and the many others to come.
A list of the conferencesfunded through this program follows:

Researching Community Psychology: Integrating Theories and Methodolo-

gies, September 1988
Psychological Well-Being in Nonhuman Captive Primates, September 1988
Psychological Research on Organ Donation, October 1988
Arizona Conference on Sleep and Cognition, January 1989
Socially Shared Cognition, February 1989
The Role of Experience in ModifyingTaste and Its Effects on Feeding,
April 1989
Perception of Structure, May 1989
Suggestibility of Children’s Recollections, June 1989
Best Methods for the Analysis of Change, October 1989
Conceptualization and Measurement of Organism-Environment Interaction,
November 1989
Cognitive Bases of Musical Communication, April 1990
Conference on Hostility, Coping/Support, and Health, November 1990
Psychological Testing of Hispanics, February 1991
Study of Cognition: Conceptual and Methodological Issues, February 1991

Gary R. VandenBos, PhD Virginia E. Holt

Acting Executive Director Manager, ScientificConferences Program
Science Directorate, APA Science Directorate, APA

xi
 PREFACE

Sleep was once a phenomenon of considerable interest to psychologists. One

of Freud’s most popular books (and arguably one of his best) dealt with
dreams. In the classic 1924 study by Jenkins and Dallenbach, sleep provided
convincing evidence in favor of the interference theory of forgetting. More
than 50 years later, and almost a century after Freud, some of the most fre-
quently asked questions in the introductory psychology course concern what
dreams mean and why we seem to forget them. Despite impressive advances
in understanding sleep as a biological phenomenon and in understanding the
nature of mental processes in the normal waking state, our knowledge of men-
tal life during sleep remains extremely impoverished.
Much of the current scientific interest in sleep can be traced to the dis-
covery in 1953, by Aserinsky and Kleitmen, of the association between rapid
eye movements and dreaming. Ironically, however, the discovery of physio-
logically different sleep stages, and their differential association with dreaming
and other mental activities, seems to have transformed sleep from a psycholog-
ical topic to a biological one. A great deal of work addresses sleep as behavior
at a purely biological level, as in studies of the ontogeny and phylogeny of
sleep and other biological rhythms. And many major medical centers offer the
services of a sleep disorders clinic oriented toward such problems as insomnia
(a common symptom in depression and normal aging), narcolepsy, sleep ap-
nea, and parasomnias such as enuresis and somnambulism. But surprisingly
little current research focuses on perceptual, memory, and thought processes
during this period of activity. There is, of course, a long tradition of work by
psychologists on the effects of sleep deprivation on human performance. But
research on cognitive processes during sleep itself has received comparatively
little attention, both on the conference circuit and in print. Examination of
Sleep Research, the annual archival publication of the Association of Profes-
sional Sleep Societies, reveals relatively little cognitive psychology. Similar
trends are apparent in the published proceedings of the biennial European
Conferences on Sleep Research and in the volumes of Advances in Sleep Re-
search. There is, of course, a large literature (arising primarily from within the
psychoanalytic tradition) concerning the interpretation of dreams. Unfortu-

xii
nately, the methodology in these studies is generally clinical apd hermeneutic,
without the careful controls characteristic of experimental psychology.
Nevertheless, important questions about sleep persist that must be ad-
dressed at a purely psychological level: How much information from the envi-
ronment can be processed by the sleeper? What sorts of mental activities are
associated with sleepwalkingand sleeptalking?Are dreams meaningful, and if
so why aren’t they better remembered? What are the possibilities for learning
during sleep? Such questions have been asked by psychologists and others,
but only rarely have they been addressed with the concepts, principles, and
methods of contemporary cognitive psychology. One factor contributing to
this situation is the isolation of psychoanalysis, with its interest in dreams,
from mainstream scientific psychology. Another is the phenomenology of
sleep itself: The sleeper typically appears oblivious to his or her environment,
and the occurrence of sleep is generally inferred from this lack of responsive-
ness and the absence of conscious awareness of events during sleep. The com-
mon identification of cognition with consciousness leaves the impression that
little or no mental activity occurs during sleep.
Of course, studies in which sleepers are awakened at various points in
the night’s sleep typically yield reports of thoughts, images, and dreams, but
these are rarely remembered spontaneously in the morning. Some individuals
engage in episodes of sleepwalking and sleeptalking, which can involve rela-
tively complex speech acts and behavioral sequences-but, again, these are
rarely if ever remembered. One explanation for this universally experienced
memory deficit is that the cortical centers that mediate complex mental pro-
cessing are disengaged or deactivated during sleep, with the result that the
sleeper does not attend to environmental and mental events occurring during
periods of sleep. Because these events are not noticed, they are not processed
in a manner that encodes accessible traces of them in memory.
Thus, the most commonly accepted explanation of postsleep amnesia is
in terms of consolidation failure. For example, it has been proposed that the
low level of cortical arousal characteristic of sleep effectively prevents the
sleeper from performingthe cognitive operationsnecessary to encode memory
traces of sleep events that are accessible in the subsequent waking state. Ac-
cording to this view, dreams and other sleep events are remembered only when
the sleeper awakens during them, permitting retrieval from short-term mem-
ory. If the sleeper awakens shortly after a dream has occurred, residual infor-
mation retrieved from short-term memory may serve as a cue to the retrieval
of a highly degraded long-term memory trace. If retrieval is delayed until all
traces of the dream have decayed or been displaced from short-term memory,
the long-term traces will be virtually inaccessible.
As intuitively appealing as this explanation is, recent empirical work on
sleep, as well as theoretical advances in the study of cognition, call it into
account. For example, the idea of consolidation failure assumes a rigid distinc-
tion between short-term and long-term memory that is not supported by the

xiii
current literature. Moreover, cognitive theorists have recently begun to distin-
guish between effortful processes, which require intention and consume atten-
tional resources, and automatic processes, which do not. More important, the
domain of automatic processing has expanded to include rather complex
mental activities-the kind that ordinarily would be expected to leave residual
traces in permanent memory. Finally, studies of both brain-damaged patients
and intact subjects support a concept of implicit perception and memory, in
which current and past events may influence ongoing experience, thought,
and action even though the individual lacks awareness (concurrent or retro-
spective) of the events themselves.
Recent research on sleep processes bears directly on these theoretical
ideas. While sleepers are ordinarily considered unable to engage in complex,
intentional cognitive activities during sleep, some evidence tends to contradict
this assumption. For example, in the phenomenon of lucid dreaming, selected
subjects report that they become aware of the fact that they are dreaming and
are able to consciously direct the contents of the dream, while remaining
asleep. This claim is difficult to verify objectively, for obvious reasons. But a
series of highly provocative and apparently well-controlled studies has shown
that subjects can make discriminative responses to verbal suggestions deliv-
ered while they are asleep. Although the possibility of sleep learning had been
firmly rejected in a widely influential review published in 1955, studies of
memory of events occurring during surgical anesthesia strongly suggest that
sleep learning may be possible, provided that its success is measured in terms
of implicit rather than explicit memory. These preliminary findings, which
have emerged from a variety of laboratories employing rather different para-
digms, suggest that cognitive activity of considerable complexity may be possi-
ble during sleep, provided that it is mediated by automated procedures and
assessed by measures that do not require awareness of the events on the part
of the subject.
Despite the research possibilities offered by current advances in theory,
most texts on sleep, whether intended for undergraduate use or scholarly refer-
ence, pay little attention to cognitive processes, aside from our ubiquitous fail-
ure to remember our dreams. A major exception to this generalization is The
Mind in Sleep (Arkin, Antrobus, & Ellman, 1978), which provided a compre-
hensive account of research on cognitive processes during sleep since the dis-
covery of the EEG correlates of dreaming. The volume focuses on such topics
as the difference between stage REM and stage NREM mental activity, factors
affecting dream recall, the effects of presleep and intrasleep stimulation, REM
deprivation, sleeptalking, and night terrors. Unfortunately, the book is now
out of date, especially with respect to current thinking in cognitive psychology;
although a new edition has been promised, it is not yet available. Another
primary source is a recent monograph, Dreaming: A Cognitive-Psychological
Analysis (Foulkes, 1985). As comprehersive as the Arkin et al. anthology is,
its contributors have almost nothing to say about the problem that concerned
Freud most: the meaning of dreams. Foulkes’ book approaches this problem
xiv
from the perspective of cognitive psychology, particularly psycholinguistics
and cognitive development. His analysis clearly indicates that sleep research-
ers and cognitive psychologists have much of interest to say to each other.
In 1986, the Neurosciences Institute of Rockefeller University hosted a
small workshop on the remembering and forgetting of dreams, which testifies
to an increased interest in sleep among neurobiologists and psychiatrists; but
cognitive psychologists were decidedly underrepresented at this meeting. A
further impetus to dialog is the recently published report of an ad-hoc commit-
tee of the National Academy of Sciencesto the Army Research Institute, con-
cerning psychological techniques to enhance human performance (Enhancing
Human Performance:Issues, Theories, and Techniques, 1988). While gener-
ally critical of most of the techniques that have been proposed for this purpose,
the Committee strongly urged that further attention be devoted to the phe-
nomenon of learning during sleep. We expect that this suggestion will lead
to a greatly improved environment for research on all aspects of cognition
in sleep.
The present volume represents the Proceedings of the Arizona Confer-
ence on Sleep and Cognition, held in Tucson January 19-22, 1989. A princi-
pal concern of the conference was the implications of recent work on implicit
memory and other aspects of information-processingoutside of awareness for
studies of cognitive processes during sleep, and the role of the sleep laboratory
as a vehicle for studying various aspects of information processing outside of
awareness, in the absence of the active deployment of attention. To this end,
selected investigators in the area of sleep who have an interest in cognitive
processes were brought together with their counterparts in the area of cogni-
tion who have an interest in sleep. The goal of the conferencewas to create an

>
environment in which representativesof these two quite different areas would
meet and exchange ideas in a spirit of open inquiry and constructive criticism
and advance the study of cognition during sleep beyond the stage of specula-
tive thought. We hope that one outcome of this meeting, and publication of
the Proceedings, will be a new agenda for research on sleep that will carry
investigat rs of both camps well into the 2 1st century.
Part ne of the volume contains two papers in which mental activity
during sle p is used as a link between neuroscience and cognitive science.
Antrobus provides a sweeping review of the literature on eye movements and
image in dreaming, with particular reference to cortical activation during
sle . imilarly, Hobson provides an update of the Hobson-McCarley Activa-
tion-Synthesis model of dreaming.
Part Two includes a number of papers on various aspects of mental ac-
tivity during sleep. Kutas offers an authoritative review of the literature on
event-related (evoked) potentials during sleep. Harsh and Badia describe a
series of experiments on the transfer to sleep of conditioned responses ac-
quired during the normal waking state. Badia continues this discussion, focus-
ing on the acquisition of new conditioned responses during sleep. Evans
contributes an overview and update of his extremely provocative studies of
xv
response to verbal suggestions during sleep, and the relation of this phenome-
non to hypnosis and other dissociative states. Eich provides a new look, origi-
nally prepared for the 1988 National Academy of Sciences and Army Re-
search Institute study of techniques for enhancing human performance, at the
possibilities for sleep learning. LaBerge summarizes his programmatic re-
search on lucid dreaming, in which the sleeper becomes aware of the fact that
he or she is dreaming, and exercises some control over the content of the
dream. Finally, Goldmann explores the parallels between information pro-
cessing during sleep and during adequate surgical anesthesia.
Part Three focusses on cognition in the hypnogogic and hypnopompic
states occumng in the transition from waking to sleeping and back again.
Mendelson reviews the physiological and cognitive effects of drugs used in
treating insomnia. Bonnet examines the factors affectingsubjective reports of
sleep onset-which, he notes, almost never agree with physiological measures.
Dinges summarizes his research on performance impairments in the period
after awakening and illustrates his argument with a dramatic example of hyp-
nopompic reverie.
Part 4 returns to the question of dreams, this time in a clinical as opposed
to an experimental context, with an emphasis on the effects of stress on
dreams. Cartwright describes a new study of the dreams of women going
through divorce and explores the implications of her findings in relation to
the question of the meaning of dreams. Kramer summarizes his ongoing work
on nightmares in Vietnam veterans suffering posttraumatic stress disorder.
The Arizona Conference on Sleep and Cognition was supported in part
by the Science Directorate of the American Psychological Association. We are
grateful to Alan G. Kraut, Virginia E. Holt, and Barbara Calkins of the Science
Directorate for their efforts. Matching funds were provided by various units
of the University of Arizona (Department of Psychology, Lee Sechrest, Head;
Committee on Cognitive Science, Menill Garret, Director; College of Nurs-
ing, L. Claire Parsons, Dean; and the Sleep Disorders Center of the Arizona
Health Sciences Center, Stuart Quan, Director) and by the UpJohn Company
(James W. Battles, Medical Sciences Liaison). Final preparation of the Pro-
ceedings was supported in part by Grant #MH-35856 from the National Insti-
tute of Mental Health.
We are grateful to a number of individuals for their efforts in behalf of
the conference. Our colleagues Larry E. Beutler, L. Claire Parsons, and Stuart
F. Quan served as chairs for the conference sessions. Also at the University of
Arizona, Carol Cantor, Lucy Canter Kihlstrom, and Martha L. Glisky pro-,
vided organizational assistance-as did Karen Shoun of the Doubletree Hotel.
Elizabeth C. Tang of Communications: Visual designed the stunning confer-
ence poster. Barbara Calkins represented the APA Science Directorate at the
Conference; Lynne Lamberg, a science writer with a special interest in sleep,
grilled us during breaks and after-hours and forced us to express our ideas in
comprehensible English. Professors Robert A. Bjork (UCLA), David Koulack

xvi
(Manitoba), and Jonathan Winson (Rockefeller)made vigorous and incisive
contributions to the conference sessions. Finally, we thank those investigators
who enriched the conference by contributing summaries of their research to
the poster sessions: Andrew Brylowski (University of Texas Southwestern
Medical Center), Mindy Engle-Friedman (Baruch College, CUNY), Irene P.
Hoyt (University of Wisconsin), Robert Nadon (University of Pennsylvania),
Kan Paller (Yale School of Medicine), and Jim Wood (University of Arizona).

Richard R. Bootzin
John F. Kihlstrom
Daniel L. Schacter

xvii
3NO I t l V d
 CHAPTER 1

THE NEUROCOGNITION
OF SLEEP MENTATION:
RAPID EYE MOVEMENTS, VISUAL IMAGERY,
AND DREAMING

JOHN ANTROBUS

The relation of eye movement to the cognitive process of dreaming during

sleep has been a source of speculation since Ladd ( 1892) suggested that such
movement might represent the sleeper’s observation of his or her dreams. Af-
ter the discovery by Aserinsky and Kleitman ( 1953)of the association between
Stage 1 REM (rapid eye movement) sleep and dreaming, REMs, which occupy
only a fraction of Stage 1 REM sleep, have been identified as one component
of a cluster of phasic activities that are associated with a pontine-geniculate-
occipital (PGO) sequence of neural activity. Despite the dramatic frequency
and amplitude of PGO activity, the fact that it is not associated with any
known function in the waking state’hasconfounded efforts to identify its func-
tion in REM sleep.
Pontine-geniculate-occipitalactivity has been implicated in the produc-
tion of at least four characteristics of sleep thought and imagery: bizarreness,
direction of visual orientation, clarity or vividness of visual imagery, and
amount of information processed from nonsensory sources. The apparent
pontine origin of the PGO sequence implies that the cerebral cortex and, there-
fore, higher cognitive processes have no role in initiating eye movement (EM)
activity in REM sleep (Hobson & McCarley, 1977).This chapter evaluates the
neurocognitiveand neurophysiological support for models that have proposed
a relation between EMS and the cognitive characteristics of the imagery and
thought of REM sleep.
The first 25 years of contemporary sleep mentation research have fo-
cused on whether dreaming is isomorphic with Stage 1 REM or whether it is
3
4 JOHN ANTROBUS

better aligned with a correlated biological characteristic such as the REMs

themselves or with concurrent phasic events like the PGO spikes that are
readily observed in cats but not in intact human subjects.
Although many studies have reported modest relations between physio-
logical markers and characteristics of Stage 1 REM sleep mentation, the re-
lations have either not been replicated or have proven to be relatively small in
magnitude. In short, relative to non-REM (NREM) sleep, REM sleep consis-
tently yields dreamlike mentation reports and descriptions of fanciful, vivid,
hallucinatory, and sometimes bizarre imagery and thought, regardless of the
presence of physiological markers at different intervals within the REM pe-
riod. Nevertheless, the behavioral drama of the burst of REMs, particularly
in the early part of REM periods, continues to dominate neurophysiological
theories of dream production and perpetually continues to puzzle students in
this field.
The relatively small payoff of recent psychophysiological studies of
dreaming, coupled with the high cost of such research, has discouraged many
investigators from continuing the search. However, recent advances in the
neurosciences, particularly in neurocognition, have clearly demonstrated that
the b r a h m i n d is much more complex than any other subject in the history of
science and that most of our working assumptions about the relations between
brain concepts and mind concepts have been too simplistic. Thus, one of the
goals of this paper is to restate some of the early models of REM sleep and
dreaming in an updated context and, thereby, to reorient the direction of fu-
ture research.
It is now established that REM bursts or clusters within Stage 1 REM
sleep are associated with PGO-like activity (McCarley, Winkelman, & Duffy,
1983). It is therefore appropriate to consider all these phasic event bursts or
clusters-REMs, phasic submental EMG suppression, periorbital integrated
potentials (PIPS),and middle ear muscle activity (MEMAS)-as a family of
neural events closely associated in time or neural space.
Because of the greater ease of measurement of REMs and their putative
association with visual imagery, most phasic mentation studies have been car-
ried out with REMs. The early phasic-tonic studies pioneered by Foulkes and
colleagues (Foulkes & Pope, 1973)tended to regard phasic events as an inten-
sive form of Stage 1 REM sleep. The search for a link to sleep mentation was
broadly directed toward dream recall and the vividness of the events experi-
enced. (For an excellent review of this literature, see Pivik, 1978).)
Four positions describing the functional relation between phasic activity
and the characteristics of dreaming, or sleep mentation, have been proposed.
The first, suggested by Ladd ( 1892), is the compelling hypothesis that REMs
represent the action of the sleeper looking at the events in his or her dream.
Associated with this position, but not dependent on it, is the magnitude hy-
pothesis, namely, that periods of “dense” REM activity are associated with an
increment in the vividness or clarity of the dreamer’s visual imagery.
The Activation-Synthesis model proposed by Hobson and McCarley
( 1977), which was founded on an articulated neurophysiological model of
Neurocognition 5

sleep stage alternation, posits that EM information is initiated by the pons

rather than by the cerebral cortex. Receiving EM information from sources
not under its control obliges the cognitive apparatus that produces dream im-
agery to, in effect, “make the best of a bad job” (p. 1 l), thereby creating menta-
tion that is often bizarre in nature.
The fourth position is a modification of the first and second positions
and is based on the across states connectionist model of imagery and thought
(ASCIT) proposed by Antrobus (1986, 1990). It argues that, within the Stage 1
REM period, the interval of high frequency REM and PGO activity is associ-
ated with a temporary increment in cortical activation. This increased activa-
tion supports a general increment in the production of imagery and thought,
which in turn leads to an increase in the information in the sleep mentation
report.
The EM visual orientation model and the Activation-Synthesis model
will be examined first because they contribute to the understanding of whether
the cortex is responsible for the initiation and orientation of EM within
REM sleep.

LOOKING RESPONSE-EYE MOVEMENT-PGO-

VISUAL IMAGERY MODEL
Well after the Aserinsky and Kleitman (1953) discovery of REM sleep, re-
searchers recovered the work of Ladd, who in 1892 first suggested that move-
ment of the eyes beneath the lids of sleeping subjects might indicate that the
sleeper is watching his or her dreams. But what exactly is the sleeper watching?
Rechtschaffen and Foulkes (1965) evaluated a model in which the internal
neural origin of the visual dream image is constructed from random receptor
activity in the retina. They taped open the eyes of the sleeper and, in REM
sleep, presented a number of illuminated visual forms. However, upon the
immediate awakening of the sleeper, they found no evidence to support a
model that required retinal information to be transmitted to cortical process-
ing centers.
Meanwhile, Rofiarg, Dement, Muzio, and Fisher ( 1962)tested a model
in which Stage 1 REM EMS are the motor component of looking responses
even though the visual cortex receives no visual information from the retina.
Problems with the original data analysis and several subsequent failures to
replicate the results left this model in a precarious state for over 20 years. Re-
cent research by Herman et al. (198 1); Herman, Barker, and Rofiarg (1983);
and Herman (1984), however, has supported the position that at least some
REMs are associated with directional looking responses within REM sleep.
In these studies, judges rated the concordance of EMS as measured by the
electrooculogram (EOG) and as inferred from reported visual imagery and
body movement. In both the 1962 and 1981 studies, the strength of the associ-
ation between EOG and imagery reports increased with the judges’ (but not
with the subjects’) confidence in their predictions. This confidence effect indi-
cates that at least some EMS are associated with looking responses during
dreaming, but the relation may be obscured during states of reduced cortical
6 JOHN ANTROBUS

activation within REM sleep so that the awakened dreamer is unable to recall
the relevant looking cues, or perhaps the frontal eye fields do not influence
EM direction until they are sufficiently activated by medial reticular forma-
tion (MRF) excitation beyond some minimal level. Note that it is assumed
that the visual dream imagery is produced regardless of whether the frontal
eye fields are sufficiently active to modify EM direction.
The Herman et al. (198 1) study provides the strongest evidence that
there is a relation between EM pattern and the visual imagery of REM sleep.
However, the evidence is consistent with both the looking hypothesis, which
implies that the cortex is the origin of EM control in REM sleep, and the
activation-synthesis model (Hobson & McCarley, 1977), which proposes that
REMs are controlled by activity that originates in the pons in the form of PGO
spike trains. According to the activation-synthesis model, the PGO activity is
carried to the occipital and frontal cortices, where the cortex synthesizes the
information about REM pattern to create a visual scenario that is consistent
with it.
Evidence that the pontine origin of PGO activity is out of cortical reach
has come from demonstrations of EMS in REM sleep in decorticate prepara-
tions (Jeannerod, 1966; Jeannerod & Mouret, 1962; Mergner & Pompeiano,
1981). Herman (in press) countered this interpretation of the decorticate data
by showing that the EM frequency is reduced by 80%,the bursting characteris-
tic is eliminated, and the EM patterns are stereotyped and repetitive. This
observation is perhaps the strongest evidence for the cortical control of REMs
and against the activation-synthesis position that subcortical units determine
the spatial and temporal patterns of REMs.
A second test of the looking response models of REM sleep EMS asks
whether the velocities of the EMSare similar to those of looking responses in
the waking state. The test is complicated by several factors: (a) the resting
muscle tonus is different in the two states, (b) the eyes are covered by the
lids during sleep, (c) waking saccadic movement is generally initiated by an
extrafoveal external visual stimulus, and (d) waking PGO activity tends to be
masked by the larger electromyogram activity in unrestrained animals. In two
comparisons of waking and REM EMS, Herman and colleagues (Herman et
al., 1981 ;Herman et al., 1983)found that REM EMSare quite similar to wak-
ing saccades with the eyes closed. The left-right patterns of EMS matched the
predictions of judges from the verbal reports of the subjects best when judges
took account of the imagined head movements of the dreamer. The temporal
distribution of EMS in REM sleep is, therefore, more similar to the waking
EM distribution when the head is unrestrained than when the head is held in
a fixed position. Herman and colleagues concluded, therefore, that the oculo-
motor system in REM sleep coordinates both eye and head movement.
Aserinsky, Lynch, Mack, TzankoE, and Hum (1985), however, argued
that, in REM sleep, the eye not only moves more slowly but also evidences a
“disengagement of the usual velocity-amplitude relationship seen in waking
movements” (p. 9). In the waking state, the angular velocity of the eye is pro-
Neurocognition 7

portional to the angular excursion of the movement. Aserinsky et al. (1985)

concluded that there is no evidence to support this relation within REM sleep.
However, they may have overstated their case. Their study compared REM
EMSwith 5.5 min and 1 1 min of arc excursion with EMSof similar excursion
under several waking conditions. All waking conditions, with eyes open or
closed, showed a significantly higher velocity during 1 1-min than during 5.5-
min excursions, and the increment was significantly greater than the incre-
ment during REM sleep. Nevertheless, the increment was significant in REM
sleep at the .005 level. It seems, therefore, that REM sleep may produce an
attenuation rather than a disengagement of the velocity-amplitude relation.

ACTIVATION-SYNTHESIS MODEL
The Activation-Synthesis model (Hobson & McCarley, 1977) consists of a
neurophysiologicalactivation model (McCarley & Hobson, 1979, which was
recently revised (Hobson, Lydic, & Baghdoyan, 1986), followed by a cognitive
synthesis process by which the cortex combines the neural information from
various activated subcortical locations to form dreamlike mentation. Their
“working sketch” (Hobson & McCarley, 1977, p. 1340)of the model assumes
that the “forebrain is tonically activated, probably via the midbrain RF that
is also responsible for its activation during waking. Thus the forebrain is made
ready to process information.”
They proposed that the stimuli from which dreams are constructed origi-
nate not in the “cognitive areas of the cortex” (Hobson & McCarley, 1977, p.
1347)but in a noncognitive or reflex process in the pontine brain stem, namely
PGO activity. Referring to the close association in time of PGO spikes and
REMs during Stage 1 REM sleep, they suggested that this process indirectly
provides spatial information that can be used in the construction of dream
imagery. Because the direction and temporal pattern of waking EMSand ves-
tibular activity provides some form of spatial information, they suggested that
the EMSand vestibular activity of REM sleep, even though they are randomly
driven by PGO spikes, might initiate similar forms of information processing
that might subsequently be synthesized into segments of a dream. This posi-
tion differs from that of the ASCIT model (Antrobus, 1990) in which these
brain stem generators provide only nonspecific activation to widespread areas
of neural networks in the association cortex and, thereby, determine the infor-
mation processing rate rather than more specific spatial information.
According to Hobson and McCarley (1 977, p. 1347), the process of
transforming this information into a dream is “likened to a computer search-
ing its addresses for key words”; “best fits to the relatively inchoate and incom-
plete data provided by the primary stimuli are called from memory, the access
to which is facilitated during dreaming sleep” by the tonically activated fore-
brain: “In other words, the forebrain may be making the best of a bad job
from the relatively noisy signals sent up to it from the brain stem.” The two
metaphors, of synthesis and of making the best of a bad job, might be restated
today in the more precise parallel distributed processes (PDP) vector models
8 JOHN ANTROBUS

in which neural nets engage in a process of “graceful degradation,” in which

the output of a particular neural net constitutes the best fit to the current input
to that net (McClelland, Rumelhart, & Hinton, 1986, p. 29).

Bizarreness: The Phasic Component of the

Activation-Synthesis Model
Hobson and McCarley (1977) suggested that the “ ‘bizarre’ formal features of
the dream” may be attributed to “properties of the brain stem neuronal gener-
ator mechanism” (p. 1347). They assumed that “the random but specific na-
ture of the generator signals could provide abnormally sequenced and shaped,
but spatiotemporally specific, frames for dream imagery; and the clustering of
runs of generator signals might constitute time-marks for dream subplots and
scene changes. Further, the activation by generator neurons of diffuse post-
synaptic forebrain elements in multiple parallel channels might account for
the disparate sensory, motor, and emotional elements that contribute to the
‘bizarreness’of dreams” (p. 1347).
Hobson and McCarley (1977) as well as Porte and Hobson (1986) fur-
ther suggested that the large number of the neuronal sources of information
activated during REM sleep contributes to the bizarreness of REM mentation.
By way of emphasizing an alternative to the psychoanalytic notion that
dreams are motivated by the need to disguise unconscious conflictual infor-
mation, Hobson and McCarley ( 1 977) concluded that the bizarreness of
dreams may be attributed to the difficulty of making sense of such random
input from subcortical, and therefore noncognitive, generators.
The assumption that phasic brainstem events cause segmentation of the
dreamer’sexperience implies that the phasic events are associatedwith discon-
tinuity in ongoing cognitive processes, regardless of whether discontinuities
are bizarre. Because phasic events are more frequent in REM sleep than in
Stage 2 sleep or in waking, the activation-synthesis model posits that disconti-
nuities should be most frequent in REM sleep. As reported by Reinsel, Antro-
bus, and Wollman (in press), however, changes in topic in mentation reports
are the least common in the waking state.
The finding that many characteristics of the REM/NREM report difference
disappear after total recall count (TRC; Antrobus, 1983)is partialled out sug-
gests that the greater bizarreness of REM reports may, in part, be contingent
on their greater length. TRC is the sum of all words in the mentation report
that describe the thought and imagery prior to the awakening or interruption.
It excludes commentary, associations, and redundant references. That is,
compared with waking thought, all sleep mentation may be somewhat bizarre
and, therefore, the more mentation recalled, the more bizarre the report will
be. Using the Antrobus (1983; N = 73) data set, Porte and Hobson (1986)
found that REM reports were, indeed, more bizarre than NREM reports. In
an unpublished analysis in our lab, we replicated this finding but found that
the bizarreness difference disappeared after TRC was partialled out. An in-
terim conclusion is that if any component of bizarreness does survive the par-
Neuroccgnition 9

tialling out of TRC, it will be only a minute part of the REM/NREM difference.
Knowledge of TRC can correctly sort 92.5% of REM/NREM pairs; knowledge
of bizarreness scores may add another 0.5-1 .O% at most.

BIZARRENESS: PHASIC VERSUS

TONIC REM AWAKENINGS
The Activation-Synthesis and ASCIT models both predict that mentation re-
ports will be more bizarre following phasic (REM, PIP, or MEMA) awaken-
ings than tonic awakenings. The ASCIT model assumes that PGO activity is
associated with a widespread increase in cortical activation, sufficient to acti-
vate the frontal eye fields that control REMs, so that both more thought and
more imagery are produced and reported. To the extent that sleep in general
is associated with more bizarre mentation, an increment in TRC should bring
with it an increment in bizarre mentation and, conversely, partialling out TRC
should eliminate the bizarreness increment. By contrast, according to the acti-
vation-synthesis model, phasic events should increase bizarreness but not
TRC with sleep stage held constant.
Rechtschaffen, Watson, Wincor, Molinari, and Barta ( 1972) reported
that PIP awakenings produced more bizarre mentation than tonic control
awakenings, but they later criticized their failure to obtain the reports under
conditions in which the interrogator was blind to the polygraph condition.
In a subsequent study, Watson, Bliwise, Friedman, Wax, and Rechtschaffen
( 1978)used blind procedures to demonstrate the PIP-bizarreness relation but
found it in only 1 of 4 subjects. Subsequently, Ogilvie, Hunt, Sawicki, and
Samanhalskyi ( 1982) found a significant association between MEMAs and
bizarreness in 9 subjects, averaged across REM and NREM reports. They re-
ported a borderline interaction (p < .lo) between REM/NREM and phasic-
tonic awakening for the dependent variable recall such that, relative to tonic
awakenings, phasic events (or MEMAs) were associated with more recall in
REM relative to NREM. A slight increment in bizarreness in phasic, relative
to tonic, REM reports was not formally tested. Recently, Reinsel, Antrobus,
and Wollman (in press) found no difference in the bizarreness of phasic and
tonic REM sleep reports of 19 subjects where phasic events were defined by
REM bursts, but there was a tendency (p = .08)for TRC to be greater follow-
ing the phasic or REM awakenings. (For a recent review of this literature, see
Watson & Deptula, in press.)
Aside from the possible contribution of interference and interruption to
bizarreness, the effects of under- or overactivation versus differential patterns
of activation among different cognitive modules remain a major question. The
notion of an underactivated system that has access to fragmentary informa-
tion implies a system prone to errors and, therefore, to bizarre output. How-
ever, we associate high activation with alert waking performance that is close
to error free, whereas high cognitive activation in the absence of an anchoring
effect of sensory input, as is the case in REM sleep, creates a very different
pattern of cognitive output.
10 JOHN ANTROBUS

If increased cortical activation is associated with a reduction in bizarre

sleep mentation, then the late morning dreams that occur when one sleeps
later on weekend mornings should be the least bizarre. Yet, the opposite is
true. Kondo (1988) recently compared the mentation reports of subjects who
were put to bed 3 hours later than usual (thus, all reports were made 3 hours
later) with those obtained on a control night. He argued that the late morning
reports would benefit from the rising phase of activation of the diurnal rhythm
in addition to the phase of the sleep REM/NREM cycle. All the predictions were
supported. The brightness and clarity of the visual imagery, the bizarreness of
the report, and the number of words used to describe the mentation, TRC
(logTRC), all increased in the late REM and NREM reports in the late-morn-
ing delayed-sleep condition. The REM/NREM difference remained unchanged.
Because the increment in cortical activation due to the diurnal rhythm is puta-
tively more widespread than the cortical activation associated with REM
sleep, the additive effects of the two sources constitutes strong evidence that
bizarreness is the consequence of an activated cognitive system, supported by
an activated cortical structure, in the presence of high sensory thresholds.
In conclusion, the cognitive data supports a general activation rather
than a disruptive activation origin for bizarre mentation. Nevertheless, al-
though the PGO- or phasic-bizarreness relation lacks substantial support at
the present, it may be premature to conclude that no relation exists. Future
studies must explicitly define and measure bizarreness and recall and must
describe effects both with and without partialling out TRC or a similar index
of total information reported.

RELATION OF REM SLEEP TO VISUAL IMAGERY:

NEUROPHYSIOLOGICAL EVIDENCE
This section examines the evidence for the role of the cerebral cortex, and
the cognitive processes it supports, in the control of EMS within REM sleep.
Although it may be assumed that many of the structures essential to visual
perception contribute to the generation of spontaneous visual imagery, there
is no strong evidence about which parts of the brain contribute to the produc-
tion of spontaneous visual imagery in either waking or sleep. It should be
cautioned that cortical models for the generation of visual imagery in response
to instruction or command (Kosslyn, 1988), where the concept or meaning
precedes the image, may be inappropriate for the production of REM sleep
imagery, where the temporal precedence of meaning and image is unknown.
Early sleep imagery studies started by examining the role of the retina in
REM sleep. Pompeiano’s (1970) finding that there is no neural transmission
from the retina to the visual cortex in REM sleep suggests that the visual im-
ages of REM sleep have an extraretinal origin and many even have multiple
sources. As previously mentioned, Rechtschaffen and Foulkes ( 1965) found
no evidence that the retina is sensitive to illuminated objects presented during
REM sleep to subjects whose eyelids had previously been taped open.
Once the cognitive system becomes unresponsive to external stimuli it
becomes, by definition, a closed system. At that point, the search for the origin
Neurocognition 11

of the information that flows through the system must be expanded to include
all the feedforward and feedback pathways that become active when the sys-
tem is not externally driven as well as an account of how these pathways are
state (waking, REM, NREM) dependent. This opens up a variety of new possi-
bilities concerning the relation between REMs and visual imagery. Spontane-
ous independent activity might occur in spatial, visual movement and higher
processing modules but might be transmitted to neighboring modules only
when the system is sufficiently activated by the MRF. At that point, a winner-
takes-all effect may allow one module to dominate another. For example, a
visual module may produce an event that higher cognitive modules are
obliged to accommodate to, or the higher cognitive module may inhibit in-
compatible productions in the lower visual modules. None of this visual
dreaming production need involve the oculomotor system unless the frontal
eye fields and the brain stem oculomotor systems are sufficiently well-
activated.
Because this kind of information is not accessible to introspection in
waking or in sleep, one must examine the patterns of neural activity in the
structures putatively responsible for the relevant component cognitive pro-
cesses in order to reduce such speculations to a reasonable model. However,
detailed evidence on the neural circuits that control EMScan only come from
animal studies, and this necessarily precludes examination of the relation of
neural activity to visual imagery in different components of this complex
system.
The general strategy pursued here is to identify the pathways that enable
EM control in waking perception and to then determine how closely this ac-
tion pattern operates in REM sleep. In waking perception, the decision to
move the eyes is distributed over many cortical and subcortical centers and is
based on a wide variety of information classes such as visual, auditory, and
eye and head position and velocity as well as on higher cognitive judgments
and values. Within the cortex, much of the perceptual and cognitive informa-
tion is funneled through the frontal eye fields that putatively decide where in
space to move the eyes and decide to communicate this, typically, to the supe-
rior colliculus, which further refines the decision and passes the command to
the oculomotor system that programs the EMS. However, the fine tuning of
the motor execution is camed out in a sideways loop into the seventh vermal
lobule of the cerebellum (Noda, Murakami, Yamada, & Aso, 1988), which
takes account of eye position and velocity relative to head and body position.
Before we examine some of the components that contribute to EMS in
waking and sleep, it is important to note that the activation of all components
is dependent on MRF activation. This activation, largely pontine in origin,
not only activates the cortical areas that produce both visual imagery and EM
decisions but also drives the final leg of the oculomotor system, namely, the
oculomotor neurons. At the most general level then, the co-occurrence of vi-
sual imagery and EMSin REM sleep is dependent on MRF activation. Within
REM periods, the occurrence of REM may also be due to a higher level of
MRF activation than that which exists during periods of ocular quiescence,
12 JOHN ANTROBUS

but evidence supporting this hypothesis is only indirect. Of particular interest

is whether activity in certain components of the oculomotor system during
REM sleep can be taken as evidence of information processing analogous to
that carried out by the structures during waking perception. We will first look
at the relations between the visual cortex, EMS,and MRF activation in waking
perception and REM sleep.
Breitmeyer’s ( 1986) analysis of visual masking and saccadic EMSoffers
a useful hunch about the function of both MRF and PGO activation of the
lateral geniculate nucleus (LGN). During waking saccades, the fovea is contin-
uously smeared with a succession of visual patterns. The images appear
smeared because the retina input summates before it reaches the cortex. This
is not a serious problem at the onset of the saccade because the visual system
is only interested in the target stimulus at the termination of the saccade. But
a problem does arise if the next-to-final retinal image summates with the target
image to smear or mask the target image.
Breitmeyer (1986) cited the work of Cohen, Feldman, and Diamond
(1969); Ogawa (1 963); Singer ( 1977a, 1977b); Singer and Bedworth ( 1973);
Singer, Tretter, and Cynader ( 1975);and Tsumoto and Suzuki ( 1976)to argue
that the MRF, possibly under the guidance of the frontal eye fields, provides
phasic activation to the LGN and the LGN-cortical pathway to assist selective
processing of the foveal image at the termination of the saccade. Breitmeyer
noted that “this reset would occur at the beginning of each new fixation inter-
val by counteracting any persisting effects of saccade suppression exerted on
the response of these sustained neurons as a result of the activation of transient
channels during the saccade” (1986, p. 77). Perhaps the waking LGN activity
level was well below that of REM sleep (McCarley et al., 1983)because of the
marked reduction of visual pattern in the waking research environment of
their cats. Extended to REM sleep, the Breitmeyer model suggeststhat looking
responses initiated in the cortex, possibly in the frontal eye fields, control with
a 60- 100-ms latency a LGN-visual cortex activation that, in the absence of
retinal input to the LGN, activates or perhaps primes the visual cortex in a
manner that supports the construction of visual imagery. This speculation, of
course, assumes that the cortex plays some role in initiating saccades in REM
sleep. This assumption will be examined after a look at the relation between
PGO activity and the occipital cortex projection area for the LGN.
McCarley et al. (1983) plotted cortical evoked potentials that were timed
from the onset of REM sleep EMSin human subjects. They observed a pane-
tal-occipital evoked response that began 27.6 ms prior to REM onset and
peaked 5.6 ms after REM onset (range = 0-12 ms), followed by a low ampli-
tude negative wave. The location of the spike relative to EM onset in REM
sleep was later than the 10-40 ms peak prior to the saccade onset that was
reported by Kurtzberg and Vaughan ( 1982)for waking human subjects. Sam-
ple sizes larger than the 5 and 6 subjects that were respectively used in the two
studies are necessary to determine if there is a significant difference in the
timing and location of the REM and waking EM-triggered cortical-evoked
potentials.
Neurocognition 13

Although McCarley et al. (1983) found that EMS were contralateral to

the occipital-parietal response, Monaco, Baghdoyan, Nelson, and Hobson
( 1984), using implanted electrodes, found the EMSto be ipsilateral to the oc-
cipital response. This apparent discrepancy may have occurred because scalp
electrodes linked to a common reference to measure evoked potentials do not
necessarily provide a measure of cortical activity immediately below the elec-
trode (Fein, Raz, Brown, & Memn, 1988) and because electrodes may re-
spond to electromagnetic projection from neural dipole that are sometimes
remote from the electrode (Goq Williamson, Van Gilder, Allison, & Fisher,
1980; Kaufman, Okada, Tripp, & Weinberg, 1982), perhaps even in the con-
tralateral hemisphere. The assumption of contralateral dipole projection
would render the EM-posterior cortex findings of the two studies congruent.
Cortical-evoked potentials have never been computed in sleep summed from
the termination point of saccades, the onset of fixation. Using an estimated
average of 25 ms for saccade duration, however, would place the peak of the
cortical response approximately 19 ms prior to fixation. This lead time seems
compatible with Breitmeyer’s (1986) model of a cortically controlled, MRF-
energized preparatory activation of the visual cortex just prior to foveation.
Further research with waking saccades is necessary to determine whether the
occipital-evoked potentials identify visual cortex enhancement processes tied
to the onset of ocular fixation, or rather, to a presaccade decision process.
Of course, there is no foveation following saccades in REM sleep. In light
of the salience of visual imagery in that state, however, one might speculate
that this phasic activation of the visual cortex supports the visual analysis of
whatever information is resident on the visual networks at the time of activa-
tion. For example, top-down processes in the association cortex might bias
neural nets in earlier modules of the visual system to produce visual images
that “fit” ongoing thought (see Antrobus, 1990). However, this conception
also presupposes some cortical influence on the timing of PGO activity, a posi-
tion that has been discounted by the identification of PGO burst cells in
the pons.
The argument for pontine control of REMs is based on evidence that
EMSin REM sleep appear to originate in the premotor neurons of the pontine
reticular formation (Buttner, Buttner-Ennever, & Henn, 1977; Henn, Butt-
ner-Ennever, & Hepp, 1982; Henn, Hepp, & Buttner-Ennever, 1982). Among
these neurons are pontine giant cells that project to the oculomotor and vestib-
ular neurons (Pompeiano, 1980). They exhibit similar patterns of discharge in
both waking and REM sleep. McCarley, Nelson, and Hobson ( 1978) identified
burst cells, primarily in the brachium conjunctivum, which is one of the neu-
ral branches that joins the pons to the cerebellum. Inasmuch as the burst cells
are necessary to PGO activity, they appear to be the origin of the PGO wave
generator system. They fire at high rates during horizontal EMS and precede
ipsilateral lateral geniculate activity (Monaco et al., 1984; Nelson, McCarley,
& Hobson, 1983).
Because the majority of the REM burst cells are in the brachium con-
junctivum and because the brachium conjunctivum carries the output of the
14 JOHN ANTROBUS

cerebellum to the brainstem, we must consider the possibility that this bursting
activity in REM sleep is triggered by cerebellar processes. This hypothesis is
also supported by the finding that discharge in the tegmental reticular nucleus
of Bechterew, which has cerebellar connections, is coherent with both PGO
waves and REMs and is also coupled with EM during waking (McCarley,Nel-
son, Hobson, & Strassman, 1981).
The role of the cerebellum in coordinating EM and head movement with
spatial search is well-known, and the cerebellum is generally thought to fine
tune the general EM control commands initiated in the frontal eye fields of
the cerebral cortex (Bruce & Goldberg, 1985). Further support for this conjec-
ture is provided by Hobson and McCarley’s (1972) demonstration that the
Purkinje cells of the cerebellum are more active in REM than in either NREM
sleep or waking and that, within REM sleep, they are more active during
REMs than ocular quiescence. Activation of the cerebellum in conjunction
with REMs is not evidence, of course, that the cerebellum is processing cortical
input, but it does demonstrate the operation ofone further link between cortex
and the oculomotor system during REM sleep EMS.
Although the major pathway from frontal eye fields to the brain stem
includes the superior colliculus, little is known about the relation between
collicular and cerebellar unit activity. The McCarley et al. (198 1) study shows
that it does not have direct input to the PGO units.
If the precise route from cerebral cortex to the burst cells of McCarley et
al. ( 1978)in REM sleep is to be identified with precision, it will require exten-
sive prior tagging in the waking state of individual neurons in the frontal eye
fields-cerebellum-burst cell sequence. Individual adjacent units, such as hold
units and fire units (particularly in the frontal eye fields), often serve quite
different functions in waking perception (Bruce & Goldberg, 1985), so that
measures of lead time and coherence between different nuclei or fields are
of little value if the comparisons are based on random samples of units in
each group.
The absence of high frequency and high amplitude burst cells in the cere-
bral cortex is not, in itself, evidence that the pontine burst cells are free of
cortical control. As described later, the oculomotor neurons have unique en-
ergy requirements, and only an inefficient brain design would place these cells
in the cerebral cortex surrounded by dense associative neuronal pathways and
distant from the oculomotor neurons. A better design would place the com-
plex decision making in the cortex and would place the neurons with high
volley rates, which it modulates, close to the oculomotor neurons, namely, in
the brain stem. The location of the burst cells and the input to the oculomotor
neurons in the MRF seem to fit this organization.

BURSTING CELLS AND STATES OF SLEEP AND WAKING

What are the implications of the shifts in firing patterns of individual units
across states of sleep and waking? This question constitutes a major puzzle
concerning the function of PGO and burst cell activity in REM sleep, particu-
Neurocognition 15

larly their relation to the visual imagery of dreaming sleep. The oculomotor
neurons always fire in bursts, proportional in duration to the projected size of
the saccade. But many neurons show a bursting pattern only when they are not
processing or transmitting information, and this is particularly true of sleep,
especially NREM sleep and the transitions from either waking or NREM to
REM sleep (Greene, Haas, & McCarley, 1986). Many of the rhythmic neural
patterns such as electroencephalogram (EEG) alpha and spindles occur when
the eyes close or in sleep states. These bursting patterns are generally termi-
nated by activation from the MRF (McCarley, Benoit, & Barrionuevo, 1983).
The bursting and rhythmic patterns within individual units can be determined
from the electrochemical properties of the units, but their functions are not
understood and may differ from structure to structure (Llinas, 1989).
It is the pattern of firing across a large set of neurons that putatively
carries the information processed by a nucleus or field. However, almost noth-
ing is known about how the shifts in spike patterns within a given unit affect
the pattern of a network of units within a given structure. Perhaps one excep-
tion to this pessimistic picture is the bursting patterns that lead to recruitment
of neighboring cells and lead to widespread rhythmic activity. In this case,
surely, the neurons carry no cognitive information other than the message that
the structure is in an off state.
The information carried by the pattern of PGO spikes is more difficult
to decode. The pontine leg shows a bursting pattern (Greene et al., 1986; Nel-
son et al., 1983), but its input to the lateral geniculate does not (Fourment &
Hirsch, 1980), even though other geniculate units do show bursting patterns
during REM sleep. Monaco et al. ( 1984) found that the primary spike in occip-
ital PGO waves is correlated with EM direction and amplitude in REM sleep
but not in waking.
In conclusion, it is clear that the information available about the EM
production system in REM sleep is, by itself, insufficient to support any strong
model regarding EM activity and the cognitive process of visual imagery pro-
duction. The evidence does not clearly support the cortical origin of EM con-
trol proposed by Herman et al. ( 198 I), but neither does it support the Hobson
and McCarley ( 1977) position that the REMs are totally free of cognitive con-
trol. Finally, the uninterpretability of cortical PGO bursting and high fre-
quency spiking does not support the activation-synthesis hypothesis that the
spikes interrupt ongoing cognitive processes to produce bizarre mentation.
Even the EM information identified in the occipital cortex by Monaco et al.
(1984) must remain suspect inasmuch as it is not seen in the waking state.
This latter point suggests one recommendation for all future research on
this issue, namely, what are the appropriate waking controls for the study of
EMSin REM sleep? If dreaming, as imaged perception, provides the context
for an EM, then the waking control condition must approximate the imagined
perceptual event. Because experimenters have no access to each subject’s
dreams, they must at the very least employ a variety of illumination levels and
visual events as well as head position and movement conditions. Obviously, if
16 JOHN ANTROBUS

a neuron’s response is invariant over a range of these conditions, the multiple

controls could be eliminated. But McIlwain (1988) has demonstrated at least
one case where it is not. The trajectory of an electrically evoked saccade is
dependent on the initial position of the eye.
The waking controls of most animal sleep studies have been poorly de-
fined, generally studying animals under low arousal and low illumination with
little or no visual pattern or movement, and often placing the animal under
restraint (with the exception of Sakai & Jouvet, 1980), so that no head move-
ments are possible. Thus, the conditions that the visual system is designed to
respond to rarely occur in the waking control conditions. Obviously, there is
no single waking control condition that is suitable for the study of REM sleep
saccades.

VESTIBULAR, CEREBELLAR, AND OCULOMOTOR SYSTEM

Vestibular activity has often been suggested as a contributor to the characteris-
tics of dream imagery. The horizontal posture of sleep has been suggested as a
factor in the production of dreams of flying and falling. The intimate relation
between head movement and eye movement in the waking state is sufficient
reason to consider the possible contribution of vestibular activity to EMS in
REM sleep. If a relation does exist during sleep, it is additional reason to com-
pare unrestrained with restrained animals in waking control conditions in
studies of the neural control of EMS.
During PGO activity in sleep, the pontine giant cells activate the vestibu-
lar neurons (Peterson, Franck, Pitts, & Daunton, 1976; Pompeiano, 1970,
1980). Pompeiano ( 1970) observed that high rates of medial and descending
vestibular nuclei firing (80- 160 spikes/s) were observed simultaneously with
EMS in REM sleep but not in waking or synchronous sleep. Nelson et al.
( 1983) observed in restrained cats that, in waking, the PGO cells fire in single
spikes rather than bursts, as in REM sleep. The similarity of the burst cell to
the EM firing pattern is a criterion for ascertaining the similarity in the func-
tion of the waking and REM sleep EM control system and the function of
PGO spikes. But are these comparisons valid ifthey fail to account for concur-
rent vestibular activity?
In the waking state, the medial and superior vestibular nuclei act on the
third and sixth nerve to coordinate EMS with head movements during sac-
cadic and tracking EMS, but during sleep there is neither retinal information
to control eye trajectories nor head movements with which the EMS must be
coordinated. The vestibular neurons extend to the paramedial reticular forma-
tion, and in the waking state, both vestibular and reticular neurons contribute
to the excitation of the burst cells that control horizontal eye movements in
the waking state. However, although the medial vestibular neurons modify
EMS, it is the descending vestibular neurons that coordinate eye and head
movement information with the cerebellum and higher cortical centers via
the medial longitudinal fasciculus to the thalamus and, possibly, from there
to the occipital cortex.
Neurocognition 17

Ito ( 1974) described the vestibular-cerebellar-ocular network by which

head movement and vestibular information modifies the cerebellar burst cell
firing rates in the waking state. This network suggests part of the circuit by
which the vestibular nuclei, which are quite active in REM sleep, can modify
EMSvia the cerebellar burst cells. Extensive afferents to the cerebellum from
the association cortex (Kornhuber, 1974) indicate pathways by which cogni-
tive processes could modify waking saccades and may similarly modify sleep
EMSwhen the cortex is sufficiently activated, as in REM sleep, by the MRF.
The active involvement of the vestibular neurons raises the possibility
that saccade control centers are programming REM EMSas though there were
simultaneous head movements, as supported by the looking response findings
of Herman et al. ( 1981). This complication emphasizes the difficulty of finding
the appropriate waking head movement and visual stimulus conditions
against which to compare REM EMS. None of the animal PGO EM sleep
studies provide sufficiently explicit descriptions of their waking procedures or
systematically vary their waking EM conditions so that an adequate assess-
ment of the REM waking EM similarities and differences can be made.
Although the role of the vestibular nuclei in the modification of sleep
mentation has been discussed only in the context of EMS, these nuclei may
provide a trigger for dreams of falling or floating. Vestibular activity, when
accompanied by a sporadic drop in the Stage 1 REM inhibition of alpha motor
neurons, may account for the dreamer’s experience of waking following a large
body jerk that is accompanied by a dreamed episode of being off balance. The
studies of the relation between REM sleep EMS and sleep mentation reports
cited earlier by Roffwarg, Herman, and colleagues indirectly support this
model. A significantly greater correlation between EM direction and the re-
ported directional shift of the dreamer’s gaze is observed when taking account
of head movements than when ignoring them. Experimental evidence for this
notion could be obtained in human sleep by means of pneumocaloric stimula-
tion of the vestibular system.

PGO-BIZARRENESS MODEL REVlSlTED

As noted earlier, the Activation-Synthesis account of bizarreness rests on the
assumption that PGO activity originates in the pons and is independent of
cortical control and, therefore, is free of any input from antecedent cognitive
processes. The cortex is informed of EMS by an independent route (Hobson
& McCarley, 1977)and does the best it can in synthesizingthe information it
receives. If, as proposed here, the cortex uses cognitive information from sev-
eral sources to initiate, control, or modify saccadic EMSvia a cerebellar route
rather than receives EM information post hoc, then the bizarreness explana-
tion of the activation-synthesis model is not warranted.

Crick and Mitchison Theory of REM Sleep and Neural Nets

Crick and Mitchison (1983, 1986)have recently proposed that the production
of REM dreams represents a cognitive process in which neural networks that
18 JOHN ANTROBUS

have been “overloaded” (Crick & Mitchison, 1986, p. 234) as a consequence

of waking perceptual and cognitive activity can once again become receptive
to new information when they reawaken. The proposal is suggested by the
finding that math models of “gridlocked” neural nets can be unlocked by ran-
dom input to the net. If PGO spikes in REM sleep constitute such a random
generator, then it is reasonable to posit such a function for PGO spikes. If, as
proposed here, however, PGO spikes are a lawful or nonrandom output of a
component of the visuomotor system, then a major assumption of the Crick
and Mitchison ( 1983, 1986) model is untenable.
The evidence reviewed here for the role of the cerebellum in the pro-
gramming of EMs in REM suggests that the possible role ofthe cortex in creat-
ing the visual imagery that is coordinated with EMS during REM sleep be
reexamined. Might not the associative cortex attempt to interpret pattern in-
formation initiated in the occipital cortex, which in turn may result in an
oculomotor decision to foveat the visual image, an apparent perception, in the
foveal projection area of the occipital cortex? And furthermore, might it not
be possible, since the retina receives no input in REM sleep and the EM cannot
therefore foveate the imaged stimulus, that the movement command is given
repeatedly, thereby producing a visual “stutter” that we see as a REM burst?

Visual Imagery and REMs

A number of studies have looked for an association between REMs and other
qualities of visual imagery such as vividness, clarity, color saturation, and
number of visual features. In “what was in some measure a post hoc analysis”
(see, Foulkes and Pope, 1973, p. 1 15) of reports from phasic and tonic Stage
1 REM and Stage 2, Molinari and Foulkes (1969) found that REM-phasic
versus tonic awakenings were associated with more reports of “primary visual
experience” (p. 351); tonic REM and Stage 2 reports were associated with
“secondary cognitive elaboration” (p. 35 I), such as reports of thinking, being
awake, recognizing, or interpreting. Unfortunately, the two scales were con-
founded. The Primary Visual Imagery scale was scored residually only if the
mentation report did not contain secondary cognitive elaboration. Further,
some categories of secondary cognitive elaboration were added “post hoc if
they proved to discriminate REM (burst) from (quiescent)” (p. 35 l), that is,
phasic from tonic intervals. This procedure of scale construction confounds
any conclusion about the relation of visual imagery to phasic Stage 1 REM
versus other states of sleep. In 1973, Foulkes and Pope attempted a replication
of the study using the same scales, but the phasic-tonic difference in primary
visual experience held only for the spontaneous portion of the sleep reports.
The authors emphasized that visual imagery, with or without secondary cogni-
tive elaboration, is reported from all phasic and tonic REM sleep awakenings.
Reinsel et al. (in press) just found that phasic, relative to tonic, REM reports
are not associated with an increase in verbally reported visual imagery as mea-
sured by the log of the number of visual nouns, modifiers, and spatial preposi-
tions taken as a rough index of the number of visual features in the report. In
Neurocognition 19

a critical review of phasic-tonic relations, Pivik ( 1978) concluded that, except

for the Foulkes and Pope (1973) study, “we appear to be dealing with a rather
weak effect” (p. 269).
A major handicap in the search for the neural substrates of visual imag-
ery is that our inferences about visual characteristics have up to now been
based exclusively on verbal reports. Rechtschaffen’s( 1983)recent success with
the use of photographs to identify visual characteristics such as color satura-
tion and visual clarity without regard to specific form or meaning offers an
experimental tool that encourages us to reassess our models of visual image
production. Rechtschaffen constructed a large number of photographic varia-
tions of a single scene. He presented a book containing the photographic varia-
tions to subjects as they awoke from different classes of sleep and asked them
to select the single photograph that was most like their preceding sleep menta-
tion. The model selection was the natural full color, full clarity photograph
that had been used to construct the wide range of clarity, brightness, color
saturation, and figure-ground variations in clarity and other distortions. One
of the most noteworthy achievements of this instrument is its ability to distin-
guish a characteristic of phasic versus tonic REM mentation that has not been
picked up by verbal report measures, specifically visual clarity, scaled in mag-
nitude of focus versus blur in the photograph. Phasic REM visual imagery was
judged higher in overall visual clarity than imagery during tonic REM sleep
(p < .02, N = 25). Color saturation, brightness, and the remaining measures
showed no change. Because Rechtschaffen did not obtain verbal mentation
reports, the relation of the visual responses to verbal indices of visual menta-
tion characteristics is unknown. If, indeed, the photoimagery instrument is
sensitive to characteristics of visual imagery in sleep that verbal reports are not
sensitive to, then the two types of instruments together may be able to provide
a more valid test of the neuropsychological models of sleep mentation that
have made predictions about variations in visual imagery. For example, an
increase in visual characteristics of mentation during phasic REM is a critical
prediction for the EM models of dreaming. The absence of support from ver-
bal reports (Reinsel et al., in press) may be rectified by the photoimagery
measure.
Recently, Antrobus, Hartwig, Rosa, Reinsel, and Fein (1987) scaled
variations on a single color photograph for brightness and clarity of focus using
a magnitude estimation technique where 100represented the focus and bright-
ness of normal waking vision. Photos were arranged in a 4 X 4, Brightness X
Focus matrix. Upon awakening, subjects selected the photo that best repre-
sented the visual quality of each of the brightest and clearest (maximum = 5)
objects or persons in their sleep mentation. Across all conditions, the bright-
ness of visual imagery approximated that of visual perception more than did
the clarity of focus, and both indices were highest during waking, slightly lower
for REM, and lowest for NREM imagery responses. Brightness means were
88,77, and 5 1, and clarity means were 79,69, and 44 for waking, REM, and
NREM states, respectively. The result suggeststhat there may be different neu-
20 JOHN ANTROBUS

ral control mechanisms for the two properties and that it may be worthwhile,
in future research, to distinguish among various properties of visual imagery.
Despite Rechtschaffen’s support for the phasic REM-visual imagery
model, the precise relation between the clarity of visual imagery, on the one
hand, and the REM sleep PGO and subcortical and cortical pattern of activa-
tion, on the other, is unclear. Although the PGO-associated occipital response
(Nelson et al., 1983) may include no external visual pattern information, it
may produce a general depolarization of pattern receptor cell units in areas
16, 17, and 18 so that higher order hypercomplex cells of the peristriate cortex
can be activated to produce visual patterns that are consonant with the cogni-
tive sequences generated in the associative cortex. Thus, mentation may be
produced throughout REM sleep but may generate more visual detail during
PGO-vestibular-related activation of the occipital cortex.
Although the role of the vestibular nuclei in the modification of sleep
mentation has been discussed only in the context of EMS, these nuclei may
provide a trigger for dreams of falling or floating. Vestibular activity, when
accompanied by a sporadic drop in the Stage 1 REM inhibition of alpha motor
neurons, may account for the dreamer’s experience of waking following a large
body jerk that is accompanied by a dreamed episode of being off balance. The
studies of the relation between REM sleep EMS and sleep mentation reports
by Roffwarg, Herman, and colleagues cited earlier indirectly support this
model. A significantly greater correlation between EM direction and the re-
ported directional shift of the dreamer’s gaze is observed when head move-
ments are considered rather than ignored. Experimental evidence for this
notion could be obtained in human sleep by means of pneumocaloric stimula-
tion of the vestibular system.

CONCLUSION
This chapter has reviewed evidence in support of the position that the PGO-
EM link is part of a feedforward system for the control of waking saccades.
Although waking saccades may be elicited by either visual or auditory input,
the executive decision to emit a saccade during waking and REM sleep appears
to be made in the cortex. Although this system is down throughout most of
sleep, the activation of the cortex, vestibular, and oculomotor system by the
MRF during portions of stage 1 REM is sufficient for the system to output
P W and REM periodically.The cognitive role in this cerebral cortex-cerebel-
lum-PW-EM sequence is demonstrated by the association of EMs in REM
sleep with the clarity and direction of visual imagery and with the total amount
of information reported. Eye movements and other phasic activity in REM
sleep are therefore interpreted as evidence of a temporary increment in the
activation of the cortex, including the visual association structures that gener-
ate the imaginal characteristics of dreaming.
The lawfulness of this control system is evidence against the interpreta-
tion of PGO activity as a random process. Therefore, it is evidence against
both Hobson and McCarley’s ( 1977) proposal that PGO is a random process
Neurocognition 21

that may, among other things, account for the bizarreness of sleep mentation
and Crick and Mitchison’s ( 1 986) proposal that PGO activity is a random
process that frees gridlocked neural nets during REM sleep.

References
Antrobus, J. S. (1983). REM and NREM sleep reports: Comparison of word frequen-
cies by cognitive classes. Psychophysiology, 20, 562-568.
Antrobus, J. ( 1986).Cortical and cognitive activation, perceptual thresholds, and sleep
mentation. Journal ofMind and Behavior, 7, 193-2 10.
Antrobus, J. ( 1990).Cognitive and cortical activation and sleep mentation. Manuscript
submitted for publication.
Antrobus, J., Hartwig, P., Rosa, D., Reinsel, R., & Fein, G. (1987). Brightness and
clarity of REM and NREM imagery: Photo response scale. Sleep Research, 16,
240.
Aserinsky, E., & Kleitman, N. (1953). Regularly occurring periods of ocular motility
and concomitant phenomena during sleep. Science, 118,361-375.
Aserinsky, E., Lynch, J. A., Mack, M. E., Tzankoff, S. P., & Hum, E. (1985). Compari-
son of eye motion in wakefulness and REM sleep. Psychophysiology, 22, 1-10.
Breitmeyer, B. C. (1986). Eye movements and visual pattern perception. In E. C.
Schwab & H. C. Nusbaum (Eds.), Pattern recognition by humans and machines
(pp. 65-87). New York Academic Press.
Bruce, C. J., & Goldberg, M. E. (1985). Primate frontal eye fields: I. Neurons discharg-
ing before saccades. Journal of Neurophysiology, 53,603-635.
Buttner, U., Buttner-Ennever, J. A., &Hem, V. (1977). Vertical eye movement related
unit activity in the rostra1 mesencephalic reticular formation of the alert monkey.
Brain Research, 130, 239-252.
Cohen, B., Feldman, H., & Diamond, S. P. (1969). Effects of eye movement, brain
stem stimulation, and alertness on transmission through lateral geniculate body
of monkey. Journal of Neurophysiology, 32, 583-594.
Crick, F., & Mitchison, G. (1983). The function ofdream sleep. Nature, 304, 11 1-1 14.
Crick, F., & Mitchison, G. (1986). REM sleep and neural nets. Journal of Mind and
Behavior, 7, 229-250.
Fein, G., Raz, J., Brown, F. F., & Memn, E. L. (1988). Common reference coherence
data is confounded by power and phase effect. Electroencephalographand Clinical
Neurophysiology, 69, 58 1-584.
Foulkes, D., & Pope, R. (1973). Primary visual experience and secondary cognitive
elaboration in stage REM: A modest confirmation and extension. Perceptual and
Motor Skills,3 7, 107- 1 18.
Fourment, A., & Hirsch, J. C. (1980). Synaptic potentials in cat’s lateral geniculate
neurons during natural sleep with special reference to paradoxical sleep. Neurosci-
ence Letters, 16, 149-154.
Goff, W. R., Williamson, P. D., Van Gilder, J. C., Allison, T., & Fisher, T. C. (1980).
Neural origins of long latency evoked potentials recorded from the depth and
cortical surface of the brain in man. In J. E. Desmedt (Ed.), Progress in clinical
neurophysiology (Vol. 7, pp. 126- 145). Basel, Switzerland: Karger.
Greene, R. W., Haas, H. L., & McCarley, R. W. (1986). A low threshold calcium spike
mediates firing pattern alternations in pontine reticular neurons. Science, 234,
738-740.
Hem, V., Buttner-Ennever, J. A., & Hepp, K. (1982). The primate oculomotor system:
I. Motoneurons. A synthesis of anatomical, physiological, and clinical data. Hu-
man Neurobiology, I , 77-85.
22 JOHN ANTROBUS

Henn, V., Hepp, K., & Buttner-Ennever,J. A. ( 1 982). The primate oculomotor system:
11. Premotor systems: A synthesis of anatomical, physiological, and clinical data.
Human Neurobiology, I, 87-95.
Herman, J. H. (1984). Experimental investigations of the psychophysiology of REM
sleep including the question of lateralization. Research Communications in Psy-
chiatry and Behavior, 9, 53-15.
Herman, J. ( 1989). Transmutative and reproductive properties of dreams: Evidence
for cortical modulation of brain stem generators. In J. A. Antrobus and M. Bertini
(Eds.), The neuropsychology of dreaming sleep. Hillsdale, NJ: Erlbaum.
Herman, J. H., Barker, D. R., & R o h a r g , H. P. (1983). Similarity of eye movement
characteristicsin REM sleep and the awake state. Psychophysiology, 20, 537-543.
Herman, J. H., Rofiarg, H. P., Taylor, M. E., Boys, R. M., Steigman, K. B., &Barker,
D. R. (1981). Saccadic velocity in REM sleep dreaming, normal visual activity,
and total darkness. Psychophysiology, 18, 188.
Hobson, J. A., Lydic, R., & Baghdoyan, H. A. ( 1 986). Evolving concepts of sleep cycle
generation: From brain centers to neuronal populations. Behavioral and Brain
Sciences, 9, 37 1-448.
Hobson, J. A., & McCarley, R. W. ( 1972). Spontaneous discharge rates of cat cerebellar
Purkinje cells in sleep and waking. Electroencephalograph and Clinical Neuro-
physiology, 33,451-469.
Hobson, J. A., & McCarley, R. W. (1977). The brain as a dream state generator: An
activation-synthesis hypothesis of the dream process.American Journal of Psychi-
atry, 134, 1335-1348.
Ito, M. ( 1 974). The control mechanisms of the cerebral motor systems. In F. 0.Schmitt
& F. G. Worden (Eds.), The neurosciences: Third study program (pp. 293-303).
Cambridge, MA: MIT Press.
Jeannerod, M. ( 1966). The phasic phenomena of paradoxical sleep. Revue Lyonnaise
du Medicine, 15, 27-44.
Jeannerod, M., & Mouret, J. ( 1 962). Etude des mouvements oculaires observes chez
I’homme au cours de la veille et du sommeil [The study of eye movements in man
observed during the course of waking and sleep]. Society de Biologie de Lyon, 25,
1407-14 10.
Kaufman, L., Okada, Y., Tripp, J., & Weinberg, H. (1982). Evoked neuromagnetic
fields. Annals of the New York Academy of Sciences, 388, 191-2 13.
Kondo, T. (1 988). Late REM activation and sleep mentation. Unpublished master’s
thesis, The City College of New York.
Kornhuber, H. H. (1 974). Cerebral cortex, cerebellum, and basal ganglia: An introduc-
tion to their motor function. In F. 0. Schmitt & F. G. Worden (Eds.), The neuro-
sciences: Third study program (pp. 267-280). Cambridge, MA: MIT Press.
Kosslyn, S. M. (1 988). Aspects of a cognitive neuroscience of mental imagery. Science,
240, 162 I - 1626.
Kurtzberg, D., & Vaughan, H. G. (1982). Topographic analysis of human cortical po-
tentials preceding self-initiated and visually triggered saccades. Brain Research,
243, 1-9.
Ladd, G. ( I 892). Contributions to the psychology of visual dreams. Mind, I, 299-304.
Llinas, R. R. (1989). The intrinsic electrophysiologicalproperties of mammalian neu-
rons: Insights into central nervous system function. Science, 242, 1654-1 664.
McCarley, R. W., Benoit, O., & Bamonuevo, G. (1983). Lateral geniculate nucleus
unitary discharge in sleep and waking: State- and rate-specific aspects. Journal of
Neurophysiology, 50,798-8 18.
McCarley, R. W., & Hobson, A. (1975). Neuronal excitability modulation over the
sleep cycle: A structural and mathematical model. Science, 189, 58-60.
McCarley, R. W., Nelson, J. P., & Hobson, J. A. (1978). Ponto-geniculo-occipital
(PGO)burst neurons: Correlativeevidence for neuronal generators of PGO waves.
Science, 201,269-272.
McCarley, R. W., Nelson, J. P., Hobson, J. A., & Strassman, A. ( 198 1). A cross-correlo-
Neurocognition 23

gram study of PGO-related neuronal activity in tegmental reticular nucleus, cen-

tral tegmental field, and superior colliculus. Sleep Research, 10, 38.
McCarley, R. W., Winkelman, J. W., & DUG,F. H. ( 1 983). Human cerebral potentials
associated with REM sleep rapid eye movements: Link to PGO waves and waking
potentials. Brain Research, 274, 359-364.
McClelland, J. L., Rumelhart, D. E., & Hinton, G. E. (1986). The appeal of parallel
distributed processing. In J. L. McClelland & D. E. Rumelhart (Eds.), Parallel
distributedprocessing: Explorations in the microstructure of cognition (Vol. 1, pp.
3-44). Cambridge, MA: MIT Press.
McIlwain, J. T. (1988). Effects of eye position on electrically evoked saccadis: A theo-
retical note. Visual Neuroscience, I , 239-244.
Mergner, T., & Pompeiano, 0.( 1 98 1). Basic mechanisms for saccadic eye movements
as revealed by sleep experiments. In A. Fuchs & W. Becker (Eds.) Progress in
oculomotor research (pp. 107- 1 14). New York: Elsevier/North Holland.
Molinari, S., & Foulkes, D. ( 1969). Tonic and phasic events during sleep: Psychological
correlates and implications. Perceptual and Motor Skills, 29, 343-368.
Monaco, A. P., Baghdoyan, H. A., Nelson, J. P., & Hobson, J. A. ( 1 984). Cortical wave
amplitude and eye movement direction are correlated in REM sleep but not in
waking. Archives Italiennes de Biologie, 122, 2 13-223.
Nelson, J. P., McCarley, R. W., & Hobson, J. A. (1983). REM sleep burst neurons,
PGO waves, and eye movement information. Journal of Neurophysiology, 50,
784-797.
Noda, M., Murakami, J., Yamada, Y., & Aso, T. (1988). Saccadic eye movements
evoked by microstimulation of the fastigial nucleus of macaque monkeys. Journal
ofNeurophysiology, 60, 1036- 1052.
Ogawa, T. (1963). Midbrain reticular influences upon single neurons in lateral genicu-
late nucleus. Science, 198, 855-857.
Ogilvie, R., Hunt, H., Sawicki, C., & Samanhalskyi, J. (1982). Psychological correlates
of spontaneous middle ear muscle activity during sleep. Sleep, 5, 1 1-27.
Peterson, B. W., Franck, J. I., Pitts, N. G., & Daunton, N. G. (1976). Changes in re-
sponses of pontomedullary reticular neurons during repetitive, cutaneous, vestib-
ular, cortical and tectal stimulation. Journal ofheurophysiology, 39, 564-58 1 .
Pivik, R. T. (1978). Tonic states and phasic events in relation to sleep mentation. In
A. M. Arkin, J. Antrobus, & S. Ellman (Eds.), The mind in sleep (pp. 245-276).
Hillsdale, NJ: Erlbaum.
Pompeiano, 0. ( 1970). Mechanisms of sensorimotor integration during sleep. In E.
Stellar & J. M. Sprague (Eds.), Progress in physiologicalpsychology (Vol. 3). New
York: Academic Press.
Pompeiano, 0. ( 1980). Cholinergic activation of reticular and vestibular mechanisms
controlling posture and eye movements. In J. A. Hobson & M. A. B. Brazier
(Eds.), The reticularformation revisited (pp. 473-5 12). New York Raven Press.
Porte, H. S., & Hobson, J. A. (1986). Bizarreness in REM and NREM reports. Sleep
Research, 15, 8 1.
Rechtschaffen, A. (1983). Visual dimensions and correlates of dream images. Sleep
Research, 12, 189.
Rechtschaffen, A., & Foulkes, D. (1965). Effect of visual stimuli on dream content.
Perceptual and Motor Skills, 20, 1 149- 1 160.
Rechtschaffen, A., Watson, R., Wincor, M. Z., Molinari, S., & Barta, S. C. (1972). The
relationship of phasic and tonic periorbital EMG activity to NREM mentation.
Sleep Research, 1, 1 14.
Reinsel, R., Antrobus, J., & Wollman, M. (in press). Bizarreness in waking and sleep
mentation: Waking,REM-NREM, and phasic-tonic differences in bizarreness. In
J. A. Antrobus and M. Bertini (Eds.), The neuropsychology of dreaming sleep.
Hillside, NJ: Erlbaum.
Roharg, H. P., Dement, W., Muzio, J., & Fisher, C. (1962). Dream imagery: Rela-
24 JOHN ANTROBUS

tionship to rapid eye movements of sleep. Archives of General Psychiatry, 7,235-

258.
Sakai, K., & Jouvet, M. ( 1980). Brain stem PGO on cells projecting directly to the cat
dorsal lateral geniculate nucleus. Brain Research, 194, 500-505.
Singer, W. (1977a). Brain stem stimulation and the hypothesis of presynaptic inhibi-
tion in cat lateral geniculate neurons. Brain Research, 61, 55-68.
Singer, W. ( 1977b). The effect of mesencephalic reticular stimulation on intracellular
potentials of cat lateral geniculate neurons. Brain Research, 61, 35-54.
Singer, W., & Bedworth, N. (1973). Inhibitory interaction between X and Y units in
cat lateral geniculate nucleus. Brain Research, 49,49 1-507.
Singer, W., Tretter, F., & Cynader, W. (1975). Organization of cat striate cortex: A
correlation of receptive-field properties with afferent and efferent connections.
Journal of Neurophysiology, 38, 1080- 1098.
Tsumoto, T., & Suzuki, D. A. (1976). Effects of frontal eye field stimulation upon
activities of the lateral geniculate body of the cat. Experimental Brain Research,
25,29 1-306.
Watson, R. K., Bliwise, D. L., Friedman, L., Wax, D., & Rechtschaffen, A. (1978).
Phasic EMG in human sleep: 11. Periorbital potentials and REM mentation. Sleep
Research, 7, 57.
Watson, R., & Deptula, D. (in press). Phasic integrated potentials and ego boundary
deficits. In J. Antrobus & M. Bertini (Eds.), The neuropsychology of dreaming
sleep. Hillsdale, NJ: Erlbaum.
 CHAPTER 2

ACTIVATION, INPUT SOURCE,

AND MODULATION:
A NEUROCOGNITIVE MODEL OF THE
STATE OF THE BRAIN-MIND

J. ALLAN HOBSON

Three factors that are of relevance to the level, source, and mode of informa-
tion processing by the brain-mind state can be quantitatively estimated from
neurophysiological data. They are (a) activation (Factor A), which estimates
the electrical energy level of the information processing system from the rate
of discharge of midbrain reticular formation neurons; (b) input source (Factor
I), which estimates the provenance of data to be processed by the system from
the ratio of external to internal stimulus strength values; and (c) mode of pro-
cessing (Factor M), which estimates the way that data will be treated by the
system by calculating the ratio of cholinergic to aminergic neuronal discharge.
The model is called AIM as both a shorthand eponym combining the
first letters of each of the three factors and as a method of conveyingthe aspira-
tions of a still evolving concept. Since each of the three factors of AIM can be
defined in cognitive terms and can be measured as a neurobiological variable,
the model is both conceptually promising and empirically anchored. The ulti-
mate goal is to unify the brain and mind in a new way. This unity is implied
by the hybrid term bruin-mind,implying a functional identity of the two
domains.
When we mathematically combine Factors A, I, and M in various ways,
both explanatory and predictive principles emerge that render the model heu-
ristically valuable whatever its ultimate verity may be when it is corrected or
elaborated in the light of new data. The key to the success of the concept-
and its major limitation-is its focus on the global states of waking, sleeping,
25
26 J. ALLAN HOBSON

and dreaming and its focus on the very superficial formal aspects of cognition
during those states while ignoring, for the time being, many fascinating cogni-
tive details that also characterize and differentiate those states.

FORMAL ASPECTS OF BRAIN-MIND STATES

The reasons for choosing the state level as a frame of reference for model con-
struction are compelling. From the cognitive point of view, mental activity
has long been known to be differentiated strikingly at this global level of analy-
sis. For example, such major cognitive faculties as perception, volition, and
memory change almost qualitatively as we pass from waking through non-
REM (NREM) or slow wave sleep to REM sleep and dreaming. For example,
consider the difference between waking vision and dream vision: In waking,
visual perception is shaped by external data that have both stable and faithful
representations in our mind’s eye; in dreaming, imagery that is no less clear
and no less vivid arises spontaneously and changes unpredictably, often with-
out reference to even historical reality.
The robust nature of these mental state changes and the relative invari-
ance of their correlation with neurobiological factors suggests that they are
deeply rooted in, if not causally dependent on, the physical state of the brain.
It is the strength of this connection that supports the idea of a unified brain-
mind system. Within the past 35 years, new and powerful neurobiological
techniques, especially in electrophysiology and neuropharmacology, have
yielded abundant, replicable, and richly diversified sets of quantitative data
regarding the states of the brain associated with these robust cognitive state
variables (Hobson, 1988a, b, 1989; Hobson & Steriade, 1986). Thus, in con-
nection with the visual perception example cited previously, we now know
from studies of waking cats and monkeys how retinal signals encoding photic
intensity are processed in the lateral geniculate body and are relayed to the
visual cortex for further elaboration as animals perform visually guided tests.
During REM sleep, many of the same central neural structures are both en-
dogenously activated and autostimulated when humans dream visually de-
tailed scenes. It is this psychophysical parallelism, so clearly differentiated
from state to state, that lies at the heart of AIM. Knowing what we already
know, we ought to be able to begin to sketch a unified model.

PROBLEMATIC PREMISES OF THE MODEL

Subjectivity
Reports by human subjects of their internal mental states, despite wide vari-
ability and notorious unreliability in special cases, have both consensual and
face validity that cannot be denied. The global level of the model accommo-
dates this problem nicely by concentrating on the most unequivocal and uni-
versal aspects of mental states, their formal properties. Thus, if 80 of 100
subjects report visually vivid, constantly animated, emotionally charged, nar-
ratively coherent but bizarrely organized mental experience on experimental
AIM Model 27

awakening from REM sleep, I tend to believe that most of the 80 subjects, and
perhaps even all 100 subjects, actually had such an experience.

Isomorphism
The correlation of these formal mental state variables with brain states as they
can now be measured in humans justifies the assumption of an isomorphic
relation between the consensually validated subjective experiences and the
objectively recordable states of the brain. This correlation may be even
stronger than is now supposed given the uncertainties of accurate reports of
subjective data out of sleep and the relative superficiality of the currently re-
cordable electroencephalogram (EEG) parameters of brain activity. Isornor-
phic is defined as a similarity of form in the two domains of brain and cogni-
tion (McCarley & Hobson, 1977). Thus, for example, if memory is poor for
dreams, we can assume that the biological substrate of memory is impaired.
If, conversely, dream vision is sharp and intense, we can assume that the visual
brain is strongly activated in a way that simulates wake-state activation and
stimulation.

Animal Models
To overcome the problem posed by the inaccessibility of the human brain to
deep neurobiological analysis, we have recourse to animal models whose
brains can be studied at the level of regions, nuclei, neuronal populations,
individual neurons, membranes, and even molecular neurotransmitters (Hob-
son, Lydic, & Baghdoyan, 1986). As payment for the rich data that are thus
obtainable, we accept the uncertain assumption that all mammalian brains,
or at least those of the human, monkey, and cat, share common mechanisms
of state control so that we can build our model up from subhuman data
sources. This strategy makes no assumptions about whether the animals stud-
ied experience consciousness as we humans know it. I will now briefly discuss
the evidence that encourages such assumptions and provides the substantive
ground for the AIM model.

PSYCHOPHYSIOLOGY OF BRAIN-MIND STATES

The brain-mind is periodically activated and deactivated as humans move
from waking through NREM to REM sleep (see Figure 1). We further know
that the electrical aspects of such activation are shared by man and cat (see
Figure 2). Assumptions underlying the quantification of activation (Factor A
in the model) are thus quite strongly justified. Indeed, the activation concept
is widely accepted and applied by cognitive psychologists, usually in models
that are considerably more specific and more limited than is AIM (Anderson,
1985; Chapter 1, by J. Antrobus in this volume). Thus, the truly innovative
aspects of the new model are related to Factor I (input source) and Factor M
(mode of processing), which have not yet been considered by model-building
cognitive psychologists. (The reader seeking a more detailed discussion of Fac-
tor A should consult Hobson, in press.)
28 J. ALLAN HOBSON

Figure 1 Three brain-mind states in humans. (States ofwaking, non-REM [NREM]

sleep, and REM sleep have behavioral, polygraphic, and psychological manifestations.
In the behavior channel, posture shifts (detectable by time-lapse photography or video)
can occur during waking and in concert with phase changes of the sleep cycle. The
sequence of these stages is represented in polygraph channel displaying EMG, EEG,
and EOG tracings. Three lower channels describe other subjective and objective state
variables.)

The source of data, or input source, changes progressively from a

strongly external stimulus drive (SE) in waking to a strongly internal stimulus
drive (SI) in REM sleep. Parallel to the decrease in activation that characterizes
the wake-NREM transition is thus a decline in the access of external data to
the system. That this decrease in external data is not a linear function of the
decrease in activation is shown by the exponential increase in threshold to
external stimulation at sleep onset. This exponential change in input source
probably depends on the sudden occurrence of thalamocortical oscillation
(seen in the EEG as spindling) that interferes with the effective transfer and
central processing of sensory data. Since the strength of the external data stim-
ulus is the reciprocal of the threshold, a value for SE can be derived by comput-
ing the inverse of threshold. We can therefore assert, as shown by Bonnet’s
studies (Chapter 1 1, by M. H. Bonnet in this volume), that values for the exter-
nal data function could be directly derived from human data.
AIM Model 29

Brain

Muscle

Brain

Figure 2 A comparison of REM sleep electrographic features in man and cat, based
on a time-lapse photographic study by Theodore Spagna.

As the NREM sleep state evolves toward REM sleep in the cat (see Figure
3), we notice an increase in SI with the evolution of the pontine-geniculate-
occipital (PGO) waves (Callaway, Lydic, Baghdoyan, & Hobson, 1987). The
increase in SI parallels the resurgence of activation, but the cross-correlation
is no more linear than the relation of activation to SE at sleep onset; rather, it
is similarly and strikingly exponential. It is also significant that the thalamo-
cortical oscillation of NREM sleep subsides inversely with the increment in
PGO wave frequency. At a later and critical point in the transition from
NREM to REM, the PGO waves are no longer the isolated, large-amplitude
single spikes and wave complexes but become clustered into groups of 6-10
smaller waves (see Figure 4).That the PGO waves are genuinely sensory stim-
uli of entirely internal provenance is made clear by the discovery that during
REM sleep eye movement direction is encoded in the left-to-right amplitude
difference of the waves in the two lateral geniculate bodies and occipital corti-
ces (Callaway et al., 1987; Monaco, Baghdoyan, Nelson, & Hobson, 1984).
This discovery further justifies the assumption that these internally generated
sensory stimuli actually arise on the motor side in keeping with two important
concepts: First, that motor commands arising in the upper brain are abundant
in REM sleep but are quenched by inhibition of the anterior horn cells in the
30 J. ALLAN HOBSON

c-(

Figure 3 Transition from non-REM (NREM) to REM sleep in the cat.

spinal cord and, second, that REM sleep is characterized by a marked increase
in the excitability of internal communication systems of the brain.
We are thus justified in postulating that in REM sleep there is not only
the diminished access to external data by the brain-mind that is typical of all
sleep but also an increase in the strength of internally generated data. It seems
likely, moreover, that these PGO waves are a reflection of the disinhibition of
the efferent copy or corollary discharge systems of the brain by which sensori-
motor integration has been conceptualized since the time of the German neu-
rophysiologist, Hermann von Helmholtz (for a discussion, see Hobson,
1988a). Although PGO waves are not easily recorded from human subjects,
recent evidence supports the assumption that they do exist in our species and
that they also signal the direction of eye movement from brainstem motor
centers to the cerebral cortex (McCarley & Ito, 1983; McCarley, Winkelman,
&Durn, 1983).
In the shift from waking through NREM sleep to REM, the brain-mind
has thus been deactivated (Factor A) with a concurrent decrease in the access
of external data to the system and then reactivated with a concurrent increase
in internal stimulus strength. We can therefore say with confidence that the
cognitive differences between waking and REM sleep could not possibly de-
pend only (if at all) on a changed level of activation. Rather, it is clear that
both the waking and REM sleep states differ from NREM sleep and that they
differ in equal degrees on this dimension. This difference parallels the cogni-
tive downshift in NREM sleep that was pointed out by Antrobus (Chapter 1,
by J. Antrobus in this volume). Input source, however, has shifted radically
from external (in waking) to internal (in REM sleep), whereas NREM sleep
 ----
AIM Model 31

EOG-

Figure 4 Pontine-geniculate-occipital(PGO)waves and relation to REM sleep and

eye movement direction. (A) NREM-REM transition showing PGO waves. (B) Side-to-
side alternation of primary waves. (C) Burst cell and PGO waves. (D) Efferent copy
model. Subscripts:C = contralateral,I = ipsilateral.

shows constantly changing intermediate values of input source. Thus, differ-

ent data are being processed by the equally activated brain in the two states.
Now the question becomes, Is the data being processed in the same or in
different ways?
Factor M, estimating the mode of processing, is designed to capture and
to estimate the strength of such shifts in information processing from waking
to dreaming, such as the shifts from linear to parallel, from logical to analogic,
and from analytic to synthetic. The most likely candidate for a neurobiological
mediator of these mode shifts is the modulatory neuronal systems of the brain
whose generic name is suggestive of such a function (Laborit 1986).
32 J. ALLAN HOBSON

I -
0 Se/ectiviv Gradient

C lnterpenetrotion Grodient

10 I ( F T G ) 3 I (DRN)

OExcitotory *Inhibitory
Figure 5 Schematic illustration of the hypothesized relation between physiological
selectivity and anatomical interpenetration of REM-on (left column) and REM-off
(right column) cells whose reciprocal interconnections are modeled in Factor A.

We know that many modulatory neuronal populations of the brainstem

show progressive changes in tonic discharge rate, and in the release of their
respective neurotransmitters, over the sleep cycle in the cat. According to this
concept, the extracellular concentration of neuromodulators over time is a
function of the tonic discharge rate of the neurons that manufacture and re-
lease them. As yet, we have no evidence that such changes occur in humans,
which makes Factor M the weakest link in the model. This is unfortunate
because it is in many ways the most conceptually interesting aspect of the
model. We must therefore regard its introduction as speculative and must
await direct proof that the human brain is biochemically altered in the same
way and to the same degree as that of the cat (see Figures 5 and 6).
With this caveat in mind, we may tentatively state that the value of Fac-
tor M decreases from a high ratio of aminergic to cholinergic neurotransmitter
concentration in the brain during waking to a low ratio in REM sleep. In
other words, the brain-mind becomes progressively less aminergic and more
cholinergic as we pass from waking to dreaming (see Figure 7).
In this view, information processing transactions take place in a radically
different biochemical climate according to whether the brain-mind is awake
AIM Model 33

BRAIN NEURONAL SYSTEM

LEVEL [ Aminergic I Reticular I Sensorimotor
I 1 I IEEG Desynchmimtim I
Cortex I
'PT
:ells I
I I
I
I I
I I lPG0 Waves
I I
Thalamus I I
I I
I I
I I

I
I
Midbrain
I
I

'
I
RF
Pons I LC
II p
I
Medulla I
I
I
I
I
I Atmia andrtlusce Twitches I
I
Spinal I
Cord I
I-----I- ---- l - - - l H
0 Inhibitory OExcitatory 5%

Figure 6 Schematic representation of the REM sleep generation process.

or is in REM sleep. This concept is an intracerebral and informational update

of Hess's ( 1954) distinction between the ergotrophic (or catabolic) functions
of the sympathetic nervous system and the trophotropic (or anabolic) func-
tions of the cholinergicnervous system. The shift from an emphasis on energy
(in Hess's concept) to an emphasis on information processing (as in the AIM
model) is prompted by the recognition that such neurotransmittersas acetyl-
choline, norepinephrine,and serotonin exert their modulatory actions via sec-
ond messengers, with effects on the activity of the intracellular metabolism of
the genome as well as on the conductance of neuronal membranes (Black et
al., 1987; Hobson, 1988b). It is now widely accepted that changes in mem-
brane conductance that constitute signals from cell to cell must lead to alter-
ations in the protein structure of ion channels, receptors, or enzymes if neu-
rons are to retain a trace (or memory) of their excitation. For example,
whether a postsynaptic cell remembers the information conveyed by its cho-
 100 - -

Ret N

*
4

@
1-
LC

RN

PN

01 I I
w S D W S D
Behoviord Sfafe
(4 (B)

b Aminergic

A/C Ratio

I I I
Waking NREM REM
Sleep Sleep
(C)
Figure 7 Behavioral state and neuronal discharge rate. (A) D-on cells. (B) D-off
cells. (C) State-dependentchanges in aminergic to cholinergic ratio.
AIM Model 35

linergically mediated input may depend on whether it is coactivated by an

aminergic neuron releasing dopamine, serotonin, or norepinephrine (Flicker,
McCarley, & Hobson, 1981).
In any case, we know that the noradrenergic and serotonergic neuronal
populations of the brainstem shift from high levels of discharge in waking to
intermediate levels in NREM sleep and to very low levels in REM sleep. This
means that the brain is aminergically deafferented or “demodulated” in REM
sleep compared with waking. (See Mamelak & Hobson, 1989, for a discussion
of this concept and its implications for cognition.)

NEUROBlOLOGlCAL QUANTIFICATION OF AIM

The activation values of several neuronal populations are shown in the left
panel of Figure 8. Although the absolute values vary widely from neuron to
neuron, it is generally true that in NREM sleep the average level is about half
that in waking and REM sleep. We may thus use the ratio 2: 1:2 as the relative
value of activation across the three states.
The estimate of input source is more complex because SE and SI vary
conversely, with the former value falling rapidly at NREM sleep onset and the
latter rising rapidly at REM sleep onset. Using Bonnet’s data (Chapter 1 1, by
M. H. Bonnet in this volume), we note that since the auditory threshold
changes from 10 dB in waking to 40 dB in NREM sleep, SE suddenly falls by
a factor of about 4: 1 at sleep onset. We do not know what happens in REM
sleep, but the data of Harsh and Badia (Chapter 4,by J. Harsh and P. Badia
in this volume) indicate that the sensory threshold actually falls a bit as REM
develops, perhaps by a factor of 2: 1. Thus, we may estimate SE in REM as 2.
If PGO spike frequency is used as the estimate of internal stimulus strength,
SI has the relative values of 1:2: 100 (based on absolute values of 0.1 c/s for
waking, 0.2 c/s for NREM, and 10.0 c/s for REM). If input source is the ratio
of SE/SI, solving for input source gives the following values: wake = 4.00,
NREM = 0.50, and REM = 0.02.
Factor M is a ratio function that measures the relative strengths of the
aminergic and cholinergic systems. From the neuronal data in the right panel
of Figure 8, we obtain values for aminergic modulation (MA) of 2.00 in wak-
ing, 1.OO in NREM sleep, and 0.0 1 in REM sleep. The values for cholinergic
modulation (MC) are not precisely estimable at this time but may be conserva-
tively quantified as being proportional to Factor A: Thus, MC has values of 2
for waking, 1 for NREM, and 2 for REM. The equation M = MA/MC yields
M values of 1.OO for waking, 1.OO for NREM, and .005for REM.

COMBINING THE VALUES OF AIM

In a previous paper, the merits of various arithmetic and graphic representa-
tions of AIM were explored (Hobson, in press). The numerical value of A X
I X M was first plotted as a function of sleep cycle phase, which revealed inter-
esting properties at sleep onset (Chapter 1 1, M. H. Bonnet in this volume) and
during REM sleep (Chapter 4,by J. Harsh and P. Badia in this volume). Then
36 J. ALLAN HOBSON

Wake
100

10 -

S@
1-

0.1-

Figure 8 At sleep onset, a sudden and precipitous drop in the value of S could be
caused by the sudden drop in the value of SE (with eye closure) after a slow decline in
activation.

a three-dimensional state-space representation was explored (see Figure 9).

When each variable is one of three dimensions of the state space, the instanta-
neous positions of points with coordinates A, I, and M represent the moment-
to-moment value of the brain-mind state. The clusters or “clouds” of instan-
taneous values of AIM for waking are seen to fall in the back, upper right
comer, in the center of the space for NREM, and in the front, lower right
comer for REM. Of course, these positions are arbitrary, but their interre-
lations are revealing.
One virtue of the state-space conceptualization of AIM is that it can also
accommodate a much wider continuity of values and a more richly varied
continuity of states than can be integrated by any other existing schema. To
demonstrate this property, we can begin to contemplate and comprehend such
otherwise puzzling dissociative phenomena as lucid dreaming (Chapter 8, by
S . LaBerge in this volume) and reverie (Chapter 12, by D. F. Dinges in this
volume).

AIM IN LUCID DREAMING

In lucid dreaming, the paradox to be explained is the recovery of self-reflective
awareness by subjects who report becoming conscious of dreaming and who
can then signal out of REM sleep, when the somatic musculature is paralyzed,
by executing complex voluntary eye movement sequences. The other physio-
AIM Model 37

State Space Waking

NREM REM Sleep

Figure 9 Three dimensional state space defined by the values for brain activation
(A), input source and strength (I), and mode of processing (M).

logical signs of REM sleep are still present. It has been difficult for many sleep
scientists to understand what physiological mechanisms could allow such
wake state motor behavior as voluntary movement to occur during REM sleep
because, by definition, motor output is blocked in REM. The AIM model
allows us to consider another possibility: that lucid dreaming is a unique and
specific state of the brain-mind. Because it has some features of waking and
some features of REM, it is a hybrid state. Where would lucid dreaming fall
in the state space?
It seems unlikely that a change in activation could account for the results
because the level of cognitive activation is, if anything, higher than in REM
sleep or in waking. Rather, one suspects a change in input source, with a tem-
porary suppression of internally generated signals, allowing cortical com-
mands to dominate the oculomotor apparatus in a wake-state fashion. Put
another way, the usually weak top-down forces gain control of the usually
strongbottom-up forces that are battling for control of the brain-mind in REM
sleep. We might imagine Factor I as increasing toward waking values (ie.,
38 J. ALLAN HOBSON

toward the back ofthe state space, though still keeping to the right side because
activation remains high). Since the REM-PGO generator network in the brain
stem is activated via aminergic disinhibition, one way that such an effect could
be enhanced is by turning on the locus coeruleus and raising the value of MA,
which would put a brake on the cholinergic REM sleep generator. This would
have the effect of moving AIM upward in the state space. The net effect of
raising Factors I and M toward waking levels, while conserving a high level of
Factor A, would be to locate lucid dreaming on the right-hand wall of the state
space in the midpath of the rapid reset trajectory of the system when REM
suddenly gives way to waking at the end of each REM cycle.
This view predicts (a) that lucid dreamers will frequently awaken from
REM sleep once dream consciousness is achieved and (b) that lucidity will be
easiest to induce at times in the night when the system is likely to be changing
from REM to waking. This might be true during the brief period of descending
to Stage 1 episodes at sleep onset but will be even more likely at the end of
REM periods late in the night. This hypothesis could be tested in longitudinal
home-based studies of experienced lucid dreamers.

REVERIE
Dramatic state dissociations occur on arousal from NREM sleep, especially
in sleep-deprived subjects (Chapter 12, by D. F. Dinges in this volume). Al-
though they are electrically awake (that is, showing signs of high levels of acti-
vation) and capable of input-output processing of the wake state type (that is,
also having high levels of input source), such individuals appear to be unable
to keep a cognitive test task in mind, and hence they perform poorly. They
may even confabulate and produce garbled language only loosely rooted in
the task structure imposed by the experimenter. Thus, we are again dealing
with a paradox: Features of wake are hybridized with features of NREM sleep.
Variations of this functionally significant anomaly occur naturally in
children with night terrors and in adults (such as physicians and other night-
shift workers) who find themselves unable to perform complex cognitive tasks
when suddenly aroused from NREM sleep by duty calls. Broughton (1 968)
has dubbed these phenomena disorders of arousal. Dinges (Chapter 12, by D.
F. Dinges in this volume) has postulated that “sleep inertia” may mediate this
phenomenon because, despite their best intentions, the subjects feel irresist-
ibly pulled back down into sleep.
What could this inertial element of the brain-mind be, according to
AIM? Factors A and I are probably not the culprits because the EEG is fully
activated and wakelike input-output processing is possible. It seems rather that
it is attention, the ability to hold a concept or task in mind and focus on
it, that is defective. This would result if Factor M remained effectively low,
so that the cortex, while electrically activated and capable of input-output
transactions, remained biochemically demodulated. How could such an
event occur?
Suppose that, early in the night, the serotonergic and noradrenergic sys-
tems of the brain have been operating at half-speed for 1 to 3 hours. And
AIM Model 39

suppose further that the concentration of these two neurotransmitters in the

cortex has fallen to 50%of waking levels. An arousal signal might immediately
turn on the noradrenergic and serotonergic neurons electrically, but the con-
centrations of serotonin and norepinephrine in the extracellular fluid of the
forebrain would climb back to their wake state levels only gradually. Ten to
fifteen minutes of sustained firing might be needed to raise the concentration
of these modulatory compounds to levels consonant with sustained attention
and with memory.
In this view, sleep inertia is a hydraulic process, like the time taken for a
large reservoir to refill when a small valve is opened, no matter how high the
pressure in the supply chamber. One of the virtues of AIM is thus its distinc-
tion between electrical phenomena or “sparks” (mediating rapid changes in
Factors A and I) and biochemical phenomena or “soup” (mediating slower
changes in Factor M). Future development of this aspect of the model must
take these even slower circadian and hormonal soup-like processes into ac-
count. In reverie, the net effect on AIM of raising Factors A and I while keep-
ing Factor M low would be to move AIM to the back, lower right-hand comer
of the state space. A testable prediction of this formulation is that treatment
of subjects with biogenic amine-reuptake blockers will protect against reverie
by maintaining higher levels of Factor M throughout sleep.

CONCLUSlON
On the assumption that the states of waking and sleep effect cognition via
neurobiological mechanisms, a three-factor model is proposed. Reflecting the
electrical energy level of the brain is Factor A (for activation), which has a
well-establishedcognitive analogue in network modeling parlance. The prove-
nance of the information to be processed is assessed by Factor I (for input
source), which measures the relative access and strength of external versus
internal stimuli or data. Factor I also relates to motor activity via the corollary
discharge or efferent copy concept. The most speculative and original compo-
nent of the model is Factor M (for modulation), the estimated ratio of aminer-
gic to cholinergic neurotransmitter concentration, which measures the mode
of information processing manifested by the brain-mind. By mediating atten-
tion, learning, and memory processes, Factor M alters the capacity of the acti-
vated brain-mind to focus and perform intentional analytic computations of
its data.

References
Anderson, J. (1985). Cognitive psychology and its implications (2nd ed.). New York:
Freeman.
Black, I. B., Adler, J. E., Dreyfus, C. F., Friedman, W. F., LaGamma, E. F., & Roach,
A. H. (1987). Biochemistry ofinformation storage in the nervous system. Science,
236, 1263-1268.
Broughton, R. (1968). Sleep disorders: Disorders of arousal. Science, 159, 1070-1078.
40 J. ALLAN HOBSON

Callaway, C. W., Lydic, R., Baghdoyan, H. A., & Hobson, J. A. (1987). Ponto-genicu-
lo-occipital waves: Spontaneous visual system activation occurring in REM sleep.
Cellular and Molecular Neurobiology, 7, 105- 149.
Flicker, C., McCarley, R. W., & Hobson, J. A. (198 1). Aminergic neurons: State con-
trol and plasticity in three model systems. Cellular and Molecular Neurobiology,
I, 123-166.
Hess, W. R. (1954). The diencephalic sleep centre. In E. D. Adrian, F. Bremer, &
H. H. Jasper (Eds.), Brain mechanisms and consciousness (pp. 1 17-1 36). Oxford,
England: Blackwell.
Hobson, J. A. (1988a). The dreaming brain. New York: Basic Books.
Hobson, J. A. ( 1988b). Homeostasis and heteroplasticity: Functional significance of
behavioral state sequences. In R. Lydic (Ed.), Clinical physiology of sleep (pp.
199-220). Bethesda, MD: American Physiological Society.
Hobson, J. A. (1989). Sleep. San Francisco: Scientific Amencan Library.
Hobson, J. A. (in press). A new model of brain-mind state: Activation level, input
source and mode of processing (AIM). In J. Antrobus & M. Bertini (Eds.), The
mind-brain in dreamingsleep: Activation andasymmetry. Hillsdale, NJ: Erlbaum.
Hobson, J. A., Lydic, R.,& Baghdoyan, H. A. (1986). Evolving concepts of sleep cycle
generation: From brain centers to neuronal populations. Behavioral and Brain
Sciences, 9, 37 1-448.
Hobson, J. A., & Steriade, M. (1986). Neuronal basis of behavioral state control. In V.
Mountcastle & F. E. Bloom (Eds.), Handbook ofphysiology: The nervous system
(Vol. 4, pp. 70 1-823). Bethesda, MD. American Physiological Society.
Laborit, H. (1986). L’Inhibition de l’action [The inhibition of action] (2nd ed). Mon-
treal, Canada: Masson.
Mamelak, A. N., & Hobson, J. A. (1989). Dream bizarreness as the cognitive correlate
of altered neuronal behavior in REM sleep. Journal of Cognitive Neuroscience, 1,
201-222.
McCarley, R. W., & Hobson, J. A. (1977). The neurobiological origins of psychoana-
lytic theory. American Journal of Psychiatry, 134, 121 1- 1221.
McCarley, R. W., & Ito, K. (1983). Intracellular evidence linking pontine reticular
formation neurons to PGO wave generation. Brain Research, 280, 343-348.
McCarley, R. W., Winkelman, J. W., & D u e , F. H. (1983). Human cerebral potentials
associated with REM sleep rapid eye movement: Links to PGO waves and waking
potentials. Brain Research, 274, 359-364.
Monaco, A. P., Baghdoyan, H. A., Nelson, J. P., & Hobson, J. A. (1984). Cortical PGO
wave amplitude and eye movement direction are correlated in REM sleep but not
in waking. Archives of Italian Biology, 122, 2 13-223.
 CHAPTER 3

EVENT-RELATED BRAIN POTENTIAL

(ERP) STUDIES OF C O G N I T I O N
D U R I N G SLEEP:
IS IT MORE THAN A DREAM?

MARTA KUTAS

Punctate stimuli presented to an awake human leave an electrical trail as they

travel from sensory transducers through the central nervous system (CNS).
This electrical signature, typically represented as a waveform of voltage in
time, is known as the evoked response or potential (EP). Current-day dogma
has it that the EP reflects the potential field created from the summation of
the excitatory and inhibitory postsynaptic potentials synchronously activated
by an incoming stimulus. With the discovery of new brain wave components,
it has become clear that some waves are more intimately related to internal
events (such as preparing to move) than to external stimuli. Hence, the term
evoked was subsumed by the general term event-related (EP being expanded
to ERP). The transient ERP comprises a series of negative and positive peaks
identified by their time synchrony to an eliciting stimulus. Each peak (also
referred to as a component) is identified by its polarity and latency relative to
stimulus onset (although on occasion a psychological label finds itself un-
abashedly attached to an ERP component). An ERP component is further
characterized by the relative distribution of its amplitude across the scalp and
by its waveshape (i.e., morphology).
Over the past 20 years, much research has been aimed at discoveringthe
specific functions that relate changes in these EP characteristics to variations
in stimulus parameters such as intensity, rate of presentation, frequency, and
modality (see Regan, 1972, 1989). More important for our present purpose,
much energy has also been spent compiling a vocabulary of components that
43
44 MARTA KUTAS

describe many of the critical information-processing transactions in the brain

that underlie cognition. A fuzzy but sizeable vocabulary exists. Thus, for ex-
ample, some component changes have been related to selective attention (e.g.,
N 1, Nd) and others have been related to anticipation and preparation (contin-
gent negative variation [CNV] and readiness potential [RP]). Still others have
been linked to sensory mismatches (N2-P3a), orienting (N2), preperceptual
mismatches (mismatch negativity [MMN]), surprise (P3b), novelty (P3a), and
semantic processing (N400) (Gaillard & Ritter, 1983; Hillyard & Picton, 1987;
Johnson, 1988; Johnson, Rohrbaugh, & Parasuraman, 1987; Kutas & Van
Petten, 1988).
One undeniable benefit of some ERP components as measures of hu-
man brain activity is that they can be recorded regardless of a person’s level
of arousal or mental state. Of course, this is not to say that a person’s state
may not alter the amplitude or latency of the electrical response recorded but
rather that a subject need not be awake or fully conscious ofthe elicitingstimu-
lus to generate a measurable response. Insofar as we know what process is
indexed by a part of an ERP waveform, the differential behavior of the peaks
and troughs of the ERP can be used to make inferences about the psychology
and physiology of the brain, be it awake or asleep. However, insofar as the
processes underlying the ERP are less clear, it may be necessary to vary arousal
states to learn more about the mappings between ERP components and cogni-
tive operations.

EVENT-RELATED POTENTIALS IN SLEEP

Given that electroencephalographic(EEG) criteria have proven indispensable
for differentiating among the various sleep stages (e.g. Rechtschaffen & Kales,
1968), it is surprising how infrequently ERP measures have been chosen as a
means of evaluating the psychological and computational capabilities of the
brain during the different sleep stages. In part, this may reflect our intuitions
and the prevailing belief that not much processing (especially of external stim-
uli) goes on during sleep. Then again, the lack of cognitive ERP studies during
sleep may be a reaction to the difficulty of pinning a cognitive operation
uniquely to a specific positive or negative wave, even under conditions in
which subjects are unquestionably awake and behaving. If one is fairly certain
that there is little in the way of cognition during sleep and one is equally uncer-
tain about interpreting a measure of cognitive activity (within existing frame-
works), then one would be a bit foolhardy to jump right into the research fray.
Few have.
Indeed, most research utilizing ERP recordings during sleep has focused
on clinical rather than cognitive uses of the technique. By this view, the ERP

I extend sincere thanks to C. Van Petten for her helpful comments on this manuscript.
Research efforts were supported by a Research Scientist Development Award from the National
Institute of Mental Health (MH00322) and by a grant from the National Institute ofchild Health
and Human Development (HD226 14).
ERP Studies 45

is a clinical tool that provides an index of the integrity of the sensory pathways
in the central nervous system. For diagnostic purposes, the most informative
EP components are those that are consistently elicited by specific stimulus
parameters in the face of fluctuations in arousal or attention. Studies over the
past 20 years have indicated that the auditory brain evoked potential (ABEP),
also known as the auditory brain stem response (ABR) or the brainstem
evoked response (BER), providesjust such a response. The ABR consists of a
series of small vertex-positive waves occurring within the first 10 ms after an
abrupt stimulus (Campbell, Picton, Wolfe, Baribeau-Braun, & Braun, 1981;
Chiappa, Gladstone, & Young 1979; Stockard, Stockard, & Sharbrough,
1978). Peak 5 of the response is usually recognizable to within 20 dB of the
behavioral auditory threshold (Galambos & Hecox, 1978; Picton, Woods,
Baribeau-Braun, & Healy, 1977; Picton, Stapells, & Campbell, 1981). While
there is considerable intersubject variability in component amplitudes, the la-
tencies of the BER peaks are remarkably stable across replications, provided
that the subject’s age, gender, and body temperature as well as the physical
properties of the eliciting stimulus and recording parameters are held constant
(e.g., Moore, 1983). Such reliability, in fact, has sanctioned the development
of a normative data base against which deviant responses can be judged. Ab-
normalities in the timing of various components relative to norms signal a
possible cause for concern in either audiological or neurological domains.
With a few exceptions that have been difficult to replicate (Lukas, 1980, 198l),
it has been demonstrated that the first five waves of the BER are insensitive to
attentional manipulations in an awake adult (Picton & Hillyard, 1974; Wol-
do@ Hansen, & Hillyard, 1987; Woods & Hillyard, 1978). Thus, ABR mea-
surement has become a routine means of objective audiometry, especially for
subjects who are unwilling or unable to undergo traditional methods of exami-
nation (e.g., neonates, young children, demented individuals).
The insensitivity of the BER to arousal levels and attentiveness in awake
subjects suggests that the BER should be the same regardless of whether the
subject is awake or asleep. The literature on this question is, however, contra-
dictory. For instance, Amadeo and Shagass ( 1973) reported latency shifts as
large as 0.25 ms in different stages of sleep. In contrast, Hellekson, Allen, Gree-
ley, Emery, and Reeves (1979) found no differences in the latency of any ERP
components between waking and light naps. Osterhammel, Shallop, and Ter-
kildsen (1985) noted only minor changes in the latency of Peak 5 for only
some subjects and only for low intensity stimuli. Campbell and Bartoli ( 1986)
suggested that such inconsistencies in the literature may be due to a number
of factors including (a) failure to control the stimulus input (that is, using loud-
speakers rather than headphones), (b) failure to control for covariation in
core temperature during the night (Jones, Stockard, & Weidner, 1980; Mar-
shall & Donchin, 198l; Stockard, Sharbrough, & Tinker, 1978), and (c) small
sample size.
Campbell and Bartoli (1986) evaluated the BER at different intensities
and rates of stimulus presentation during wakefulness, Stages 2 and 4, and
46 MARTA KUTAS

REM sleep while maintaining precise control over the stimulus input (with an
ear-mold hearing aid device) and monitoring temperature in 9 female subjects
( 18-25 years old). They reported that regardless of (a) the stage of sleep, (b)
the time of night, (c) the rate of stimulus presentation, and (d) the intensity of
the stimulus, sleep did not appear to have a significant effect on any of the
components of the click-evoked ABR. They concluded “that auditory infor-
mation is transmitted without alteration from the periphery through the brain-
stem relay centres during sleep” (p. 146). Before we accept this conclusion, we
must remember that the BER is a sensitive index of the timing operations of
the auditory system but is silent about the processing of the contents of the
auditory signal.
Components of the auditory evoked response occumng between 10 and
100 ms have also been called into service for clinical hearing assessments.
These components, referred to as middle latency responses (MLR), come in
the form of transient or steady-stateresponses. Transient responses are evoked
whenever stimuli are presented at slow rates (lO/s or less), whereas steady-
state responses arise when stimuli are presented at such a high repetition rate
that the responses to successive stimuli overlap (e.g., Stapells, Linden, Suffield,
Hamel, & Picton, 1984). The auditory MLR has received considerable atten-
tion in audiometry because of the possibility that it might allow determination
of frequency-specific auditory thresholds. And indeed, threshold estimates
from MLRs correlate well with behavioral thresholds, with the estimates from
the steady-state response being less variable than those derived from transient
responses (Galambos, Makeig, & Talmachoff, 1981; Goldstein & Rodman,
1967; Klein, 1983; Mendel et al., 1975; Shallop & Osterhammel, 1983; Sta-
pells, 1984). Moreover, the effects of stimulus rate on the amplitude of the
steady-state MLR are quite similar during wakefulness and sleep. For exam-
ple, although the amplitude of the response is smaller during sleep, the maxi-
mum amplitude is still recorded at stimulus rates between 30-50 tones/s. The
amplitude of the response increases as the intensity of the stimulus increases.
A similar stimulus-intensity/response-amplitudefunction occurs at all stages
of sleep, although the slope of the suprathreshold intensity-amplitude relation
is lower during sleep than waking. Likewise, a similar intensity phase (i.e.,
latency) relation occurs in all stages of sleep and wakefulness, with the phase
of the response decreasing as the suprathreshold intensity of the stimulus in-
creases. Thus, while the steady-stateMLR is sensitive to changes between sleep
and wakefulness, it is little affected by transitions across the different sleep
stages (e.g., Linden, Campbell, Hamel, & Picton, 1985). As such, it holds
promise as a viable technique for generating electrophysiological audiograms
at different frequencies during sleep.

COGNITIVE EVENT-RELATED POTENTIALS

There are few reports comparing components of the auditory ERPs later than
100 ms to standardized stimuli in the different states of wakefulness and sleep.
Although during sleep there seems to be an overall reduction in the amplitude
ERP Studies 47

of electrical responses to sensory stimulation, this is not true of every compo-

nent. Thus, it is important to consider different time regions of the ERP sepa-
rately. For example, the amplitude of the N1 (around 100 ms) component is
usually reduced during sleep (Anch, 1977; Bell & Campbell, 1988), whereas
the amplitude of the following P2 component has been found to be highly
variable (Buchsbaum, Gillin, & Pfefferbaum, 1975). The most profound
change in the longer latency ERP components is the addition of a large nega-
tive N2 wave with a peak latency of 300-500 ms occasionally followed by a
P3 component around 800 ms (e.g., Kevanishvili & von Specht, 1979;Ornitz,
Ritvo, Carr, Panman, & Walter, 1967;Picton, Hillyard, Krausz, & Galambos,
1974; Ujszaszi & Halasz, 1986; Weitzman & Kremen, 1965;Williams, Tepas,
& Morlock, 1962).' Components of the auditory ERP later than 100 ms have
been found to be too variable to be useful in the clinic.
These data clearly demonstrate that sensory stimuli are afforded some
processing during sleep, often with stimulus-response functions, that resem-
ble those of wakefulness. These observations, together with the data demon-
strating that people are capable of making both simple and discriminative
manual responses during sleep (Granda & Hammack, 1961;Williams, Mor-
lock, & Morlock, 1966), suggest the possibility of applying cognitive ERP par-
adigms. Very few investigations of this type have been reported. Those that
have have focused on ERPs associated with anticipation or with physical mis-
matches embedded in a stimulus sequence.
During wakefulness, to be forewarned to is to be forearmed. That is,
being warned of an impending stimulus makes it possible to react more
quickly and more accurately. In most theories, the facilitatory influence of the
preparatory set on motor reactions and perceptual judgment is attributed to
different subprocesses of attention. Does this adage hold for sleep as well? Is it
possible to talk about different types of attention during sleep? In the waking
adult, differential preparation is reflected in specific patterns of CNS activity
that develop across preparatory intervals. For example, in both human and
nonhuman animals it has been found that, during intervals of preparation and
anticipation, there develops an expectancy wave; this potential is referred to
as the contingent negative variation (CNV), contingent because it is assumed
to index the contingency between the warning and imperative stimulus and
variation because that is the way the British say potential.
Walter, Cooper, Aldridge, McCallum, & Winter ( 1964) first observed
the CNV in a reaction time task with a fixed foreperiod between the warning
stimulus and the imperative stimulus. Accordingly, they proposed that this
cortical electronegativity reflected the priming of the frontal lobes for action.
Current-day views of the CNV based on human and animal experiments also
have proposed major involvement of the frontal lobes as well as the ascending
reticular activating system (ARAS) in CNV generation and modulation (Skin-

' This N2 and P3 should not be mistaken for the N2 and P3 components that are referred
to in the ERP studies of human information processing in awake adults.
48 MARTA KUTAS

ner & Yingling, 1977). Moreover, Marczynski (1978) argued that it is the cho-
linergic component of the ARAS that is involved in the genesis of surface
negative potentials of this type.
Studies demonstrating that CNVs can be elicited without a motor re-
sponse requirement have shifted the emphasis from response priming to mobi-
lization of effort for anticipated activity, wherein the amplitude of the CNV
indexes the total allowance of effort devoted to preparatory activity (for re-
views see Hillyard, 1973; Tecce, 1972). Most investigationsthus have empha-
sized the relation between the CNV and intensive (e.g., alertness, arousal, con-
centration, effort, expectancy) rather than selective aspects of attention. And
to some extent this seems reasonable. Generally, attentiveness to the eliciting
stimuli yields larger CNVs than distraction, more difficult tasks yield larger
CNVs than easy tasks, and anticipation of noxious or informative stimuli
yields larger CNVs than benign or uninformative stimuli (for a review, see
Donchin, Ritter, & McCallum, 1978). However, the relation is by no means
a simple one: Too difficult a task, too stressful a task, or too many simulta-
neous task requirements often yield a CNV of a diminished amplitude. De-
spite this complex function between CNV amplitude, attentional resources,
and arousal (e.g., Tecce & Hamilton, 1973), it is possible to determine the
relation under specified circumstances during waking and to contrast it with
that obtained under the same conditions during sleep.
One such study was conducted by Salamy, Lester, and Jones (1975). In
this experiment, 3 subjects were asked to respond (i.e., press a microswitch
taped to their right hand) to white noise forewarned by a tone occumng 1 s
ahead on 3 consecutive nights. Recordingswere obtained prior to sleep, during
REM sleep, and immediately after morning awakening. The results suggested
that, although the subjects responded correctly 65% of the time in REM sleep,
a CNV either did not develop or was dramatically attenuated during REM
sleep. However, there are aspects of the data that warrant caution. For in-
stance, it would be informative in interpreting the ERPs to compare the wave-
forms associated with a response against those elicited on no-response trials.
Given the limited number of trials per average (n = 16), however, it is unlikely
that the data could have been sorted on the basis of behavior (e.g., by reaction
time). The noise level of the data presented, especially in the REM condition,
indicates that caution is well-advised. Moreover, as the authors noted but dis-
missed, the reaction times during REM were significantly slower (five- to six-
fold) than when the subjects were awake. While the relation between CNV
amplitude and reaction time is typically not linear, very prolonged reaction
times tend to be associated with very small CNVs. Finally, the recordings were
restricted to two electrode locations, vertex (Cz) and the left occiput, each
referred to the left mastoid. Given that the latter part of a CNV elicited by a
simple, right-handed response typically has a slight contralateral predomi-
nance over the motor strip, recordings from the central scalp locations and
over the left hemisphere might provide a more precise test of the hypothesis.
In any case, a replication of this type of study as well as recordings of CNV in
ERP Studies 49

a discriminative reaction time task would be worthwhile before we close the

book on slow negative potentials and REM sleep.
The relation between CNV amplitude and non-REM (NREM) sleep has
been studied by Bruneau, Martineau, Ragazzoni, and Roux (1980). Like the
study of Salamy et al. (1979, the recordings were limited to the vertex and
occiput (in this case the right occiput) referenced to the left earlobe. Record-
ings were taken when subjects were awake and while they were asleep. During
sleep, three fifths of the recordings were obtained during Stages 1-2, approxi-
mately two fifths were obtained during Stages 2-3, and several more were
obtained during Stages 3-4. The two conditions in which CNVs were antici-
pated involved a weak sound (i.e., click) followed 880 ms later by an ankle
jerk or Hoffman reflex (a monosynaptic spinal segmental reflex triggered by
electrical stimulation of the 1A fibers of the muscle spindles in the calf [soleus
muscle]). The ankle jerk was elicited in one case by the fall of a hammer beside
the Achilles tendon and in the other by electrical stimulation of the sciatic
nerve. Given the lack of an explicit task, this is not the most auspicious para-
digm for eliciting CNVs. Moreover, the trials in which the click was paired
with an ankle jerk were interspersed with blocks in which the click and the
ankle jerk were each presented alone in the series, without pairing. Again, this
type of equivocation in the warning stimulus to imperative stimulus contin-
gency has been shown to reduce CNV amplitudes somewhat, albeit most se-
verely when the single and paired trials are interspersed within a single experi-
mental run. Although the data from sleep and waking were collected in
counterbalanced order across the subjects, there is no indication that order of
presentation made a difference on the size of the effects. On the whole, it
seemed as if a CNV developed between the warning stimulus and the imper-
ative stimulus during waking, whereas during NREM sleep the results were
more variable with a clear large positivity developing prior to the imperative
stimulus in some cases. The authors suggested that all of the ERPs recorded
during sleep were probably K complexes (biphasic, negative-positive) and that
the slow potential changes obtained by the sound and ankle jerk pairing were
abolished. Again, I believe that more testing with a design better suited for
CNV-elicitation and a more detailed mapping of the potentials across the scalp
are warranted before we generalize this conclusion. Given the controversy sur-
rounding the existence of a CNV as a unitary phenomenon as opposed to a
combination of an orienting response to the warning stimulus (0wave) and a
preparation for movement in response to the imperative stimulus (E wave), it
might also be informative to manipulate the foreperiod interval during sleep
studies. This latter approach has been used to separate the two components in
waking subjects (e.g., Rohrbaugh & Gaillard, 1983).
Thus, is being forewarned helpful in sleep as well? The answer, based on
ERPs, is that we do not know. First, one would want to know whether a
warned signal is responded to more quickly than an unwarned signal and
whether the effect of warning varies with sleep stage. The results of such a
behavioral study would aid in the interpretation of slow potential studies and
50 MARTA KUTAS

would be less costly to pursue. If the CNV requires active engagement of atten-
tional mechanisms and this is not possible during sleep, then no CNVs will be
expected. However, we must question whether the concept of attention can
be invoked in the same breath as sleep. If the CNV is truly an indicator of
CNS excitability, one might expect differences in CNV generation and mainte-
nance between REM and NREM sleep. While it would not be surprising to
find that CNVs are smaller during sleep than during waking, it is theoretically
crucial to determine whether there are any significant differences in CNV am-
plitude, morphology, and scalp distribution during the different stages of sleep.
The critical role of muscarinic cholinergic mechanisms both in REM sleep
(e.g., Hobson, 1988)and in the generation of slow negative potentials such as
the CNV (Marczynski, 1978; Pirch, Corbus, Rigdon, & Lyness, 1986) suggest
that, if CNVs are to be recorded during sleep, they should vary across the
stages, especially between REM and NREM. In addition, if the proposed in-
volvements of REM sleep and CNV amplitude in some aspect of memory
have some validity, one might expect differential effects of sleep stages on
CNV generation and amplitude. Certainly, any study of these issues should
attempt to gather the relevant ERP data from the same subjects during the
different sleep stages under the same experimental conditions (difficult as
this may be).
To date, intuition and evidence lead me to argue that the so-called cogni-
tive ERP components require at least a modicum of attention for their elicita-
tion. The most likely candidate as an exception to this rule is the MMN.’
The MMN was first identified and was subsequently studied by Naatanen and
associates (Naatanen, 1986; Naatanen & Gaillard, 1983; Naatanen, Gaillard,
& Mantysalo, 1978; Naatanen, Sams, & Alho, 1986). The MMN is a negative
component beginning around 100 ms after a stimulus change in a repetitive
homogeneous stimulus stream, even when the change is not consciously per-
ceived. That is, if subjects are asked to attend to one ear in a dichotic listening
task and to ignore stimuli in the other ear, tones in the unattended ear that
deviate from the ongoing sequence of tones elicit an MMN. The larger or more
obtrusive the deviation and the less frequently it occurs, the larger the MMN.
Naatanen has proposed that the MMN reflects the activity of an automatic,
preperceptual neuronal mismatch process; namely, the neurophysiological
basis of echoic memory in audition. The MMN has been found to be modality
specific and is presumed to be independent of and insensitive to attentional
manipulations.
The latter presumption was tested in a situation considered to provide
“the most extreme case of attention to environmental stimuli, namely sleep”
(Paavilainen et al., 1987, p. 247). Seven subjects who were instructed to wake
up early in the morning took part in an experiment that same evening, during
which they listened to tones as they read a book; gradually, they fell asleep.

However, I could easily argue that it does not properly belong in the arsenal of cognitive
components.
ERP Studies 51

Tone bursts (50 ms of a sine wave at 75 dB SPL) were presented at a constant

interstimulus interval of 5 10 ms. Ninety percent of the tones were 1000 Hz,
whereas 10%were 1050 Hz. While awake, subjects read a book and ignored
the tones; afterward they slept while sequences of 800 tones were presented.
The ERPs to the frequent (ie., standard) tones were characterized by a P1 and
an N 1 component. As had been previously reported, the N 1 component was
reduced significantly as the subjects grew drowsy, and it disappeared as they
entered Stage 1 sleep. In contrast, the PI component grew larger in amplitude
with progressing stages of NREM sleep. The reading condition was as ex-
pected The less frequent (i.e., deviant) stimulus was characterized by a large
MMN peaking between 150 and 200 ms, followed by a P3a-like deflection
between 250 and 300 ms. As drowsiness set in, 4 of the subjects continued to
produce significant MMNs and a nonsignificant hint of the P3a. By Stage 1
sleep and thereafter, none of the ERPs to deviant tones contained an MMN.
Three of the subjects continued to show a small, late positivity in response
to the deviant stimulus during Stage 1 sleep. By Stage 2 sleep there were no
differencesin the ERPs elicited by standard and deviant tones. Although some
data were recorded during Stages 3 and 4, the ERPs to deviant tones were
described as distorted by “the low frequency and high amplitude brain waves
characteristicto these sleep stages” (Paavilainen et al., 1987, p. 249). In addi-
tion, no ERPs were recorded during REM because the subjects’sleep architec-
ture was abnormal. These peculiarities included longer-than-normal Stage 2,
spontaneous awakeningafter one or two sleep cycles, and insufficient numbers
of reliable periods of REM. This last observation aside, the results of this
experiment clearly show that the MMN is not elicited by deviant stimuli
during sleep.
It is important to comment that Paavilainen et al. (1987) were the
first to propose that additional experiments with the mismatch negativity
during sleep be carried out and to suggest that such follow-up experiments
should use larger stimulus deviations. They based this recommendation on
the fact that larger deviations are associated with larger MMNs and on the
observation by Csepe, Karmos, and Molnar ( 1987) of an MMN (30-70 ms)
during slow wave sleep in cats. Csepe et al. presented 3- and 4-kHz tones
as deviant and standard tones, respectively, at a number of different target-
to-standard probability ratios (five levels between 5% and 50%)and found
MMNs of an increased latency at some recording sites but only at the lowest
deviant probabilities. While it is sometimes difficult to compare ERP com-
ponents across species, the cat recordings met enough of the necessary cri-
teria to be considered similar to the MMN recorded in man. As such, they
also suggested that Paavilainen et al.’s ( 1987) study with humans be modi-
fied and tried again. In addition to using a greater deviation as they pro-
posed, I also recommend using (a) variable intervals between stimuli (e.g.,
between 300 and 500 ms), (b) recording epochs of at least 800 ms poststim-
ulus, and (c) acclimating the subjects to sleeping in the laboratory so that
their sleep patterns are normal.
52 MARTA KUTAS

FUTURE DIRECTIONS AND POSSlBlLlTlES

As is evident from this review of the scant literature on cognitive electrophysi-
ological investigations of sleep, the door is wide open for studies of this type.
What follows are some of my thoughts on possible directions to pursue.
Although the CNV requires that subjects somehow appreciate the con-
tingency between two stimuli, there are other slow, long-lasting negative po-
tentials of the same family that are easier to record. For example, recordings
made prior to limb movements indicate the presence of a slow-rising negativ-
ity that begins in some cases as early as 1.5 s before movement onset (measured
relative to the response in the moving muscles with electromyography
[EMG]). This negativity, variously called the readiness potential, bereit-
schuftspotentiul, and N 1 of the motor potential, is largest over the motor strip
and, at least in its latter half, has a strong contralateral predominance (e.g.,
Kornhuber & Deecke, 1965). In a series of experiments, Libet (1985) made
inferences about the timing of conscious awareness based on the onset time
of the readiness potential relative to each subject’s verbal report of the inten-
tion to move. Within an experimental psychology tradition, Coles and col-
leagues (e.g., Coles & Gratton, 1986; Gratton, Coles, Sirevaag, Eriksen, &
Donchin, 1988; see also de Jong, Wierda, Mulder, & Mulder, 1988)have used
the asymmetric portion of the readiness potential as a measure of differential
preparation and have used it to make inferences about information flow
through the black box we call a brain (Coles, Gratton, Bashore, Eriksen, &
Donchin, 1985). Do we need to be paying attention to make use of partial
information? What are the odds of recording RPs to spontaneous movement
during sleep-during any stage? If a lucid dreamer can make voluntary move-
ments based on presleep instructions, will these movements be preceded by a
readiness potential? And if RPs are recorded during sleep, must we modify
our conceptions of preparation and voluntariness?
The past 10 years of ERP research have witnessed a strong push in the
areas of language processing and memory, sometimes in combination (Kutas,
1988; Kutas & Van Petten, 1988;Neville, Kutas, Chesney, & Schmidt, 1986).
For example, ERPs to words and pictures show clear repetition effects (Nagy
& Rugg, 1989; Rugg, 1985,1987; R u g , Furda, & Lorist, 1988; Rugg & Nagy,
1987). To what extent is attention necessary to yield this repetition effect?
How will the brain respond to an auditory passage that it has been exposed to
earlier in the day? How will it respond the next morning to a passage that it
has heard during sleep, either slow wave or REM? Will anomalous sentences
seem less strange to subjects if they have been experienced previously during
sleep? Both behavioral and ERP measures may provide answers to such ques-
tions, albeit not necessarily the same ones. The ERP measure has on occasion
proven to be more sensitive to implicit memories than to reaction time mea-
sures (e.g., Besson, Fischler, B o a , & Raney, 1989). If forced to bet, I would
put my money on the ability of implicit memories or procedural learning (lin-
guistic and nonlinguistic) to penetrate the sleeping brain to a greater degree
than those of an explicit or declarative nature. If semantic processing truly
ERP Studies 53

reflects lexical processing without awareness, as claimed by some researchers

(for a review, see Holender, 1986), then two associatively or semantically re-
lated words might prime each other during sleep as well. Examining ERP signs
of semantic association during sleep might shed light on these issues (e.g., Ku-
tas & Hillyard, 1 989).3Certainly, my views on language processing, ERPs, and
sleep would be affected by empirical evidence on these matters. In fact, it is
precisely because my intuitions run counter to any notion wherein people
can simultaneously cogitate about external events and enjoy sleep that I have
proposed ways to prove myself wrong.

References
Amadeo, M., & Shagass, C. ( I 973). Brief latency click-evoked potentials during waking
and sleep in man. Psychophysiology, 10,244-250,
Anch, M. ( 1 977). The auditory evoked brain response during adult human sleep. Wak-
ing and Sleeping, 1, 189- 194.
Bell, I., &Campbell, K. (1 988). In T. W. Picton (Ed.), Human event-related potentials.
EEG handbook (Vol. 3, p. 364). Amsterdam: Elsevier.
Besson, M., Fischler, I., Boaz, T. L., & Raney, G. (1989). Effects ofautomaticsemantic
priming on explicit and implicit memory tests. Manuscript submitted for publica-
tion.
Bruneau, N., Martineau, J., Ragazzoni, A., & Roux, S. (1980). Event-related slow po-
tentials evoked during ankle jerk conditioning in wakefulness and NREM sleep.
Electroencephalography and Clinical Neurophysiology, 49,93- 101 .
Buchsbaum, M., Gillin, J. C., & Pfefferbaum, A. (1 975). Effect of sleep stage and stimu-
lus intensity on auditory average evoked responses. Psychophysiology, 12, 707-
712.
Campbell, K. B., & Bartoli, E. A. (1986). Human auditory evoked potentials during
natural sleep: The early components. Electroencephalographyand Clinical Neuro-
physiology, 65, 142-149.
Campbell, K. B., Picton, T. W., Wolfe, R. G., Baribeau-Braun, J., & Braun, C. (198 I).
Clinical evoked potential studies: 11. Auditory brainstem responses. Sensus, I, 9-
19.
Chiappa, K. H., Gladstone, K. J., & Young, R. R. (1979). Brainstem auditory evoked
responses: Studies of waveform variation in 50 normal human subjects. Archives
ofNeurology, 36, 8 1-87.
Coles, M., & Gratton, G. (1986). Cognitive psychophysiology and the study of states
and processes. In E. R. J. Hockey, A. W. K. Gaillard, & M. G. H. Coles (Eds.),
Energetics and human information processing. Dordrecht, The Netherlands: Mar-
tinus Nijhof.
Coles, M. G. H., Gratton, G., Bashore, T. R., Eriksen, C. W., & Donchin, E. (1985). A
psychophysiological investigation of the continuous flow model of human infor-
mation processing. Journal of Experimental Psychology: Human Perception and
Performance, 11, 529-553.

In a fantasy world of science, I would also be interested in knowing whether segments of

the EEG activity of any given individual recorded during the day would correlate highly with a
segment of equal length recorded later that evening during REM sleep. That is, is there any evi-
dence for the notion that certain processing sequences are replayed for the purpose of consolida-
tion during sleep?
54 MARTA KUTAS

Csepe, V., Karmos, G., & Molnar, M. (1987). Evoked potentials correlates of stimulus
deviance during wakefulness and sleep in cat-animal model of mismatch negativ-
ity. Electroencephalographyand Clinical Neurophysiology, 66, 57 1-578.
de Jong, R., Wierda, M., Mulder, G., & Mulder, L. J. M. (1988). Use ofpartial stimulus
information in response processing. Journal of Experimental Psychology: Human
Perception and Performance, 14, 682-692.
Donchin, E., Ritter, W., & McCallum, W. C. (1978). Cognitive psychophysiology:The
endogenous components of the ERP. In E. Callaway, P. Tueting, & S. H. Koslow
(Eds.), Event-related brain potentials in man (pp. 349-44 1). New York: Academic
Press.
Gaillard, A. W. K., & Ritter, W. (Eds.). (1983). Tutorials in event related potential
research:Endogenous components. Amsterdam: Elsevier/North-Holland.
Galambos, R., & Hecox, K. E. (1978). Clinical applications of the auditory brainstem
response. The Otolaryngologic Clinics of North America, I I , 709-722.
Galambos, R. S., Makeig, S., & Talmachoff, P. J. (1981). A 40-Hz auditory potential
recorded from the human scalp. Proceedings of the National Academy of Sciences,
78,2643-2647.
Goldstein, R., & Rodman, L. (1967). Early components of averaged evoked responses
to rapidly repeated auditory stimuli. Journal of Speech and Hearing Research, 10,
607-705.
Granda, A. M., & Hammack, J. T. (1961). Operant behavior during sleep. Science,
133, 1485-1486.
Gratton, G., Coles, M. G. H., Sirevaag, E., Eriksen, C. W., & Donchin, E. (1988). Pre-
and poststimulus activation of response channels: A psychophysiologicalanalysis.
Journal of Experimental Psychology: Human Perception and Performance, 14,
33 1-344.
Hellekson, C., Allen, A., Greeley, H., Emery, S., & Reeves, A. (1979). Comparison
of inter-wave latencies of brain stem auditory evoked responses in narcoleptics,
primary insomniacs and normal controls. Electroencephalography and Clinical
Neurophysiology, 47, 742-744.
Hillyard, S. A. (1973). The CNV and human behavior. Electroencephalography and
Clinical Neurophysiology, Suppl. 33. Amsterdam: Elsevier.
Hillyard, S. A., & Picton, T. W. (1987). Electrophysiology of cognition. In F. Plum
(Ed.), Handbook ofphysiology, Section I : The nervous system: Vol5. Higherfunc-
tions of the nervous system, Part 2 (pp. 5 19-584). American Physiological Society,
pp. 5 19-584.
Hobson, J. A. (1988). The dreaming brain. New York: Basic Books.
Holender, D. ( 1986). Semantic activation without conscious identification. Behavioral
and Brain Sciences, 9, 1-66.
Johnson, R., Jr. (1988). The amplitude of the P300 component of the event-related
potential: Review and synthesis. In P. K. Ackles, J. R. Jennings, & M. G. H. Coles
(Eds.), Advances in Psychophysiology (Vol. 3, pp. 69-138). Greenwich, C T JAI
Press.
Johnson, R., Jr., Rohrbaugh, J. W., & Parasuraman, R. (Eds.). (1987). Current trends
in event-related potential research, Electroencephalography and Clinical Neuro-
physiology, Suppl. 40. Amsterdam: Elsevier.
Jones, T. A., Stockard, J. J., & Weidner, W. J. (1980). The effects of temperature and
acute alcohol intoxication on the brain stem auditory evoked potential in the cat.
Electroencephalographyand Clinical Neurophysiology, 49,23-30.
Kevanishvili, Z., & von Specht, H. (1979). Human slow auditory evoked potentials
during natural and drug-induced sleep. Electroencephalographyand Clinical Neu-
rophysiology, 47,280-288.
Klein, A. J. (1983). Properties of the brain-stem slow wave component: 1. Latency,
amplitude, and threshold sensitivity.Archives of Otolaryngology, 109, 6- 12.
Kornhuber, H. H., & Deecke, L. (1965). Hirnpotentialanderungen bei Wilkurbeweg-
ungen und passiven Bewegungen des Menschen: Bereitschaftspotential und re-
afferente Potentiale [Changes in brain potential during voluntary and passive
ERP Studies 55

movements in man: The readiness potential and reafferent potential]. Pfruegers

Archivfir GesammeltePhysiologie. 284, 1-1 7.
Kutas, M. (1988). Review of event-related potential studies of memory. In M. S. Gaz-
zaniga (Ed.), Perspectives in memory research (pp. I8 1-2 18). Cambridge, MA:
MIT Press.
Kutas, M., & Hillyard, S. A. (1989). An electrophysiological probe of semantic associa-
tion. Journal of Cognitive Neuroscience, 1,38-49.
Kutas, M., & Van Petten, C. (1988). Event-related brain potential studies of language.
In P. K. Ackles, J. R. Jennings, & M. G. H. Coles (Eds.), Advances inpsychophysi-
ology (Vol. 3, pp. 139- 187). Greenwich, C T JAI Press.
Libet, B. ( 1985). Unconscious cerebral initiative and the role of conscious will in volun-
tary action. Behavioral and Brain Sciences, 8, 529-566.
Linden, R. D., Campbell, K. B., Hamel, G., & Picton, T. W. (1985). Human auditory
steady state evoked potentials during sleep. Ear and Hearing, 6, 167- 174.
Lukas, J. H. (1980). Human attention: The olivo-cochlear bundle may function as a
peripheral filter. Psychophysiology, 17,444-452.
Lukas, J. H. (198 1). The role of efferent inhibition in human auditory attention: An
examination of the auditory brainstem potential. International Journal of Neuro-
science, 12, 137-145.
Marczynski, T. J. ( 1978). Neurochemical mechanisms in the genesis of slow potentials:
A review and some clinical implications. In D. Otto (Ed.), Multidisciplinary per-
spectives in event-related brain potential research (pp. 25-36). Washington, DC
U.S. Government Printing Office.
Marshall, N. K., & Donchin, E. (198 1). Circadian variation in the latency of brainstem
responses: Its relation to body temperature. Science, 212, 356-358.
Mendel, M. I., Hosick, C. E., Windman, T. R., Davis, H., Hirsh, S. K., & Dinges,
D. F. (1975). Audiometric comparison of the middle and late components of the
adult auditory evoked potentials awake and asleep. Electroencephalography and
Clinical Neurophysiology, 38,27-33.
Moore, E. J. (Ed.). (1983). Bases of auditory brain-stem evoked responses. New York:
Grune & Stratton.
Naatanen, R. (1986). Neurophysiological basis of the echoic memory as suggested by
event-related potentials and magnetoencephalogram. In F. Klix & H. Hagendorf
(Eds.), Human memory and cognitive capabilities (pp. 6 15-628). Amsterdam:
Elsevier/North-Holland.
Naatanen, R., & Gaillard, A. W. K. (1983). The orienting reflex and the N2 deflection
of the ERP. In A. W. K. Gaillard & W. Ritter (Eds.), Tutorials in event-related
potential research: Endogenous components (pp. 1 19-329). Amsterdam: Elsevier/
North-Holland.
Naatanen, R., Gaillard, A. W. K., & Mantysalo, S. (1978). The early selective attention
effects on evoked potential reinerpreted. Acta Psychologica, 42, 3 13-329.
Naatanen, R., Sams, M., & Alho, K. (1986). The mismatch negativity: The ERP sign
of a cerebral mismatch process. In W. C. McCallum, R. Zappoli, & F. Denoth
(Eds.), Cerebral psychophysiology, electroencephalography and clinical neuro-
physiology (Suppl. 38, pp. 172- 178). Amsterdam: Elsevier.
Nagy, M. E., & Rugg, M. D. (1989). Modulation of event-related potentials by word
repetition: The effects of inter-item lag. Psychophysiology, 26,43 1-436.
Neville, H. J., Kutas, M., Chesney, G., & Schmidt, A. (1986). Event-related brain po-
tentials during the initial encoding and subsequent recognition memory of con-
gruous and incongruous words. Journal of Memory and Language, 25,75-92.
Ornitz, E. M., Ritvo, E. R., Carr, E. M., Panman, L. M., & Walter, R. D. (1967). The
variability of the auditory averaged evoked response during sleep and dreaming
in children and adults. Electroencephalographyand Clinical Neurophysiology, 22,
5 14-524.
Osterhammel, P. A., Shallop, J. K., & Terkildsen, K. (1985). The effect of sleep on
the auditory brainstem response (ABR) and the middle latency response (MLR).
Scandinavian Audiology, 14,47-50.
56 MARTA KUTAS

Paavilainen, P., Cammann, R., Alho, K., Reinikainen, K., Sams, M., & Naatanen, R.
(1987). Event-related potentials to pitch change in an auditory stimulus sequence
during sleep. In R. Johnson, Jr., J. W. Rohrbaugh, & R. Parasuraman (Eds.), Cur-
rent trends in event-related potential research (EEG Suppl. 40). Amsterdam: Else-
vier/North Holland.
Picton, T. W., & Hillyard, S. A. (1974). Human auditory evoked potentials: 11. Effects
of attention. Electroencephalographyand Clinical Neurophysiology, 36, 19 1-200.
Picton, T. W., Hillyard, S. A., Krausz, H. I., & Galambos, R. (1974). Human auditory
evoked potentials: I. Evaluation of components. Electroencephalography and
Clinical Neurophysiology, 36, 179- 190.
Picton, T. W., Stapells, D. R., & Campbell, K. B. (198 1). Auditory evoked potentials
from the human cochlea and brainstem. Journal of Otolaryngology, 10, 1-41.
Picton, T. W., Woods, D. L., Baribeau-Braun, J., & Healey, T. M. J. (1977). Evoked
potential audiometry. Journal of Otolaryngology, 6, 90- 1 19.
Pirch, J. H., Corbus, M. J., Rigdon, G. C., & Lyness, W. H. (1986). Generation of
cortical event-related slow potentials in the rat involves nucleus basalis cholinergic
innervation. Electroencephalographyand Clinical Neurophysiology, 63,464-475.
Rechtschaffen, A., & Kales, A. (1968). A manual of standardized terminology, tech-
niques, and scoring system for sleep stages of human subjects. Washington, DC:
U.S. Government Printing Office.
Regan, D. (1972). Evoked potentials in psychology, sensory physiology and clinical
medicine. London: Chapman & Hall.
Regan, D. ( 1989). Human brain electrophysiology: Evokedpotentials and evoked mag-
neticfields in science and medicine. Amsterdam: Elsevier/North Holland.
Rohrbaugh, J. W., & Gaillard, A. W. K. (1983). Sensory and motor aspects of the
contingent negative variation. In A. W. K. Gaillard & W. Ritter (Eds.), Tutorials
in event related potential research: Endogenous components (pp. 269-3 10). Am-
sterdam: Elsevier/North-Holland.
Rugg, M. D. (1985). The effects of semantic priming and word repetition on event-
related potentials. Psychophysiology, 22,642-647.
Rugg, M. D. (1987). Dissociation of semantic priming, word and non-word repetition
by event-related potentials. Quarterly Journal of Experimental Psychology, 3YA,
123-148.
R u g , M. D., Furda, J., & Lorist, M. (1988). The effects of task on the modulation of
event-related potentials by word repetition. Psychophysiology, 25, 55-63.
R u g , M. D., & Nagy, M. E. (1987). Lexical contribution to non-word repetition
effects: Evidence from event-related potentials. Memory and Cognition, 15, 473-
481.
Salamy, J., Lester, J., &Jones, K. ( 1975). REM sleep and contingent negative variation
development. Electroencephalography and Clinical Neurophysiology, 39, 20 1-
204.
Shallop, J. K., & Osterhammel, P. A. (1983). A comparative study of measurements
of SN- 10 and the 40/sec middle latency responses in newborns. Scandinavian Au-
diology, 12,91-95.
Skinner, J. E., & Yingling, C. D. (1977). Central gating mechanisms that regulate
event-related potentials and behavior. In J. Desmedt (Ed.), Progress in clinical
neurophysiology: Attention, voluntary contraction, and event-related cerebral po-
tentials (Vol. l, pp. 30-69). Basel, Switzerland: Karger.
Stapells, D. R. ( 1984). Studies in evoked potential audiometry. Unpublished doctoral
dissertation, University of Ottawa, Canada.
Stapells, D. R., Linden, R. D., Suffield, J. B., Hamel, G., & Picton, T. W. (1984).
Human auditory steady state potentials. Ear and Hearing, 5, 105- 1 13.
Stockard, J. J., Sharbrough, F. W., & Tinker, J. A. (1978). Effects of hypothermia on
the human brainstem auditory response. Annals of Neurology, 3, 368-370.
Stockard, J. J., Stockard, J. E., & Sharbrough, F. W. (1978). Nonpathologic factors
influencing brainstem auditory evoked potentials. American Journal of EEG
Technology, 18, 177-209.
ERP Studies 57

Tecce, J. J. ( 1 972). Contingent negative variation (CNV) and psychological processes

in man. Psychological Bulletin, 77, 73-108.
Tecce, J. J., & Hamilton, B. T. (1973). CNV reduction by sustained cognitive activity
(distraction). In W. C. McCallum & J. R. Knott (Eds.), Event-related slow poten-
tials of the brain. Electroencephalography and Clinical Neurophysiology, Suppl.
33,229-237.
Ujszaszi, J., & Halasz, P. (1986). Late component variants of single auditory evoked
responses during NREM sleep Stage 2 in man. Electroencephalographyand Clini-
cal Neurophysiology, 64,260-268.
Walter, W. G., Cooper, R., Aldridge, U. J., McCallum, W. C., & Winter, A. L. (1964).
Contingent negative variation: An electric sign of sensorimotor association and
expectancy in the human brain. Nature, 203, 380-384.
Weitzman, E. E., & Kremen, H. (1965). Auditory evoked responses during different
stages of sleep in man. Electroencephalographyand Clinical Neurophysiology, 18,
65-70.
Williams, H. L., Morlock, H. C., & Morlock, J. U. (1966). Instrumental behavior dur-
ing sleep. Psychophysiology, 2,208.
Williams, H. L., Tepas, D. I., & Morlock, H. C., Jr. (1962). Evoked responses to clicks
and electroencephalographic stages of sleep in man. Science, 138,685-686.
Woldort€, M., Hansen, J. C., & Hillyard, S. A. (1987). Evidence of selective attention
in the midlatency range ofthe human auditory event-related potential. In R. John-
son, J. W. Rohrbaugh, & R. Parasuraman (Eds.), Current trends in event-related
potential research (pp. 146-1 54). Amsterdam: Elsevier.
Woods, D. L., & Hillyard, S. A. (1978). Attention at the cocktail party: Brainstem
evoked responses reveal no peripheral gating. In D. A. Otto (Ed.), Multidisciplin-
ary perspectives in event-related brain potential research (pp. 230-233). Washing-
ton, DC: U.S. Government Printing Office.
 CHAPTER 4

STIMULUS CONTROL AND SLEEP

JOHN HARSH AND PIETRO BADIA

The transfer of learning from the wake to the sleep state has been the subject
of several studies involving conditioned discrimination (e.g., Beh & Barratt,
1 965) and operant control procedures (e.g., Williams, Morlock, & Morlock,
1966). These studies show that stimulus control transfers from waking to
sleep. The degree to which learned behavior can be controlled, however, has
been found to be greatly reduced by sleep. One of the goals of this research was
to investigate factors related to the level of control exerted by stimuli presented
during sleep. An additional goal was to provide further information about the
effects of stimulus control on sleep. Addressing these issues may increase our
understanding of brain organization in sleep and may also lead to procedures
for controlling sleep-related behavioral disorders.

REINFORCEMENT AND RESPONDING

Because earlier reports suggested that reinforcement affects the probability of
responding (e.g., Williams et al., 1966), reinforcement contingencies were var-
ied in our first studies (e.g., Badia et al., 1984; Harsh et al., 1987). Our studies
used young adult volunteer subjects who were told while they were awake that
tones would be presented repeatedly while they were asleep and that their task
would be to respond to the tones by taking a deep breath. Four experimental
conditions were studied. In each case, the interstimulus interval averaged 4
min. Further details regarding the procedures, the control conditions, and the
rationale for response selection can be found in Badia et al. (1984) and Harsh
et al. (1987). The conditions were (a) instructions only, in which subjects were
told that their task was to respond to the tone (65 dB of 30-s duration); (b)
escape, in which subjects were told to respond to the tone and that responding
would terminate the tone (65 dB of 30-s maximum duration); (c) escape/
58
Stimulus Control 59

Figure 1 Mean response probability and mean response latency for responses dur-
ing non-REM (NREM) and REM sleep under different test conditions.

avoidance, in which subjects were told to respond to terminate the tone (65
dB of 30-s maximum duration) and that failure to do so would result in a
louder (90 dB) tone; and (d) incremental escape/avoidance, in which subjects
were told to respond to terminate the tone that was initially 45 dB but, in the
absence of a response, increased in 10 dB increments every 10 s up to 95 dB.
The results of these studies for the first night of testing are shown in
Figure 1. Note first the data for the three procedures that involved responding
to a fixed intensity (65 dB) tone. In the instructions only group, the percentage
of responding varied across sleep stages, with responses during sleep Stages 3
and 4 occumng relatively infrequently (see top of Figure 1). Having subjects
respond by terminating the tones (escape group) or avoiding the presentation
of louder tones (escape/avoidance group) resulted in an increased likelihood
of responding. Responding occurred more quickly in the escape/avoidance
group (see bottom of Figure 1). Thus, with these three procedures, responding
to a 65-dB tone varied as a function of the consequences of the responding.
These findings are consistent with those of earlier studies concerning rein-
forcement for responding in sleep (cf. Williams et al., 1966).
The incremental escape/avoidance procedure resulted in the greatest
control over behavior. Responding occurred on 90-100% of the trials in each
stage of sleep. Note that the average response latency with the incremental
procedure was 20-25 s during Stages 2,3, and 4 and during REM sleep. This
means that the tone intensity had increased from 45 dB to 65 dB at the time
of responding, which was the decibel level of the fixed-intensity tone under
the other three procedures. Presumably, the greater behavioral control with
60 JOHN HARSH AND PIETRO BADIA

the incremental procedure, at the same average decibel level as the other pro-
cedures, is attributable to fluctuations in response thresholds during all stages
of sleep.
The data from the four experimental procedures demonstrate that reli-
able control of behavior in all stages of sleep can be obtained with the selective
use of reinforcement contingencies. Significantly, the control was obtained
with stimuli presented at a high rate (every 4-5 min). Although the data in
Figure 1 are from only the first night of testing, reliable responding with the
incremental procedure has been maintained across 10 consecutive test nights
(Badia, Harsh, & Balkin, 1986). Because the incremental procedure proved to
be the most successful in maintaining responding, it was used in subsequent
studies.

EFFECTS ON SLEEP
Earlier investigators concerned with behavioral control in sleep studied the
electroencephalographic (EEG) record at the time of test stimuli to assess
whether subjects had responded while they were asleep or following arousal
from sleep (e.g., Williams et al., 1966). Using EEG alpha as evidence of awak-
ening, it has generally been concluded that subjects are able to respond during
sleep. Responding without alpha activity has frequently been observed in our
studies. However, we approached the question of whether sleep was disturbed
by the behavioral control procedure more generally and included subjective
sleep reports, physiological measures of variables related to sleep, and mea-
sures of daytime sleepiness.

Subject Reports
The morning reports we obtained in early experiments (e.g., Badia et al., 1984;
Badia, Harsh, Balkin, ORourke, & Burton, 1985; Harsh et al., 1987)assessed
subjects’ perceptions of sleep quality and also assessed recall of stimulus/re-
sponse events that occurred in sleep. In these experiments, subjects reliably
responded to stimuli presented on an average of once every 4-8 min. When
asked to report on the quality of their sleep, subjects typically reported that
their sleep was not disturbed or was only minimally disturbed. When asked
how many tones were presented during the night, subjects typically reported
that only 6- 10 tones (up to 100 were actually presented) had been presented.
Thus, although behavior was reliably controlled by frequently presented stim-
uli, there was little awareness of the stimulus presentations and no or minimal
sleep disruption was attributed to the behavioral control procedure.
The results were different when subject reports were obtained by waking
subjectsjust after tone presentations. As part of a study of the relation between
cognitive activity and response variables (see Burton, Harsh, & Badia, 1988),
subjects were awakened 30 s after tones were presented during Stage 2 and
REM sleep. The stimulus presentation procedure in this experiment was tai-
lored to provide reports following both response and no response trials. Addi-
tionally, the tones were presented relatively infrequently (a variable interval
Stimulus Control 61

Figure 2 Percentage of time spent awake in non-REM (NREM) sleep stages and
in REM sleep under control condition and 4-min and 1-min interstimulus interval
experimental conditions.

45-min schedule). As with the morning reports, the results support the conclu-
sion that subjects are capable of responding without the perception of waking.
That is, subjects responded without reporting awakenings on an appreciable
number of trials. However, following approximately 50% of the responses,
subjects reported that they had been awakened by the tones. This contrasts
with the morning recall of less than 10% of the tones. Thus, subject reports
obtained more closely in time to the stimulus/response events were more
likely to indicate perceived wakefulness.

Physiological Variables
Electrophysiological data were recorded in each of our experiments. Thirty-
second epochs of sleep records were scored following conventional guidelines
(Rechtschaffen & Kales, 1968). The results obtained across experiments pro-
vide a consistent picture of the effects of behavioral control on sleep structure.
Thus, when subjects are tested with an average interstimulus interval of 4 rnin
or more, the only effect on structure is a reduction in Stage 4 sleep (Badia et
al., 1984; Badia et al., 1985; Harsh et al., 1987). The observation of minimal
change in sleep structure under these test conditions is in accord with the sub-
jects’ morning reports (described previously) of minimal sleep disturbance.
Data obtained by Magee, Harsh, and Badia (1987) indicate that inter-
stimulus intervals shorter than 4 rnin can result in severe sleep disruption.
Magee et al. presented stimuli after every 4 rnin or after every 1 rnin of sleep.
As in previous studies, under the 4-min condition, sleep structure was mini-
mally disturbed. However, when stimuli were presented after every 1 rnin of
sleep, sleep structure was profoundly altered. Figure 2, which is based on data
62 JOHN HARSH AND PIETRO BADIA

Figure 3 The percentage of trials followed by awakenings, stage shifts, brief changes,
and no changes in the sleep record.

from the Badia et al. (1985), Harsh et al. ( 1987), and Magee et al. (1987) stud-
ies, illustrates this outcome. It appears that, with young adults, sleep structure
is minimally disturbed with interstimulus intervals of 4 min or more but is
greatly disturbed with an interstimulus interval of 1 min. It should be noted
that although older adults are as behaviorally responsive, their sleep is much
more vulnerable to disruption (Harsh, Purvis, Badia, & Magee, 1989).
The recordings at the time of and just after tone presentations were also
examined for evidence of brief electrophysiological changes attributable to
stimulus presentation and/or response emission. Figure 3 presents data on
the brief electrophysiological changes. Included is the percentage of tones and
responses accompanied by awakenings (as scored in 30-s epochs), shifts in
sleep stage, alpha events (<15-sbursts), EEG speeding (< 15-sincrease in EEG
frequency), and increases in electromyograph (EMG) recording. Also pre-
sented is the percentage of trials accompanied by none of the changes (no
change). The data were obtained using the incremental procedure, with tones
occurring every 4 min on the average. Tone intensity began at 45 dB and
increased in 10 dB increments every 10 s up to 95 dB. It can be seen that
awakenings occurred only infrequently, and short bursts of alpha, which are
sometimes taken as a sign of brief awakening or arousal, occurred during or
followingonly 20%of the tones. Thus, responding occurred during polygraph-
ically defined sleep on the large majority of trials.
Although awakenings may have been infrequent, some changes were
found on all but 16%of the trials. The significance of these changes in relation
to the sleep/wake continuum and to overall sleep quality is unclear. It is inter-
esting in this regard that, when tones begin at a higher dB level and increase
more rapidly (see Magee et al., 1987), there is a much greater probability of
brief electrophysiological changes; however, as long as the interstimulus inter-
val is not less than 4 min, there are minimal effects on sleep structure and
daytime sleepiness.
Stimulus Control 63

Data reported by Harsh, Badia, ORourke, and Burton (1983) indicate

that the effect of behavioral control on sleep includes suppression of sleep spin-
dles. These investigators selected tone/response trials during Stage 2 sleep that
did not result in awakenings or stage shifts and studied the distribution of
K-complexes and spindles during the 1-min period immediately preceding
stimulus presentations and during the 1-min period immediately following
stimulus termination. No significant effects were found for the distribution of
K-complexes; however, spindle production was very obviously suppressed.
The time course of the effect on spindle distribution is very clear, lasting ap-
proximately 40 s. As with the measures discussed previously, it is not clear
what the suppression of spindles means with regard to the sleep/wake contin-
uum or to sleep quality. Further research is needed to clarify these relations.

Daytime Sleepiness
One index of sleep quality is level of daytime sleepiness. Badia et al. (1985)
assessed daytime sleepiness under three test conditions. Under one condition,
stimuli were presented with an average interstimulus interval of 8 min. Under
a second condition, the average interstimulus interval was 4 min. The third
condition was a control condition during which stimuli were not presented
and subjects slept without disturbance. The nighttime test conditions were
continued for 4 consecutive nights, and daytime sleepiness was assessed in
morning and afternoon naps. No evidence was found of differential daytime
sleepiness.
Magee et al. (1987) added to these data. Subjects were tested for 1 night
under two experimental conditions, that is, stimuli were presented after every
4 min of sleep or after every 1 min of sleep. Daytime sleepiness was not in-
creased under the 4-min condition relative to the control condition, replicat-
ing the findings of Badia et al. ( 1985). Thus, presenting signal stimuli as fre-
quently as once every 4 min of sleep is not sufficient to disrupt the process
underlying sleep’s function of minimizing daytime sleepiness. This finding is
consistent with the subjects’ morning reports (described earlier) that their sleep
had been normal or nearly normal. The finding also fits well with the observa-
tion that sleep structure is not substantially altered under these conditions. In
contrast, under the 1-min condition, subjects were profoundly sleepy. The
sleepiness data from this condition are consistent with the finding of a greatly
altered sleep structure. This outcome reinforces the conclusion that there are
limits on the ability to respond to stimuli without disrupting fundamental
processes.

STIMULUS CONTROL AND SLEEP STAGE

The data from the studies reviewed thus far indicate that, with the appropriate
contingencies,response probability can be maintained at a high level through-
out the night without greatly disrupting sleep. However, variability in response
latencieswas found in these studies. Some subjects responded far more rapidly
than others, and for all subjects, great variability in response latency occurred
64 JOHN HARSH AND PIETRO BADIA

across the night. At least some of this between-subjects and within-subject

variability is associated with sleep stages. Figure 1 depicts the variability in
response latencies associated with sleep stage under the escape, escape/avoid-
ance, and incremental escape/avoidance procedures described earlier. As can
be seen, sleep stage was more strongly related to response latencies under the
escape and escape/avoidance procedure (significantdifferenceswere found be-
tween all sleep stages; Badia et al., 1984). With the incremental procedure,
however, only Stage 1 latencies were reliably different from those of other
stages (see Badia et al., 1985; Magee et al., 1987). The absence of stage differ-
ences in responsiveness prompted a closer look at the within-night variability
under the incremental procedure.
Harsh et al. ( 1987),using the incremental procedure, examined the aver-
age response latencies for each hour of the night and reported that response
latencies increased linearly across the night. This function was independent of
sleep stage effects. That is, response latencies increased when only Stage 2 or
only REM responses were included in the analysis. More recent research ex-
amining hour by hour variation in latencies suggests that the time-of-night
function is actually curvilinear and is closely related to the body temperature
curve (Lammers, Carpenter, Harsh, & Badia, 1989). These data establish that
response variability during the night is due to more than sleep stage variation.
Several interpretations of the time-of-night functions described have been
considered (e.g., Harsh et al., 1987). Although more than one factor may be
involved, the curvilinearity observed and the correspondence between respon-
siveness and body temperature suggest that responsivity is strongly influenced
by a circadian process.
Recent analyses have provided a clearer indication of the relation be-
tween responsiveness and sleep stage variation. Harsh, Stone, Leiker, and
Badia (1989) computed the mean response latencies for successive 10-min
periods of the night for each of the subjects tested in the Magee et al. (1987)
study and compared the rhythmic variation evident in these time series with
the cycling of REM and non-REM (NREM) sleep. Spectral analysis of the
response latency time series indicated a periodicity in the range of 100-133
min for several of the subjects. Cross correlation analysis failed to provide
evidence of a close relation between response rhythmicity and sleep stages.
However, further analysis (see Harsh et al., 1989, for details) indicated that
response latency is complexly related to time of night, sleep stage, and the
order and duration of REMlNREM Cycles. Figure 4 illustrates these relations.
The data are from the 6 subjects tested under the 4-min condition who had
four identifiable REMINREM cycles. Mean response latencies are plotted for
successivetenths of each REM cycle. The approximate onset and termination
of each REM period is indicated. Several effects can be seen in this figure,
including (a) an increase in mean response latency across the night, (b) a rhyth-
mic variation in response latencies beginning after the first REMINREM cycle,
and (c) a tendency for REM to occur in the troughs of the response cycles.
Stimulus Control 65

REM REM REM REM

1 2 3 4

i\

;"
6 10
I
6
- 10 6

T E N T H S OF REM CYCLE
10 6 10

Figure 4 Mean response latencies across consecutive tenths of REM cycles. Means
are based on subjects ( N = 6) with four identifiable REY/NREM cycles during the night.

CONCLUSlON
The findings of this research suggest the followingconclusions. First, with rein-
forcement for responding, control of learned behavior can be maintained reli-
ably by stimuli presented during sleep. Second, when stimuli are presented 4
min or more apart, behavioral control results in little or no change in sleep
structure, in daytime sleepiness, or in perceptions of sleep quality. Neither
perceived wakefulness nor wakefulness as it is scored on the sleep record are
necessary for responding, although stimulus/response events typically result
in brief EEG or EMG change. Third, within-subject, within-night variance in
responsiveness is complexly related to time of night, sleep stage, and REM/
NREM Cycle.

References
Badia, P., Harsh, J., & Balkin, T. (1986). Behavioral control over sleeping respiration
for 10 consecutive nights. Psychophysiology,23,409-4 1 1.
Badia, P., Harsh, J., Balkin, T., Cantrell, P., Klempert, A., O'Rourke, D., & Schoen,
L. (1984). Behavioral control of respiration in sleep. Psychophysiology,21, 494-
500.
66 JOHN HARSH AND PIETRO BADlA

Badia, P., Harsh, J., Balkin, T., ORourke, D., & Burton, S. (1985). Behavioral control
of respiration in sleep and sleepiness due to signal-induced sleep fragmentation.
Psychophysiology, 22, 5 17-524.
Beh, H. C., & Barratt, P.E. H. (1965). Discrimination and conditioning during sleep
as indicated by the electroencephalogram.Science, 147, 1470- 1471.
Burton, S., Harsh, P., & Badia, P. (1988). Cognitive activity in sleep and responsiveness
to external stimuli. Sleep, 11, 6 1-68.
Harsh, J., Badia, P., ORourke, D., & Burton, S. (1983). K-complexes and sleep spin-
dles in relation to behavioral responses elicited during sleep. Psychophysiology,
20,444.
Harsh, J., Badia, P., ORourke, D., Burton, S., Revis, C., & Magee, J. (1987). Factors
related to behavioral control by stimuli presented during sleep. Psychophysiology,
24,528-534.
Harsh, J., Purvis, P., Badia, P., & Magee, J. (in press). Behavioral responsiveness in
sleeping older adults. Biological Psychology.
Harsh, J., Stone, P., Leiker, M., & Badia, P. (1989). Ultradian rhythms in responsive-
ness to stimuli presented during sleep. Manuscript submitted for publication.
Lammers, W., Carpenter, G., Harsh, J., & Badia, P. (1989, June). Responsivity in sleep,
temperature, and time of night. Paper presented at the annual meeting of the Asso-
ciation of Professional Sleep Societies, Washington, DC.
Magee, J., Harsh, J., & Badia, P. (1987). Effects of experimentally induced sleep frag-
mentation on sleep and sleepiness. Psychophysiology, 24, 528-534.
Rechtschaffen, A., & Kales, A. (Eds.) (1968). A manual of standardized terminology,
techniques, and scoring system for sleep stages of human subjects. Los Angeles:
Brain Information Service/Brain Research Institute, University of California at
Los Angeles.
Williams, H. L., Morlock, H. C., & Morlock, J. V. (1966). Instrumental behavior dur-
ing sleep. Psychophysiology, 2,208-2 16.
 CHAPTER 3

MEMORIES IN SLEEP:
OLD AND NEW
PIETRO BADIA

Observations of learning, memory formation, and information processing in

sleep have resulted in a confusing body of literature and in contradictory con-
clusions concerning what does and does not occur. To clarify some of these
issues, I distinguish between forming new memories in sleep (learning) and
retrieving old memories in sleep (i.e., information processing). Several ques-
tions are considered. Can individuals access established memories and process
information in sleep? Are individuals generally or selectively amnesic to new
events occumng in sleep? That is, can new memories be formed in sleep? If so,
what kind? I conclude with two proposals. One deals with a set of organizing
principles for learning in sleep that relate to categories of knowledge: what can
be remembered and what cannot-in brief, the limits of memory in sleep.
The second proposal deals with a theoretical model of brain (hippocampal)
functioning in sleep that may underlie these memory limits.
Individuals interested in information processing and memory in sleep
have used different research strategies to approach the problem. The differ-
ences relate to the state in which material is presented and the state in which
testing takes place. Researchersexplicitly interested in memory formation use
either a sleep-sleep procedure or a sleep-wake procedure. With a sleep-sleep
procedure, material is presented in the sleeping state, and evidence for mem-
ory is also tested in the sleeping state. That is, the investigation focuses on
whether information can be acquired in the sleep state and can also be altered
in the sleep state (acquisition and extinction of conditioning).
With the sleepwake procedure, information is also presented in the
sleeping state but, in this case, tests of retention occur in the waking state. The
question asked is whether memories formed in sleep carry over to waking. Of
67
68 PIETRO BADlA

the two, the sleep-wake procedure is the one most commonly used. Two vari-
ants of the latter are used. Some researchers test participants when they
awaken in the morning after presenting material repeatedly through the night,
with the focus on the formation of long-term memory (i.e., learning in sleep).
Other researchers awaken the participants immediately or shortly after pre-
senting the material and focus on the duration of short-term memory.
There is a third procedure that relates to the transfer of training and
retrieval issues and not to the formation of new memories. Researchers using
this procedure present information in the waking state and then test for reten-
tion in the sleeping state. With this wake-sleep procedure, the focus is usually
on information processing (i.e., on retrieval in sleep of material learned in the
waking state). The wake-sleep procedure does not assess the formation of new
memories in sleep. It is important to distinguish studies focusing on the re-
trieval of old memories in sleep (information processing) from those focusing
on the formation of new memories in sleep. There is no doubt that informa-
tion processing occurs in sleep; there is some doubt about the formation of
new memories in sleep.

CREATING VERSUS RETRIEVING MEMORIES

Several studies illustrate the distinction between retrieving old memories
in sleep and creating new memories in sleep (e.g., Badia, Harsh, Balkin,
ORourke, & Burton, 1985; Koukkou & Lehmann, 1968).Let me give you an
example by taking you through the procedure. We can instruct subjects in the
waking state that a tone will be presented while they are sleeping. We instruct
them that whenever the tone is perceived they are to make a behavioral re-
sponse (close a microswitch, take a deep breath). A few practice trials are pre-
sented in waking to ensure that they understand and can follow the instruc-
tions. We then allow them to sleep and, periodically during the night (e.g.,
every 4 min), present a series of the tones to them. We find that participants
respond to the tone in sleep at a very high level, nearly 100%of the time in
some cases, and on most trials they respond without arousing. They demon-
strate that instructions given while awake (i.e., learning occurs in waking) can
be retrieved while asleep.
The fact that subjects detect the stimulus, recall the instructions, and
select the proper response while asleep indicates that information can be pro-
cessed in sleep and that both short-term memory and long-term memory are
functioning. However, when subjects are asked the following morning the
number of tones presented and the number to which they responded, they
report virtually no memory for hearing the tones or for having responded. It
does not matter whether they are questioned in the morning or awakened
after the behavioral response at night. Subjects are able to process, in sleep,
information acquired in waking, but they cannot remember doing so.
Why are these new events during sleep (hearing tones and responding to
them) not reported? One could argue that no new memories were formed in
sleep. Subjects heard the tones and were able to retrieve the old memory (“Re-
Memories in Sleep 69

spond when you hear the tone”) that they learned while in the awake state.
They were able to process the information and to respond to the tones, but
they were not able to form the new memory of having detected the tones and
having responded to them.
You may have noted that such findings show a marked similarity to
anterograde amnesia, that is, old experiences can be retrieved but new experi-
ences are not remembered. For instance, this deficit closely resembles the
memory deficit present in people with hippocampal damage. Indeed, later I
will suggest that the basis for “amnesia” in sleep results directly from a hippo-
campal deficit. Let me return to the literature on memory formation and sum-
marize the findings without giving the details of each study.
To an extent, we are all familiar with the basic findings concerning mem-
ory formation (learning) in sleep. This literature is quite consistent. When the
material to be learned or remembered is factual (names, numbers, events,
etc.), the evidence for learning or for memory in sleep is systematically nega-
tive. The conclusions drawn about 30 years ago by Emmons and Simon (1956)
and Simon and Emmons (1956) have been confirmed, elaborated, and
strengthened, with few exceptions (Aarons, 1976; Eich, 1989). In essence, the
conclusions concerning memory formation of factual material relate to the
occurrence of an electroencephalographic (EEG) waking pattern or alpha. In
brief, the probability of learning in sleep is directly related to the duration of
alpha emitted at the time the material is presented.’ The exact durations of
alpha necessary for the formation of long-term memory vary slightly and de-
pend on the delay interval of testing. In general, it is thought that alpha dura-
tions less than 5-10 s at the time material is presented will not permit even
weak learning to occur (Koukkou & Lehmann, 1968). In addition, studies
have indicated that the longer the duration of alpha, the more accurate and
longer the memory (Lehmann & Koukkou, 1974; Shimizu et al., 1977). I
should note, however, that when short-term memory is of primary interest
and subjects are awakened immediately after the information is presented, the
findings are less consistent. Some researchers report short-term memory in the
absence of alpha (Oltman et al., 1977)and others do not (Shimizu et al., 1977).
The generalization regarding alpha responding applies to the formation
of new memories in sleep and does not apply to accessing old memories in

’ I should note a series of experiments conducted by Evans, Gustafson, OConnell, Ome,

and Shor (1970) that Seem to conflict with the statements made here. Hypnotizable subjects pre-
sented with suggestions and cue words in REM sleep apparently were able to memorize the sugges-
tions in that they followed instructions when given the cue words. They had no memory for doing
so in the awake state. The phenomenon described by Evans et al. may be very different from
that discussed here because (a) only selective subjects (deeply hypnotizable) demonstrated the
phenomenon; (b) responses occurred only when suggestions and cue words were presented in
REM; (c) responses to suggestions were very low, varying from 14% to 2 1% for the various experi-
ments; (d) mean latency to respond was very long, varying from 32 s to I 10 s across experiments;
and (e) alpha was evoked by the procedure on occasion. In addition, items c and d suggest that
the words presented by Evans et al. may have been incorporated into a dream Sequence. There is
considerable evidence that “incorporation” does occur.
70 PIETRO BADlA

sleep. Subjects can retrieve material already stored in the long-term memory
and process the information in short-term (working) memory in the absence
of alpha.
How have these findings on memory formation in sleep been inter-
preted? One interpretation focuses on state dependency and retrieval pro-
cesses. The argument is that memory is state dependent; if memory occurs in
the state of sleep, it cannot be retrieved in the state of awake. In the previous
example, a retrieval view would argue that the new memory of responding to
the tones in sleep does exist; it simply cannot be retrieved in the waking state.
The state-dependency hypothesis has received considerable support in assess-
ments of drug and nondrug states. However, I should note that the transfer of
memory from wake to sleep, and not from sleep to wake, suggests that state
dependency, if it is operating, is asymmetrical. In addition, other data that I
will note shortly are incompatible with the state dependency view.
One could reject the state-dependency and retrieval view and, as noted,
argue that new memories in sleep are not formed. We could then theorize
about why they are not formed. For example, it could be argued that new
memories cannot be formed in sleep because the duration of short-term mem-
ory is too brief to permit rehearsal. Without rehearsal, transfer to long-term
memory and consolidation cannot occur.
Or one could argue that the link between short-term and long-term
memory is severed in sleep. In a sense, the argument is that the brain in sleep
is organized in a way that does not permit the formation of memories. Further-
more, the functional state of this organization is partially revealed by the pre-
vailing EEG pattern. For memories to be formed, alpha responding (an EEG
awake pattern) must be present during the presentation of the material to be
learned (short-term memory trace). I will discuss this later. Let me now turn
to some positive findings concerning learning in sleep. These findings involve
procedures that require less cognitive loading.
In brief, studies have reported that classical conditioning and habitua-
tion (for humans and nonhumans) occur in the sleeping state and that the
effects carry over to the waking state. These studies demonstrate that learning,
in the absence of alpha, is possible. Interpreting both positive and negative
instances of learning in sleep thus becomes a special challenge for emerging
theories.

CONDITIONING IN SLEEP
Several well-conducted studies have used classical conditioning in humans to
demonstrate learning (memory) in sleep. An elegant study composed of sev-
eral experiments was reported by Beh and Barratt (1965). In their first experi-
ment, they showed that classical conditioning in the waking state transfers to
the sleeping state, a finding that others have since documented. As noted, such
findings relate to the transfer of learning from waking to sleep and not to mem-
ory formation per se. Beh and Barratt’s second study is of greater theoretical
interest and relates directly to memory formation in sleep. They presented
Memories in Sleep 71

conditioned stimulus (CS) and unconditioned stimulus (US) trials in Stage 2

of sleep using K complexes as a measure of conditioning. Different tones
served as CS+ and CS-, and finger shock served as the unconditioned stimu-
lus. Experimental subjects had CS and US pairings; control subjects had un-
paired presentations. Pilot work revealed the shock intensity needed to avoid
stage shifts or the elicitation of alpha. The authors emphasized that there were
no awakenings for the pairings.
Clear evidence for conditioning in Stage 2 sleep occurred. Once the ac-
quisition of conditioning was demonstrated, Beh and Barratt ( 1965) tested
whether extinction would occur in sleep by giving CS+ alone trials. Extinction
of the response was also demonstrated. Then in a later session during waking,
they tested whether a CS+ that was conditioned in sleep would result in greater
alpha blocking than the CS-. They found that significantly greater alpha
blocking occurred to the CS+ than to the CS-. These results provide powerful
evidence for the acquisition of learning in sleep and for the transfer of this
learning to the waking state. Research in lower animals has also indicated that
the acquisition of conditioning can occur in sleep. For example, a study by
Maho and Bloch ( 1983) showed that rats could be conditioned in REM sleep
and that the conditioning effects transferred to the waking state.
Other studies with humans, not as complete as the Beh and Barratt
( 1965) study, provide additional support for memory formation (learning) in
sleep. McDonald, Schicht, Frazier, Shallenberger, and Edwards ( 1975) classi-
cally conditioned humans in the waking state and tested them in the sleeping
state. They showed that conditioning carried over to the sleeping state and,
more important for our interests, that conditioning could be extinguished in
sleep. The effects were found with vasoconstriction, heart rate, and K com-
plexes. Conditioning findings very similar to those reported by McDonald et
al. ( 1975)were also reported with cats.
In addition to conditioning, a number of studies have focused on habitu-
ation in sleep. Habituation is thought to be a simple form of learning and is
used frequently as a learning paradigm. Unfortunately, the findings on habitu-
ation with humans are less consistent. Some studies have reported evidence
for habituation in sleep, while others have not. This inconsistency may relate
to the existence of greater individual variability in sleep than in waking, which
may render studies less sensitive to detecting an effect. A description of factors
affecting responsivity in sleep was summarized by Bonnet ( 1982).
In any event, several recent, well-conducted studies provide support for
habituation in sleep. Johnson, Townsend, and Wilson (1975) showed that ha-
bituation occurred in sleep using vasoconstriction and heart rate responses.
They also found that habituation in sleep carried over to the waking state.
Similarly, McDonald and Carpenter ( 1975)found with humans that both ha-
bituation and dishabituation occurred in all stages of sleep (vasoconstriction
and heart rate response measures). Several other studies using heart rate as a
measure have also shown that habituation in sleep occurs (Firth, 1973).
72 PIETRO BADlA

Although the literature on classical conditioning and habituation is not

extensive, the studies thus far reviewed provide clear evidence that learning
occurs in sleep and that it transfers to the waking state.
Let me summarize the findings on memory in sleep to this point. The
evidence for short-term memory in sleep and retrieval from long-term mem-
ory is substantial. Material learned while awake can be transferred to sleep,
retrieved from long-term memory, and worked upon. There is also evidence
that conditioning and habituation occur in sleep and that they transfer to the
waking state. However, there is no evidence that factual information (i.e.,
events, words, and numbers) presented, perceived, and responded to in sleep
is converted to long-term memory and remembered in waking.
These data suggest that brain functioning is different in waking and
sleeping. The finding that the learning of factual knowledge in sleep depends
on the presence of alpha suggests that the neural gateway transmitting this
type of knowledge to long-term memory functions only in waking. However,
simple conditioning and habituation do occur in sleep and in the absence of
alpha. How can we account for these two sets of data? Current hypotheses
concerning memory do not adequately organize these data, nor do they pro-
vide insights concerning brain functioning in sleep. New organizing principles
are needed.
As a prelude to suggesting such principles, let me note the salient findings
that must be addressed concerning memory in sleep: (a) short-term (working)
memory is functional, (b) long-term memory is accessible, (c) learning such
as conditioning and habituation does occur, and (d) learning of new factual
material does not occur.
As noted earlier, these findings are strikingly similar to another set of
findings related to lesions in the medial temporal region of humans and to the
hippocampus in nonhumans. The similarity is especially striking for certain
anterograde amnesia patients. This similarity suggests that brain functioning
in amnesic patients, and in animals with similar pathology, might serve as a
model for brain functioning in sleep (or conversely, brain functioning in sleep
could serve as a model for this pathology).
Several benefits will accrue if this model is appropriate: (a) The knowl-
edge gained from the study of this pathology may advance our understanding
of brain functioning and learning in sleep; (b) the organizing principles that
come with the model may be applied to the positive and negative findings
on learning; (c) further, the model may suggest the brain structures that are
responsible for the findings; and (d) the model may point to new research
directions to pursue.

DECLARATIVE VERSUS PROCEDURAL KNOWLEDGE

To develop the argument, it is necessary to distinguish between two other
forms of memory in addition to short-term and long-term memory, namely,
declarative and procedural. This distinction has been helpful in understanding
the memory deficits in amnesia patients, and it provides an organization for
Memories in Sleep 73

interpreting the data on learning and memory in sleep. I should note that the
distinction between declarative and procedural knowledge is also supported
by a rich literature involving brain lesioned animals (Isaacson, 1982). The
distinction first appeared in the literature on artificial intelligence and was later
applied to biological memory by Squire (1987). Other memory distinctions
also exist, but they have not achieved the same biological anchoring (see
Sherry & Schacter, 1987, for an excellent analysis on the topic).
Generally, the distinction between declarative and procedural knowl-
edge is clear. Declarative knowledge is explicit, everyday remembering of fac-
tual content, sentences, words, numbers, events, names, and so forth. The
information is acquired quickly and is verbalized with ease. The acquisition
of declarative knowledge is severely impaired in patients with anterograde am-
nesia and is dependent on the integrity of the medial temporal region.
In contrast, procedural knowledge is implicit, is acquired more slowly, is
not easily verbalized, and is closely tied to procedure. It includes skill learning,
perceptual learning, classical conditioning, habituation, sensitization, and
priming. By its nature, procedural knowledge does not depend on one brain
structure or region and is little, if at all, impaired in patients with anterograde
amnesia.

THE BRAIN AND MEMORY

Evidence suggestingthat the brain is organized in a way compatible with these
memory distinctions comes from several sources. One source is the study of
patients with amnesia. A particularly well-known patient is referred to as RB.
The characteristicsof this patient are well-summarized in Squire’sinformative
book (1987) entitled Memory and Brain. The descriptions that follow were
drawn from this book. In general, there was no discernible impairment of
higher cognitive functioning for RB except for his marked incapability to form
new memories for factual material.
At age 52, RB had an ischemic episode as a complication of surgery and
demonstrated anterograde amnesia after the episode. His IQ was 11 1, there
was no aphasia, and no neuropsychological signs of frontal lobe involvement
were evident. No cognitive deficits other than amnesia were ever detected. He
was tested over a 5-year period and, during this time, asked the same ques-
tions, told the same stories, and did not learn the names or faces of people
seen regularly. He also failed to learn paired associates. Both short-term mem-
ory and remote long-term memory were intact.
An examination of RB’s brain following his death revealed that signifi-
cant bilateral lesions occurred only in the hippocampal formation. There was
also a patchy loss of cerebellar Purkinje cells. The lesions in the hippocampus
were limited to the pyramidal cells of the CAI field. Other surrounding struc-
tures on both sides of the brain were normal. This case firmly established that
restricted hippocampal damage can result in global anterograde amnesia for
factual knowledge.
74 PIETRO BADlA

The different effects of procedural and declarative knowledge on amne-

sic patients are interesting and informative. Simple classical conditioning and
habituation can be demonstrated, without recall of the event predicted by the
CS. Similarly, amnesics can show marked improvement on a mirror tracing
task yet cannot remember that they had multiple training sessions. Puzzles
can also be learned with increasing speed. Priming occurs, despite the failure
of recall and recognition, but only when testing is treated as a nonmem-
ory task.
Animals with hippocampal lesions show similar differences between
procedural and declarative knowledge. For example, simple classical condi-
tioning occurs with these animals. Thus, procedural knowledge is intact. How-
ever, lesioned animals, compared with controls, behave very differently when
conditioning involves declarative knowledge. For example, hippocampal-le-
sioned animals (rats, cats, primates) do not show blocking or overshadowing
(e.g., Rickert, Lorden, Dawson, Smyly, & Callahan, 1979), reversal learning
(Berger & Orr, 1983),latent inhibition, or conditional discrimination (Squire,
1987). In each of these instances, for learning to occur, the animal must retain
factual knowledge concerning the outcomes of past experiences. In short,
there are obvious similarities between the memory deficits of normal organ-
isms that are asleep and awake organisms with bilateral damage to the hippo-
campus.
Let me summarize the salient findings related to memory in sleep and
also to hippocampal lesions. The following exists for both: (a) short-term
memory is operational; (b) established long-term memory can be accessed; (c)
new memories for declarative knowledge (including certain kinds of condi-
tioning) cannot be formed; (d) new memories for procedural knowledge (sim-
ple conditioning, habituation, etc.) can be formed. While priming effects occur
under certain conditions with amnesics, these effects have not yet been ade-
quately tested in sleep.
These findings, taken together with numerous studies of hippocampal
lesions in various species, suggest that declarative knowledge relates directly
to a functional hippocampus, while procedural knowledge does not. The dis-
tinction between declarative and procedural knowledge characterizes learning
in sleep. In sleep, procedural knowledge is stored, and declarative knowledge
is not. One implication of this analysis is that the hippocampus, in the pres-
ence of an EEG sleep pattern, is not functional for the acquisition of declara-
tive knowledge; that is, it serves its neural gating function of transferring short-
term to long-term memory only in the presence of an EEG waking pattern.
Evidence supporting hippocampal gating that is contingent on the behavioral
state (e.g., awake or asleep) is available in rats (Winson, 1986). Based on his
findings, Winson suggested that neural gating in the hippocampal trisynaptic
circuit is analogous to an open or closed logic gate in that, depending on the
behavioral state, information either flows freely or is almost totally restricted.
Others have also noted that the hippocampal contribution to information flow
is highly state dependent (Hobson & Schmajuk, 1988).
Memories in Sleep 75

The analysis presented here provides guidelines for understanding the

limits of learning in sleep and suggests some interesting theoretical tests and
predictions. The theoretical tests involve the effects of primiilg while asleep on
subsequent responding while awake. They also include the conditions under
which complex conditioning should and should not be demonstrated.
However, the analysis also implies that the range of material that can be
learned in sleep, while interesting and important, is circumscribed. The
broader value of this analysis may be in its focus on brain function in sleep. It
may also bear on the related issue concerning the function of sleep. Various
theorists (Home, 1988) have speculated that sleep is a time for central nervous
system restoration and for memory consolidation of waking experiences. The
latter function may explain why the link between short-term and long-term
memory is severed in sleep, that is, why declarative material presented in sleep
is not learned. It may be that the hippocampus, in addition to providing a
gating function to the long-term memory in waking, also plays the active role
during sleep of consolidating previous waking experiences. Perhaps in sleep,
gating and consolidating cannot occur concurrently. Thus, the gating of new
information and the continued consolidation of old information may be in-
compatible functions related to waking and sleeping, respectively.

References
Aarons, L. ( 1976). Sleep assisted instruction. Psychological Bulletin, 83, 1-40.
Badia, P., Harsh, J., Balkin, T., ORourke, D., & Burton, S. (1985). Behavioral control
of respiration in sleep and sleepiness due to signal-induced sleep fragmentation.
Psychophysiology, 22, 5 17-524.
Beh, H. C., & Barratt, P. E. H. (1965). Discrimination and conditioning during sleep
as indicated by the electroencephalogram. Science, 147, 1470- 147 1 .
Berger, T. W., & Orr, W. B. (1983). Hippocampectomy selectivelydisrupts discrimina-
tion reversal learning of the rabbit nictitating membrane response. Behavioral
Brain Research, 8,49-68.
Bonnet, M. (1982). Performance during sleep. In W. Webb (Ed.), Biological rhythms,
sleep and performance (pp. 205-237). New York: Wiley.
Eich, E. (1 989). Learning during sleep. Paper presented at the Arizona Conference on
Sleep and Cognition, Tucson, AZ.
Emmons, W. H., & Simon, C. W. (1956). The non-recall of material presented during
sleep. American Journal of Psychology, 6, 76-8 1 .
Evans, F. J., Gustafson, L. A., OConnell, D. N., Ome, M. T., & Shor, R. E. (1970).
Verbally induced behavioral responses during sleep. Journal of Nervous and Men-
tal Disease, 148,467-476.
Firth, H. (1973). Habituation during sleep. Psychophysiology, 10,43-5 1.
Hobson, J. A., & Schmajuk, N. A. (1988). Brain state and plasticity: An integration of
the reciprocal interaction model of sleep cycle oscillation with attentional models
of hippocampal function. Archives Italiennes de Biologie, 126, 209-224.
Home, J. A. (1988). Why we sleep. New York Oxford University Press.
Isaacson, R. L. (1982). The limbic system. New York: Plenum Press.
Johnson, L. C., Townsend, R. E., & Wilson, M. R. (1 975). Habituation during sleeping
and waking. Psychophysiology, 12, 574-584.
76 PIETRO BADIA

Koukkou, M., & Lehmann, D. (1968). EEG and memory storage in sleep experiments
with humans. Electroencephalography and Clinical Neurophysiology, 25, 455-
462.
Lehmann, D., & Koukkou, M. (1974). Computer analysis of EEG wakefulness-sleep
patterns during learning of novel and familiar sentences. Electroencephalography
and Clinical Neurophysiology, 37, 73-84.
Maho, C., & Bloch, V. (1983). Acquisition of classical conditioning during REM sleep
in rats. Sleep Research, 12, 160.
McDonald, D. G., & Carpenter, F. A. (1975). Habituation of the orienting response in
sleep. Psychophysiology, 12,618-623.
McDonald, D. H., Schicht, W. W., Frazier, R. E., Shallenberger, H. D., & Edwards,
D. J. (1975). Studies of information processing in sleep. Psychophysiology, 12,
624-629.
Oltman, P. K., Goodenough, D. R., Koulack, D., Maclin, E., Schroeder, H. R., &
Flannagan, M. J. (1977). Short-term memory during Stage 2 sleep. Psychophysiol-
Ogy, 14,439-444.
Rickert, E. J., Lorden, J. F., Dawson, R., Jr., Smyly, E., & Callahan, M. F. (1979).
Stimulus processing and stimulus election in rats with hippocampal lesions. Be-
havioral Neural Biology, 27, 454-465.
Sherry, D. F., & Schacter, D. L. (1987). The evolution of multiple memory systems.
Psychological Review, 94,439-454.
Shimizu, A., Takehashi, H., Sumitsiyi, N., Tanaka, M., Yoshida, I., & Kaneko, Z.
(1977). Memory retention of stimulations during REM and NREM stages of
sleep. Electroencephalography and Clinical Neurophysiology, 43, 658-665.
Simon, C. W., & Emmons, W. H. (1956). Responses to material presented during
various levels of sleep. Journal of Experimental Psychology, 51,89-97.
Squire, L. R. (1987). Memory and brain. New York: Oxford University Press.
Winson, J. (1986). Behaviorally dependent neuronal gating in the hippocampus. In
R. L. Isaacson & K. H. Pribram (Eds.), The hippocampus (Vol. 4,pp. 77-92). New
York Plenum Press.
 CHAPTER 6

BEHAVIORAL RESPONSES
DURING SLEEP
FREDERICK J. EVANS

Cognitive activity during sleep was virtually ignored prior to the discovery of
the link between REM sleep and dreaming (Aserinsky & Kleitman, 1953).
Freud had rejected hypnotic recall and abreaction in favor of the dream as the
royal road to the unconscious. For Freud (1922), the fact that the dream was
a product of sleep was quite incidental; indeed, it was a nuisance given the
difficulty of eliciting dream recall in some patients.
Historically,cognitive activity during sleep was the basis of the first form
of psychotherapy. In the fourth century BC, on the small Turkish island of
Cos, the Aesculapian temples were the site of a systematic method of healing
centered around suggestions of cure and well-being administered by priests to
the sleeping ill. Collison (1988) likened the procedures of the Aesculapians to
hypnosis rather than sleep. Certainly, an association between hypnosis and
sleep has a long tradition. However, the electroencephalograph(EEG) of hyp-
nosis reveals a waking rather than a sleep state (Evans, 1979b). Nevertheless,
this fascination with possible behavioral and subjective similarities between
sleeping and hypnotic behavior along with our conviction that cognitive activ-
ity during sleep is richer than mere dreaming led a group of us to conduct the
studies to be described.

SLEEP-INDUCED BEHAVIORAL RESPONSE

Anecdotal examples support the notion that complex cognitive behavior (in
addition to dreaming) can occur during sleep, particularly if an appropriate
“set” of instructions has been provided before sleep. One rarely falls out of
bed, a remarkable feat considering the amount of bodily activity that occurs
throughout the night. Clear territorial boundaries are worked out with our
77
78 FREDERICK J. EVANS

sleeping partners. Loud noises that are repetitive or familiar are less likely to
wake the sleeping person than soft, strange, novel, or meaningful noises.
Wake/sleep phenomena have been studied in the laboratory. For example,
some people claim that they awaken regularly at a preselected time. Oswald
( 1962)found that, when appropriate waking instructions were given, sleeping
subjects woke up only when specified names of friends were spoken. Condi-
tioned responses and discrimination between auditory stimuli may be elicited
during sleep if a waking response tendency has been established (Beh & Bar-
ratt, 1965; Granda & Hammack, 1961; McDonald, 1966; Weinberg, 1966;
Williams, Morlock, & Morlock, 1966; Zung & Wilson, 1961). Sleep/wake
phenomena, especially sleep learning, has been studied extensively in Russia
and in a few controlled studies in the United States (Simon & Emmons, 1956).
A controversial claim is that hypnoticlike suggestions can be given suc-
cessfully during sleep (Bernheim, 1889; Bertrand, 1826; Fresacher, 1951; Gill
& Brenman, 1959; Schilder & Kauders, 1927). Barber ( 1956)whispered hyp-
noticlike suggestions to 22 subjects who were asleep in their rooms. Some
subjects responded to the suggestions; physiological and EEG criteria were
not used to monitor sleep. Using a sleep/sleep paradigm, Cobb et al. (1965)
documented cognitive activity behaviorally in a pilot EEG study with proce-
dures similar to the one described next.

Procedure
For 2 nights, 19 male student nurses slept in a standard sleep laboratory using
EEG monitoring techniques. During on-line visual diagnosis of alpha-free
REM sleep, suggestions were presented verbally to subjects. Typical sugges-
tions were “Whenever I say the word itch, your nose will feel itchy until you
scratch it”; “Whenever I say the word pillow, your pillow will feel uncomfort-
able until you move it.” The suggestions were tested by saying the cue word
(itch or pillow) once. An attempt was made to test each cue word on at least
two separate occasions during the same REM period in which the suggestion
was given, during all subsequent REM periods that night, and during REM
periods of the second night. The suggestion itself was not readministered on
the second night. Suggestionswere not repeated after their initial presentation.
However, two new suggestions were presented each night whenever possible.
Suggestions were administered and cues were tested only during REM
sleep. At least 120 s of alpha-free REM sleep were required between cue-word
presentations. It was conservatively assumed that visually detected alpha indi-
cates arousal. During the sleep period, the presence of even minimal alpha
(three cycles) was signaled by the technician using a small light outside the line
of sight of the subject. Only subjects who displayed an alpha density exceeding
40% during an eyes closed, waking-rest trial were included in the study.
The subject’s behavior was observed by the experimenter, who was in
the same room. When the subject awakened in the morning, his memory for
the session was tested both directly (during an interview) and indirectly (by
response to cue words in the context of a word association test). Any behav-
Sleep Suggestion 79

ioral or physiological response to the critical cue word was observed. When the
subject awakened after Night 2, a more detailed inquiry was made to evaluate
memory for the sleep events.
The subject was not told before either session that suggestions would be
given, but he was told that sleep cycles were being studied.

Characteristics of Sleep-Induced Behavior

A detailed parametric description of the important characteristics of sleep-
induced behavior is presented elsewhere (Evans, Gustafson, OConnell, Orne,
& Shor, 1969, 1970).

Response Rate
During the 2 nights, 416 cue words were presented during sleep, and 89 correct
responses were observed. On the average, the 19 subjects responded to a mean
of 2 1.2%of all cue words; the highest response rate by a subject was 48%.

Stimulation and EEG Alpha

Uninterrupted REM sleep continued for at least 30 s for 7 1% of all cues ad-
ministered; that is, alpha activity did not occur before or while the cue word
was administered. Many responses were obtained without eliciting alpha ac-
tivity during the suggestion, after the cue words were administered, or before
or after the response.

Specificity of Sleep-Induced Behavior

Once a suggestion was successfully completed (i.e., without eliciting alpha ac-
tivity), it was not repeated in a subsequent REM epoch, nor was it repeated
during Night 2 if it had been administered during Night 1. Cue words were
tested in several subsequent REM periods. There are three possible temporal
sequencesbetween the administration of a suggestion and the subsequent cue-
word testing of it: (a) Immediate testing, with the cue word administered dur-
ing the same REM period as the suggestion; (b) delayed testing, with the cue
word tested the same night but in a REM period following the one in which
the suggestion had been administered; and (c) carry-over testing, with the cue
word tested during Night 2 (after it had been suggested during Night 1). One
day intervened between the administration and testing of these suggestions.
For immediate, delayed, and carry-over conditions, correct responses
respectively followed 20%, 239'0, and 23% of all cues given. Of the total 89
responses observed, 16 occurred during Night 2 in response to suggestions
administered on Night 1. All of these responses were obtained from 6 of the
9 subjects who had been classified as responders based on their Night 1 per-
formance.

Alpha and Response Type

When alpha followed a successful response, it was significantly slower than
waking alpha frequency but was not significantly different from the slowed
80 FREDERICK J. EVANS

frequency that occurs occasionally (but spontaneously) during unstimulated

REM sleep. On those occasions when alpha was elicited to successful re-
sponses, it was significantly slower than waking state alpha (10.0 cps vs. 9.0 1
cps), t(df= 18) = 2 . 1 5 , ~< .05.

Response Latency
A successful response tendency was mobilized slowly. The average response
latency was 32 s. Latency increased as the temporal dissociation between the
administration of the suggestion and the cue word increased. For example,
latencies for immediate and cany-over responses were 19 s and 59 s, respec-
tively (p < .001). In contrast, a similar suggestion given during hypnosis or
tested posthypnoticallyproduces a response latency of only a few seconds.

Subsequent Waking Recall

After the subject awakened, he did not remember the verbally presented mate-
rial, nor could he remember responding. No differencewas found between the
latency of cue word associates to which a response had occurred and the la-
tency of control word associates (.92 s vs. .99 s, respectively). This lack of
recall involved amnesia rather than forgetting, because the material was still
available for future responding during sleep. When the subject returned to
sleep the next night, the mere repetition of the relevant cue word (without
repetition of the suggestion itself) was sometimes sufficient to elicit the appro-
priate response. The behavioral response appeared to be specific to sleep in
spite of the intervening amnesia.
The behavioral response could not be elicited by repeating the cue word
in the waking state. Behavior analogous to posthypnotic suggestion was not
observed.

Long-Term Response Specificity

Six subjects returned to sleep for a 3rd night about 5 months after their initial
participation. Four of these subjects were responders who had shown carry-
over responses on Night 2 to a Night 1 suggestion. Cues for the suggestions
given during the initial 2 nights were presented without administering the sug-
gestion again. The 4 subjects responded consistently to the 5-month-old sug-
gestions, even though there was no intervening waking memory about the
procedure or the specific suggestions.

Replication Study
A subsequent study (Perry, Evans, O’Connell, Orne, & Orne, 1978)confirmed
these characteristics, although a lower response rate ( 14%vs. 18%) and even
longer response latencies were found. Three additional findings confirmed
that the motor behavior during sleep was not random but was directly elicited
by the suggestions and their associated cue words: (a) Correct videotaped re-
sponses were discriminated from random body movement behavior by blind
raters; (b) interspersed dummy cue words not associated with suggestionsthat
Sleep Suggestion 81

had been administered did not elicit behavioral responses appropriate to any
suggestion; and (c) if a cue word was presented before the suggestion had been
given, it did not elicit the specific behavioral response that would later be asso-
ciated with the suggestion.

SLEEP RESPONSE AND RESPONSIVITY TO HYPNOSIS

Part of the original aim of these studies was to investigate the relation between
hypnotizability and sleep-induced response rate. Hypnotizability was assessed
several weeks after the two sleep nights by experimenters who were blind to
the subjects’ rate of responding during sleep. Hypnotic responsiveness was
assessed using the Harvard Group Scale of Hypnotic Susceptibility, Form A
(HGSHS:A; Shor & Orne, 1962)and the Stanford Hypnotic SusceptibilityScale,
Form C ( S H S S : ~ Weitzenhoffer
, & Hilgard, 1962).
Those subjects who responded most frequently to sleepinduced sugges-
tions were more responsive tohypnosis. Pearson correlations between suscep-
tibility to hypnosis and the frequencies of cue administration, response, and
response rate were positive and consistent for the measures, indicating that
sleep-induced response and susceptibilityto hypnosis are related in some man-
ner. The relation is complex, however, and several factors influence the inter-
pretation of these correlations.

Response Frequency
Hypnotizable subjects were found to awaken following stimulation less fre-
quently than insusceptiblesubjects. Consequently, they slept longer, and more
cues were tested. The correlations between response-rate percentage (which
effectively controls the difference in the frequency of cue administrations) and
both HGSHS:A ( r = .42) and SHSS:C ( r = .39) were of borderline significance
(p < .lo, two-tailed).

Hypnosis and Temporal Dissociation of Sleep Response

The correlations between sleep suggestibility and responsivity to hypnosis
were higher for percentage of delayed responses than for percentage of imme-
diate responses. Responsivity to hypnosis more successfully predicted the abil-
ity to respond to sleep-induced suggestion when there was a temporal dissocia-
tion between the administration of the suggestion and the cue word, that is,
when the response was elicited by a cue word during the second night to a
suggestion that had been administered only during the first night. The corre-
lations between the carry-over response rate with HGSHS:A and SHSS:C were
.64 (p < .01) and .60 (p < .Ol), respectively.

Sleep Response and Type of Hypnotic Phenomena

Neither the frequency of cues nor the frequency of responses correlated with
waking suggestibility, hypnotic motor suggestibility, or challenge suggestibil-
ity. This is surprising because the sleep behaviors suggested involved motor
responses. The correlations with the hallucinatory-reverieand the posthypnot-
82 FREDERICK J. EVANS

ic-dissociative clusters of hypnotic behavior tested by HGSHS:A and SHSS:C

were significant for the percentage rate of carry-over (dissociative) responses
(n = .52 and .58, ps < .01) but not for the immediate responses. These two
hypnotic clusters include phenomena experienced only by the few subjects
who can be deeply hypnotized.

Hypnotizability, Sleep-lnduced Response, and Arousal

Subjects who were able to remain asleep while responding to the verbal sugges-
tions administered during sleep were more responsive to hypnosis and slept
more soundly than subjects who did not respond. This joint relation reflects
the hypnotizable responder’s ability to sleep well, both in his verbal claims
about his sleeping habits and in his ability to sleep without awakening in the
experimental situation. By waking more often, both spontaneously and fol-
lowing stimulation, the unresponsive subject provided himself with less op-
portunity to respond because fewer cue word administrations were possible.
The interesting theoretical question is not so much the extent to which sleep
suggestion is analogous to hypnotic phenomena but rather why the hypnotiz-
able subject sleeps better than the more easily aroused insusceptible subject.
Ranking subjects by all available hypnosis scores, we found that the 6 subjects
who were least susceptible to hypnosis accounted for 48% of all awakenings
that occurred during the two experimental nights. In contrast, the 6 subjects
who were most hypnotizable accounted for only 26% of the total awakenings
( p < .01).
Surprisingly,the relation between hypnotizability and response to sleep-
induced suggestions was not significantin the later study by Perry et al. (1978).
Whether differences in the presleep set, the complexity of the stimulus dis-
crimination necessary to process suggestions cognitively, or some other factor
accounts for this puzzling finding is not known.

SLEEP INDUCED RESPONSE AS A

DlSSOClATlVE PROCESS
About 5 months later, 7 of the 19 subjects were invited to sleep another night
in the laboratory. In general, the procedures already described were used, com-
bined with various exploratory methods utilizing hypnosis to test the limits of
sleep responding.
An attempt was made to hypnotize each subject at the beginning of the
sleep session. No special suggestions regarding responding or recall were given
to 5 of the 7 subjects. One of these subjects was allowed to fall asleep while
hypnotized. The remaining 2 subjects were given strong suggestions to re-
spond during sleep and to recall when awake. During sleep, several cue words
were administered that were appropriate to the suggestions of the 2 nights
from 5 months previously. New suggestions were also given. Waking recall
was tested as before, but each subject was hypnotized in an attempt to elicit
further recall. A variety of hypnotic techniques, including regression, was used
Sleep Suggestion 83

with the more hypnotizable subjects in an attempt to reverse the amnesia for
the sleep experiences (Evans, 1979b).
Some subjects responded much more frequently than they had during
the original 2 nights. While the induction of hypnosis and the hypnotic experi-
ence gained since the original nights may have contributed to these increases,
the evidence indicates that neither hypnotic depth nor the interpersonal vari-
ables could account for all of the increases.
Specific presleep suggestions aimed at increasing sleep responsivity may
have been helpful. However, hypnotic and/or set influences did not merely
affect new suggestions. Many responses were obtained to cue words associated
with suggestions that had been neither repeated nor apparently recalled since
the first night 5 months before.
Similar problems of interpretation recurred when attempts were made
to utilize hypnosis, either before or after the sleep session, to obtain recall of
the sleep events. Hypnosis helped, but again the results were not a direct func-
tion of hypnotic depth. Some hitherto unrecalled old suggestionswere recalled
with hypnotic techniques other than regression. This result may indicate that
the techniques originally used to probe morning recall were insufficiently sen-
sitive. On the other hand, the possible timeless or contextless effects of hypno-
sis have been documented by the work on source amnesia (Evans, 1979a; Ev-
ans & Thorn, 1966).

HYPNOSIS AND SLEEP LEARNING

Since the carefully controlled EEG studies by Simon and Emmons (1956),
learning during sleep has not been seriously considered. Simon and Emmons
reported that stimulus material presented during sleep was not recalled later
when the subject awakened unless alpha activity occurred simultaneously with
the stimulus material. Because alpha during sleep indicates arousal, they felt
that any learning occurred in a waking state. In sleep learning studies, the lack
of retention upon awakening has been considered as evidence that registration
or acquisition did not occur during sleep. However, we have shown that reten-
tion (response to cue words) occurs during sleep. The subject not only can
respond to sleep-induced suggestions while remaining asleep but also can re-
spond on subsequent nights without repetition of the suggestion and without
any apparent waking memory of the suggestions or his responses to them. The
problem remaining unsolved is that of the retrieval of sleep-acquired material.
In this sense, the problem of the retrieval of sleep-acquired material is similar
to the problem of the retrieval of dreams when the subject awakens in the
morning.
In Soviet countries, hypnopaedia or sleep learning has been widely prac-
ticed, especially in the 1950sand 1960s. Languages have been taught not only
to individuals during sleep but also, it has been claimed, to whole villages by
radio at night. These studies have been adequately reviewed (Hoskovec, 1966;
Rubin, 1968).The concept of hypnopaedia implies hypnosis, and it is typically
claimed that learning occurs only with “suggestible” subjects. It is not clear
a4 FREDERICK J. EVANS

whether “suggestible” implies hypnotizable or whether a strong waking set

is induced to convince the person that sleep learning is possible. Simon and
Emmons (1956) did not induce such a set.
An attempt was made to maximize the possibility of sleep learning. To
obtain waking recall of sleep-acquiredmaterial, it was felt that four conditions
would have to be fulfilled: (a) subjects must be chosen who could respond
during sleep to sleep-administered suggestion, using the procedures described
previously; (b) some subjects must be included who were highly responsive to
hypnosis; (c) any stimulus presented that was accompanied simultaneously by
alpha could not be included or scored (it was, therefore, not possible to repli-
cate the findings by Simon and Emmons, 1956); acquisition during arousal
was ruled out by eliminating stimuli presented during alpha; and (d) a strong
waking set must establish that sleep learning is possible. The sleep-induced
responses were described, and the subjects were told about their own sleep
induced responding during the first sleep session. Soviet hypnopaedia claims
were reviewed extensively, and subjects were motivated both by the competi-
tive aim to duplicate Russian studies (during the Cold War era of the early
1960s) and by their special qualifications as likely candidates for successful
sleep learning.
Nine subjects were tested in a sleep learning study. They qualified by
responding to a sleep-induced suggestion at least twice while remaining asleep.
As in the earlier studies, these subjects had no subsequent waking recall of
these suggestions. An appropriate presleep set was then established before the
sleep learning session. The subjects slept for about 6 hours. Several subjects
were included who did not receive this set, although the experimenter was
not blind.
Material (“A is for apple”; “P is for palace,” etc.) was presented during
REM sleep and Stages 2 and 4, defined by standard EEG criteria. Waking
recall was tested by asking the subject to check any familiar word on a list of
10 words beginning with the letterA and again from 10 words beginning with
the letter P,and so forth. Eight stimulus words, each beginning with a different
letter, were presented twice to each subject. Where possible, at least two
different letter-word pairs were presented during each stage. After awakening,
the subject received the eight appropriate 10-word lists along with two similar
dummy lists containing letter-word pairs not used during sleep. Thus, the con-
servative probability of checking one correct word by guessing was 0.10 for
each of the eight relevant lists.
Although recall was partial, it exceeded conservatively estimated chance
recall (in fact, none of the dummy list words were ever checked). Of the letter-
word combinations administered during the REM stage, 28% of the adminis-
tered words were correctly checked. In addition, the subject was able to select
with certainty the correct letter (without specifying a word) in an additional
17% of all lists. Only those letters were counted that the subject was quite
certain he heard. In fact, subjects did not guess a letter unless they were con-
vinced it had been spoken. Although words were rarely recalled from Stages 2
Sleep Suggestion 85

and 4, subjects often recognized letters from these stages. The incidence of
guessing, that is, of incorrectly recalled words of letters, was virtually zero. No
control subject (without a presleep set) recalled any words correctly.
A secondary result is important theoretically. Although none of the
words were presented simultaneously with alpha activity, whenever words pre-
sented during the REM stage were subsequently recalled, transient slower fre-
quency alpha (10.25 Hz vs. 9.64 Hz, p < .01) had been evoked within 30 s
after the presentation of the stimuli during sleep.
A third important finding of this study confirms some of the Russian
claims. Of the 9 subjects, 7 had been administered the HGSHS:A and the SHSS:
c . The correlations between the total (all stages) recall of words and the HGSHS:
A and the SHSS:C were .69 and .42, respectively. The respective rank corre-
lations with REM stage recall were .4 1 and .49.
It would seem that under optimal conditions in the laboratory, sleep
learning of relatively easy material can occur with subsequent waking recall.
The practical significance of these results is uncertain. Hypnosis may play a
role in further exploration of the theoretically exciting but practically limited
phenomenon of sleep learning.

SUMMARY
These studies have documented that meaningful behavioral suggestions can
be administered and responded to during sleep, without necessarily evoking
arousal from sleep. Although these sleep-induced responses are based on both
a subjective experience and a related behavioral response, the repertoire of
cognitive activity shown by some subjects is somewhat primitive, as revealed
by the slow response latency or the simplistic word association material suc-
cessfully learned during unintenupted sleep. The sleep-induced behavior was
not remembered using typical recall paradigms in the subsequent waking state
but, rather, remained available for recovery on later nights, even up to 5
months later, sometimes in a dramatic and transformed fashion. This amnesia
has some similarities to another induced amnesia, posthypnotic amnesia.
Posthypnotic amnesia is a temporary failure to utilize temporal sequence cues
as an organizational strategy (Evans & Kihlstrom, 1973). However, posthyp-
notic amnesia differs from the waking state amnesia of the sleep-induced be-
havior in an important way: Posthypnotic amnesia is easily reversed on cue,
but the sleep-induced response to suggestion is not.
The limit of sleep-induced suggestion is not known because there is no
way to evaluate whether the failure to respond is intentional or is simply due
to a failure of stimulus registration. We have no way of knowing if a subject
actually hears the test cue word (or suggestions) or the target words for sleep
learning unless (a) there is arousal, which disqualifies the material, or (b) there
is a response. The response rates reported must be seen as lower bounds.
The presence of the dissociative sleep response behavior is closely related
to other forms of dissociative behavior, especially responsivity to hypnosis.
The highly hypnotizable person is able to fall asleep easily, to take naps (Evans,
86 FREDERICK J. EVANS

1977), to respond to sleep administered suggestionswithout arousal, and to do

some learning during sleep. Evans (1989) referred to an individual difference
dimension of the ability to (volitionally)control cognitive sets, psychological
states, or altered states of consciousness. This dissociative skill is a dimension
that is indicated by a cluster of intercorrelated variables, including hypnotic
responsivity; napping; ease of falling asleep and the ability to control sleep
onset in a variety of circumstances; variable punctuality for appointments;
absorption; phobic, stress-related, and bulimic forms of psychopathology;
hyperemesis in childbirth and cancer; and the speed of recovery from psy-
chiatric and (probably) medical symptomatology regardless of diagnosis
(Evans, 1989).
The 1960s research on the successful response to sleep-induced sugges-
tion and its ramifications for the study of dissociative processes has meaning-
fully influenced later research on hypnotic behavior and the ability to control
states of consciousness. Yet the dramatic nature of individual responses to
sleep-induced suggestionsmay remain the least understood and most difficult
area to study. Little progress can be made in studying the limits of cognitive
processing of external stimuli during sleep until technology provides a method
for measuring arousal during stimulation independently of the presence or
absence of the targeted response. In the meantime, perhaps we should judge
the Aesculapian method of psychotherapy during sleep a success. At least it
deserves a modern test.

References
Aserinsky, E., & Kleitman, N. (1953). Regularly occurring periods of eye modality,
and concomitant phenomena, during sleep. Science, 118,213-309.
Barber, T. X. (1956). Comparison of suggestibility during “light sleep” and hypnosis.
Science, 124, 405.
Beh, H. C., & Barratt, P. E. H. (1965). Discrimination and conditioning during sleep
as indicated by the electroencephalogram. Science, 147, 1470-147 1.
Bernheim, H. (1889). Suggestive therapeutics:A treatise on the nature and uses of hyp-
nosis. New York: Putnam.
Bertrand, A. J. F. ( 1826). Du magnetisme animal en France et des jugement qu’en ont
portes les societes savantes [On animal magnetism in France and the judgements
of the scientific societies]. Paris: Balliere.
Cobb, J. C., Evans, F. J., Gustafson, L. A., OConnell, D. N., Ome, M. T., & Shor,
E. R. ( 1965). Specificmotor response during sleep to sleepadministered meaning-
ful suggestion: An exploratory investigation. Perceptual and Motor Skills, 20,
629-636.
Collison, D. R. (1988, August). A visit to Cos: In search of origins and transitions.
Presidential address at the 1 I th Triennial Congress of Hypnosis and Psychoso-
matic Medicine, The Hague.
Evans, F. J. ( 1977). The subjectivecharacteristics of sleep efficiency. Journal ofAbnor-
ma1 Psychology, 86, 561-564.
Evans, F. J. ( 1979a). Contextual forgetting: Posthypnotic source amnesia. Journal of
Abnormal Psychology, 88,556-563.
Sleep Suggestion 87

Evans, F. J. (1979b). Hypnosis and sleep: Techniques for exploring cognitive activity
during sleep. In E. Fromm & R. E. Shor (Eds.), Hypnosis: Research developments
andperspective (rev. ed., pp. 139- 183). Chicago: Aldine.
Evans, F. J. (1989). The hypnotizable subject, the hypnotizable patient: Studies in dis-
sociative processes. In A. F. Bennett & K. M. McConkey (Eds.), Cognition in
individual and social processes (Vol. 3). Amsterdam, Holland: Elsevier Science
Publishers, B.V.
Evans, F. J., & Kihlstrom, J. F. (1973). Posthypnotic amnesia as disrupted retrieval.
Journal ofAbnorma1 Psychology, 82,317-323.
Evans, F. J., Gustafson, L. A., OConnell, D. N., Ome, M. T., & Shor, R. E. (1969).
Sleep-induced behavioral response: Relationship to susceptibilityto hypnosis and
laboratory sleep patterns. Journal of Nervous and Mental Disease, 148,467-476.
Evans, F. J., Gustafson, L. A., O’Connell, D. N., Ome, M. T., & Shor, R. E. (1970).
Verbally induced behavioral responses during sleep. Journal of Nervous and Men-
talDisease, 150, 171-187.
Evans, F. J., & Thorn, W. A. F. (1966). Two types of posthypnotic amnesia: Recall
amnesia and source amnesia. International Journal of Clinical and Experimental
Hypnosis, 14, 162-179.
Fresacher, L. A. (195 1). A way into the hypnotic state. British Journal ofMedica1 Hyp-
notism, 3, 12-13.
Freud, S. ( 1922). Introductory lectures on psycho-analysis (J. Riviere, Trans.). London:
Allen & Unwin.
Gill, M. M., & Brenman, M. (1959). Hypnosis and related states: Psychoanalytic stud-
ies in regression. New York: International Universities Press.
Granda, A. M., & Hammack, J. T. (1961). Operant behavior during sleep. Science,
133, 1485-1486.
Hoskovec, J. ( 1966).Hypnopedia in the Soviet Union: A critical review of recent major
experiments. International Journal of Clinical and Experimental Hypnosis, 14,
308-3 15.
McDonald, D. G. ( 1966, January). Conditional and unconditional automatic responses
during sleep (Report No. 65-28, Ad. 48 1520). San Diego, CA: U.S. Navy Medical
Neuropsychiatric Research Unit.
Oswald, I. (1962). Sleeping and waking: Physiology and psychology. New York: El-
sevier.
Perry, C. W., Evans, F. J., OConnell, D. N., Ome, M. T., & Orne, E. C. (1978). Behav-
ioral response to verbal stimuli administered and tested during REM sleep: A
further investigation. Waking and Sleeping, 2, 3 17-329.
Rubin, F. (Ed.). (1968). Current research in hypnopedia. New York: Elsevier.
Schilder, P., & Kauders, 0. (1927). Hypnosis. In S. Rothenberg (Trans.), Nervous and
Mental Disease Monograph Series (No. 46).
Shor, R. E., & Orne, E. C. ( 1962). The Harvard Group Scale of Hypnotic Susceptibility,
Form A. Palo Alto, CA: Consulting Psychologists Press.
Simon, C. W., & Emmons, W. H. (1956). EEG, consciousness,and sleep. Science, 124,
1066- 1069.
Weinberg, H. (1966). Evidence suggesting the acquisition of a simple discrimination
during sleep. Canadian Journal of Psychology, 20, 1- 1 1.
Weitzenhoffer, A. M., & Hilgard, E. R. (1962). The Stanford Hypnotic Susceptibility
Scale, Form C . Palo Alto, CA: Consulting Psychologists Press.
Williams, H. L., Morlock, H. C., & Morlock, J. V. (1966). Instrumental behavior dur-
ing sleep. Psychophysiology, 2, 208-2 16.
Zung, W. W. K., & Wilson, W. P. (1961). Response to auditory stimulation during
sleep: Discrimination and arousal as studied with electroencephalography. Ar-
chives of General Psychology, 4, 548-552.
 CHAPTER 7

LEARNING DURING SLEEP

ERIC ElCH

Is it possible for people to register and retain what is said in their presence
while they sleep? If it is possible, is the learning that takes place during sleep
efficient enough to be of practical as well as theoretical significance?These are
the questions of chief concern in this chapter. To address these issues, research
dealing with a number of factors that may have an important influence on
sleep learning is summarized in the second section of the chapter, while in
the third section, some tentative conclusions concerning the possibility and
practicality of learning during sleep are offered, and prospects for future re-
search are outlined. Much of the material covered in both of these sections
has been culled from a remarkably thorough and trenchant review of the sleep
learning literature by Aarons ( 1976), which I strongly recommend to inter-
ested readers.
As will become apparent in the course of this discussion, solid facts about
sleep learning are scarce, and only one of the variables to be considered-the
level of electroencephalographic(EEG) activation that accompanies or follows
the presentation of a to-be-learned or target item-has been researched in a
rigorous manner. Though the present dearth of reliable data is unfortunate, it
is also understandable.For many years followingthe publication ofthe carefully
controlled EEG experimentsby Emmons and Simon ( 1956; Simon & Emmons,
1956), sleep learning was a dead issue. These investigators demonstrated that
verbal information presented during sleep was irretrievable upon awakening
unless presentation coincided with alpha activity, an EEG indicator of arousal
or wakefulness. Their negative results, in combination with a critical commen-
tary (Simon & Emmons, 1955) on the positive results that had been obtained
by others (e.g., Fox & Robbins, 1952; Leuba & Bateman, 1952), caused most
researchers in the United States and other Western countries to abandon the
idea that people may be able to learn while they sleep.
88
Learning During Sleep 89

In more recent times, however, there has been a modest revival of inter-
est in the possibility of sleep learning, owing to three developments. First, a
number of studies have shown that during slow wave (alpha free) sleep, sub-
jects are able to make complex discriminations between repetitive auditory
signals (e.g., Oswald, Taylor, & Treisman, 1960) and to perform, when cued
with appropriate sensory stimuli, motor responses that they had learned while
awake (e.g., Okuma, Nakamura, Hayashi, & Fujimori, 1966). One implica-
tion of these and related results (see Koulack & Goodenough, 1976; Lehmann
& Koukkou, 1974) is that, even during deep sleep, short-term storage of new
information is possible, as is access to old information in long-term memory.
Second, evidence from several sources (see Firth, 1973; Goodenough, 1978)
suggests that habituation or conditioning of various physiological responses,
such as heart rate and galvanic skin response (GSR), can occur during sleep,
albeit at a slower rate than occurs during wakefulness. Because both habitua-
tion and conditioning represent forms of learning, this evidence implies that
the inability to remember information presented during sleep may be attribut-
able not to problems in storing the information but, rather, to a failure to
retrieve the information on waking (Koukkou & Lehmann, 1983; Koulack &
Goodenough, 1976).Third, there have been numerous reports from the Soviet
Union and other Eastern European nations of success in demonstrating sleep
learning (see Hoskovec, 1966; Rubin, 1968, 1971). Though there can be no
doubt that learning is dramatically impaired during sleep (see Goodenough,
1978), these reports recommend a reappraisal of the conclusion that sleep
learning is impossible and raise a number of interesting questions concerning
the conditions under which learning may occur. It is to these conditions that
I now turn.

SLEEP LEARNING: METHODOLOGY AND PHENOMENOLOGY

As Aarons ( 1976) observed, whether or not learning during sleep occurs de-
pends on an intricate interplay of numerous psychological and physiological
factors. In this section, I survey those factors that, in my opinion, have the
most promise of being important moderators of sleep learning. For ease of
exposition, the specific factors to be considered are classified according to four
general types: sleep, item, task, and subject.

Sleep Factors

EEG Activation During and Following /tern Presentation

The research of Simon and Emmons ( 1956)revealed that alpha activity during
the presentation of a target item is a necessary condition for the later recollec-
tion of that item. Evidence also suggests a strong association between memory

This article was prepared with the aid of the National Academy of Sciences (American)
and the Natural Sciences and Engineering Research Council (Canadian), and it profited from the
insightful comments of Jennifer Campbell and Damn Lehman.
90 ERIC ElCH

performance and both the level and the duration of EEG wakefulness or acti-
vation patterns that follow item input. Evidence of this sort has been supplied
by a number of studies (e.g., Jus & Jus, 1972; Koukkou & Lehmann, 1968;
Lehmann & Koukkou, 1974; Oltman et al., 1977), one of which is described
next for the purpose of illustration.
In the study by Koukkou and Lehmann (1 968), short sentences were
auditorily presented to subjects during slow wave (Stage 2 or 3) sleep, and the
duration of the EEG activation (alpha) pattern produced by the presentation
of each sentence was measured. Upon awakening, the subjects completed a
test of nominally noncued or “spontaneous” recall, which was succeeded by
a test of old/new sentence-recognition memory.
The results showed that the duration of EEG activation that followed
the presentation of a given sentence was quite short (M = 9 s) for sentences
that were neither recalled nor recognized, was significantly longer (M = 26 s)
for sentences that were recognized but not recalled, and was longer still (M =
165 s) for sentences that were spontaneously recalled verbatim (Koukkou &
Lehmann, 1968, Table IIB). Thus, the postsleep recollection of sentences pre-
sented during slow wave sleep was systematically related to the duration of
EEG activation that occurred after presentation. (In later work, Lehmann and
Koukkou, 1974, demonstrated an analogous correlation between memory
performance and the level [i.e., EEG wave frequency] of postpresentation acti-
vation.) The fact that intermediate durations of activation were associated
with successful recognition but with unsuccessful recall suggests that recogni-
tion may be a more sensitive measure of memory for sleep-presented material
than is spontaneous recall, a point to which I will return later.
In an effort to provide a theoretical rationale for their results, Koukkou
and Lehmann ( 1968) proposed that the duration (and level; see Lehmann &
Koukkou, 1974)of EEG activation that occurs after the presentation of a tar-
get item reflects the time available for the long-term storage of that item. This
proposal is reminiscent of the consolidation interpretation of sleep-learning
problems put forth by Hebb (1949), who theorized that there are two distinct
forms of memorial representation: a short-term store in the form of reverberat-
ing neural circuits and a long-term store involving the development of more
permanent neural “knobs.” According to this account, it is the transformation
or consolidation of information from a short- to a long-term representation
that is curtailed by the absence of EEG activation.
Several observations are compatible with the consolidation account (see
Goodenough, 1978; Lehmann & Koukkou, 1974). For example, somnambu-
lists can carry out complex motor actions and can respond appropriately to
sensory input during very deep (Stage 4) sleep, but they cannot recall their
actions and responses once they awaken (Jacobson, Lales, Lehmann, & Zwei-
zig, 1965). Also, the apparent accuracy of dream recall is high if sleepers are
awakened during Stage 1 periods of REM sleep (a stage characterized by a
fairly active EEG), but without sleep interruption, dream recall decreases with
increased time spent in slow wave sleep after the end of the REM period (De-
Learning During Sleep 91

ment & Kleitman, 1957). And a numerical stimulus presented during deep
sleep that is not followed by appreciable EEG activation can be recalled if the
subject is intentionally and rapidly awakened before the short-term trace of
the stimulus ceases to exist (Oltman et al., 1977).
Though much of the difficulty in recalling events that take place during
sleep may reflect the impaired consolidation or long-term storage of these
events, the possibility that recall difficulties may be due to deficient retrieval
should not be overlooked. Within the last 20 years, several retrieval-based ac-
counts of sleep-learning problems have been advanced (see Foulkes, 1966;
Goodenough, 1978).One of the more recent of these, the functional state-shift
hypothesis of Koukkou and Lehmann (1983), is framed around the concept
of state-specijic memory: the idea that what has been learned in a particular
state of mind or brain is best remembered in that state (see Eich, 1989; Over-
ton, 1984). According to Koukkou and Lehmann (1983), the forgetting of
events that transpire during sleep (whether internally generated dreams or ex-
ternally presented items) is a function of the magnitude of the difference be-
tween the functional (EEG defined) states in which storage and retrieval of
the events take place. Their hypothesis accords well with a number of diverse
findings, including the aforementioned fact that if a transient period of wake-
fulness (as indicated by an increase in EEG activation) occurs soon after the
presentation of a target item, then the subsequent recall of that item will be
possible during full wakefulness. In addition, the state-shift hypothesis cames
the intriguing implication that information acquired during sleep may be ac-
cessible for retrieval during later occasions of sleep, though not during inter-
vening periods of wakefulness. Evidence pertinent to this implication will be
examined shortly. But first, I would like to make one other point concerning
the correlation between EEG activation and memory performance.
As noted earlier, a number of Soviet and Eastern European studies have
reported success in producing reliable, sometimes robust, sleep learning
effects. In these studies, presentations of the material to be learned are not
regulated according to particular EEG patterns (as is customary in Western
studies) but are timed to correspond with sleep onset, initial sleep, and early
morning sleep, which are optimal times for eliciting EEG activations with al-
pha waves (Aarons, 1976). Thus it is possible, indeed probable, that partici-
pants in these studies are not really asleep when presentation occurs but in-
stead are in a rather drowsy, but nonetheless conscious, state. Is it any wonder,
then, that so-called sleep learning is possible under such circumstances? The
obvious answer is no, but there is more to it than that. Unlike their Western
counterparts, Eastern researchers generally do not find the question of whether
subjects are really asleep during presentation of the learning material to be
important or meaningful. Their primary concern is not with the theoretical
possibility of learning during deep sleep, but with the practical purpose it
serves to present learning material to superficially sleeping subjects. This is
one of several salient differences (others will be discussed in due course) that
distinguishesthe prototypical Western study of sleep learning from the proto-
92 ERIC ElCH

typical Eastern study. As Aarons (1976) argued, these differences probably

account for why Western researchers frequently fail to find evidence of sleep
learning, whereas Eastern investigators often succeed.

Sleep-SpecificMemory
In 19 10, Prince speculated that many people have difficulty remembering
their dreams not because they do not want to remember, as Freud (1900/1953)
and other psychodynamically oriented theorists of the day were claiming, but
rather because they cannot remember, due to the mismatch between the states
of natural sleep and ordinary wakefulness. Intuitively, Prince’s idea seems
plausible, and so does Koukkou and Lehmann’s (1983) hypothesis that fail-
ures of waking memory for experimentally devised materials (such as sen-
tences) presented during sleep are attributable to the shift from sleeping to
waking states. Plausibility is one thing, however, and proof is quite another.
What empirical evidence is there to support the proposition that memory for
events occumng during sleep is specific to the sleep state?
To my knowledge, only one study-described briefly by Evans, Gustaf-
son, OConnell, Orne, and Shor ( 1966) and more elaborately by Evans ( 1972)
and Evans et al. (1969, 1970)-has sought to secure such evidence.
In this study, 18 student nurses slept in a laboratory for 2 or 3 nights.
During the first night, suggestions ( e g , “Whenever I say the word itch, your
nose will feel itchy until you scratch it”) were auditorily presented to subjects
while they were in alpha-free Stage 1 sleep. The suggestions were then tested
immediately by saying a cue word (e.g., itch) and observing the subjects’ be-
havioral responses. Of the 18 subjects tested, 1 1 were able to perform the sug-
gested responses while remaining in Stage 1 sleep. After the subjects awakened,
they did not remember the verbally presented suggestions, nor did they re-
member responding to them. In addition, when presented with the same cue
words that had elicited an appropriate response during sleep, the subjects
did not respond behaviorally when awake. Thus, the subjects appeared to
have a dense waking amnesia for events that had occurred during the prior
night’s sleep.
That the absence of waking memory reflected transient amnesia rather
than permanent forgetting is implied by the observation that, of the 1 1 subjects
who had responded to cue words during the first night, 7 responded to the
same cues during the second night. Thus, in the majority of cases, successful
second-night responding to cue words during sleep occurred even though (a)
the suggestions themselves were not readministered and (b) the subjects had
no intervening waking recollection of the suggestions or of their responses
during sleep.
After an interval of approximately 5 months, 7 subjects were retested on
a third sleep night. None of these subjects remembered the events of either
earlier evening, and 5 of the 7 had responded on both prior nights to the cue
words of the initial night. These 5 subjects responded, while in Stage 1 sleep,
to cue words from the first night’s sleep, despite the fact that the suggestions
had neither been readministered nor recalled in the intervening months.
Learning During Sleep 93

To summarize, the results reported by Evans et al. ( 1966)suggest that at

least some subjects can respond to suggestions for specific motor actions while
they remain in Stage 1 sleep. Further, these responses can be elicited during
Stage 1 sleep of a following night and even in the same sleep stage several
months later, without further reinstatement of the suggestion. This retention
occurs even though the subjects, when awake, are unable to either verbalize
their sleep experiences or to perform the sleep-acquired responses.
As noted earlier, the Evans et al. (1966) experiment is the only one of
which I am aware that directly examined whether memory for events experi-
enced during sleep is specific to the sleep state. Accordingly, their results,
though strongly suggestive of sleep specific memory, should be viewed with
caution. Why no efforts have been made to replicate and extend these results
remains a mystery.
There is one other aspect of the relation between state specificity and
sleep learning that deserves attention, and it concerns the asymmetric form
in which state-specific effects sometimes appear. In several studies involving
alcohol or other depressant drugs (e.g., Eich, Weingartner, Stillman, & Gillin,
1975; Goodwin, Powell, Bremer, Hoine, & Stern, 1969), it has been shown
that, although events encoded in an intoxicated state are difficult to retrieve
under conditions of sobriety, events experienced in the drug-free state tend
not to be state specific and can often be accessed as efficiently in the presence
of alcohol as in its absence. An analogous pattern of results has been obtained
in research involving stimulant drugs, such as nicotine (Peters & McGee,
I982), as well as in experiments entailing alterations of affect or mood. Bartlett
and Santrock ( 1979),for example, found that if preschoolers learned narrative
material while they were feeling especially happy, they remembered more of
this material when tested for recall in a happy than in a neutral mood. How-
ever, stories studied in a neutral affective state were equally well recalled re-
gardless of whether the children were tested in a neutral or a happy mood. The
implication of these and other studies (see Eich, 1989; Overton, 1984) is that
information transfers more completely from an ordinary or typical state of
mind or brain (such as sobriety or neutral affect) to an altered or atypical state
(such as alcohol intoxication or extreme happiness) than it does in the reverse
direction. The main point I want to make now is that asymmetric state-specific
memory may also be involved in sleep. That is to say, while it is evident that
knowledge acquired during wakefulness is expressible during sleep-we do,
after all, tend to dream about things we perceived while awake-events experi-
enced during sleep are difficult, if not impossible,to access during wakefulness.
Why asymmetric state specificity should occur in conjunction with sleep, or
any other experiential state (such as intoxication or happiness) for that matter,
is an open issue. One possible reason relates to the concept of cue overload,
the idea that the effectiveness of a given retrieval cue is inversely related to
the number of discrete events it subserves (Watkins, 1979). Because the vast
majority of our perceptual experiences occur while we are awake, the state of
wakefulnesscannot act as an effective cue for the retrieval of these experiences:
94 ERIC ElCH

It is simply too overloaded. Sleep, in contrast, may constitute a much more

salient or distinctive context for encoding and, thus, may serve as a powerful
cue for the retrieval of events that were encoded in the sleep state. It remains
to be seen whether this reasoning can be developed into a satisfying account
of asymmetric state-specific memory as it occurs in sleep or other experien-
tial states.

Item Factors

Methods of Item Presentation

Though analyses of dream reports indicate that sleep mentation can be reli-
ably and systematically modified by the external presentation of either visual
or tactual stimuli (Arkin & Antrobus, 1978), it is principally through audition
that sleepers maintain contact with the external environment (Aarons, 1976).
For this reason, and in the interest of practicality, audition has been the sen-
sory channel of choice in studies of sleep learning.
Two methods of transmitting auditory information are available to the
sleep-learning researcher: air conduction (loud speaker; e.g., Lehmann &
Koukkou, 1974; Simon & Emmons, 1956) and bone conduction (pillow
speaker; e.g., Bruce, Evans, Fenwick, & Spencer, 1970;Zukhar, Kaplan, Mak-
simov, & Pushkina, 1965/ 1968). Though the former method has been used
more often in past research, the latter may be more conducive to the demon-
stration of sleep-learning effects. As Aarons ( 1976) noted, bone transmits
mainly in the low-frequency range of speech, which includes the fundamental
frequency of the speaker’s voice, and may therefore enhance the fidelity of the
spoken message. Moreover, bone conduction has the curious effect of shifting
the phenomenal source of speech from the outside to the inside of one’s head.
That this may be beneficial for sleep learning is suggested by the idea (Foulkes,
1966) that the extent to which external stimuli are ignored during sleep is
reciprocally related to the sleepers’ preoccupation with their own internal
mentation. Thus, it is possible that sleepers may be more receptive to, and
hence more retentive of, information that seems to originate in their own
minds rather than in the outside world. Whether this possibility is real or re-
mote is a matter that merits serious consideration.

Characteristicsof the Target Items

The list of variables that have a significant impact on the learning of verbal
items in the waking state is extremely long and includes such factors as the
frequency and spacing of item presentations as well as the meaningfulness
and familiarity of the items themselves (see Adams, 1980; Baddeley, 1976).
Unfortunately, the effect that these and other variables have on the efficiency
of verbal learning during sleep is virtually unknown.
In regard to the frequency of item presentation, Simon and Emmons
(1955) asserted that sleep learning, if it is to occur at all, may require that a
massive number of item presentations take place, but they did not offer any
Learning During Sleep 95

clear empirical evidence to back their claim. Bliznitchenko (1968; also cited
in Aarons, 1976), a pioneer in applied Soviet research on sleep learning, ar-
gued that repeated item presentations in the same sequence are a prerequisite
to improvements in learning during sleep, but he too supplied no solid sup-
porting data.
With respect to the spacing of item presentations, an early experiment
by Coyne, which is described in detail by Simon and Emmons (1955), indi-
cated that distributed repetitions of the material to be learned (number-word
pairs) produced better sleep learning than did massed repetitions. However,
Coyne’s results are hopelessly confounded by the fact that the distributed repe-
titions occurred during the period just preceding wakefulness (typically a light,
drowsy state), while the massed repetitions occurred during deeper and possi-
bly less receptive stages of sleep (see Simon & Emmons, 1955).
The conflicting results revealed by several studies involving nonsense
syllables, common words, simple sentences, and even Chinese-English paired
associates prompted Simon and Emmons ( 1955)to conclude that meaningful-
ness is not a critical determinant of sleep learning. However, as Aarons ( 1976)
commented, in no sleep learning study have the semantic or denotative di-
mensions of the learning material been systematically manipulated. Aarons
also remarked that, apart from whatever role that semantic meaningfulness
may play in sleep learning, the personal meaningfulness or affective signifi-
cance of the learning material may be important. Indeed, hand movements
and electrographic (K complex) responses occur more frequently when sub-
jects are presented, during deep sleep, with their own rather than with some-
one else’s name (Oswald et al., 1960), and emotionally toned words, in con-
trast to neutral items (e.g., dumb, sin vs. drum, sit), provoke more pronounced
eye movements and cardiovascular changes when presented during sleep than
during wakefulness (Minard, Loiselle, Ingledue, & Duatlich, 1968).Given that
personally or affectively meaningful material is more apt to be registered dur-
ing sleep, it would be interesting to know whether such material is also more
likely to be retained upon awakening.
Earlier I pointed out that a salient difference between Western and East-
ern studies of sleep learning is that in the former, material is presented only
during EEG-defined sleep, whereas in the latter, presentation occurs at the
beginning and the end of the normal sleep cycle. Another significant difference
concerns the learner’s familiarity with the material. In most Western studies,
subjects do not know what it is they will hear while they sleep, and they usually
participate in only a single sleep-learningsession. In contrast, Eastern investi-
gators have developed a “hypnopaedic tutorial system” (see Rubin, 1970) in
which the presentation of material during superficial sleep is coordinated with
ongoing audiovisual presleep and postsleep instruction that lasts for several
weeks. Though it has been claimed that this system acceleratesthe learning of
telegraphy, foreign-language vocabulary, and other types of practical knowl-
edge (see Aarons, 1976; Bliznitchenko, 1968; Rubin, 1971), the absence of
appropriate controls makes it impossible to determine how much of the learn-
96 ERIC ElCH

ing is attributable to waking instruction alone. Still, the possibility exists that
the presentation of learning material during sleep improves performance in
the waking state, provided that the material is familiar to the learners prior to
its presentation.
Preliminary support for this possibility has been provided by Tilley
( 1979). Subjects in his experiment examined a set of 20 pictures of common
objects at bedtime. Later that night, a tape-recorded list of 10 words-the
names of half of the pictures in the original set-was presented 10 times during
either Stage 2 or REM sleep. The following morning, the subjects were tested
for free recall and recognition of the complete set of 20 picture names.
In comparison with items that had been studied at bedtime only, those
that had been presented both before and during sleep were significantly better
recalled and recognized. Curiously, the beneficial effect of repetition during
sleep was much more evident in the morning retention of items that had been
repeated during Stage 2 sleep than those that had been repeated during REM.
This finding is curious in that REM, which is characterized by low-amplitude
EEG activity and the periodic appearance of alpha frequencies, would seem
to be more conducive to the processing of incoming information than would
non-REM (in this case, Stage 2) sleep. At any rate, Tilley’s (1979) results are
clearly consistent with the Soviet claim that presleep learning can be strength-
ened or reinforced through within-sleep repetition. It is equally clear, however,
that Tilley’s results need to be replicated and, if possible, extended to other
types of learning materials and retention tasks.

Task Factors

Recall Versus Recognition

Although a few studies have used savings in relearning to assess the retention
of sleep-presentedmaterials (often with contradictory outcomes; compare, for
instance, the positive results obtained by Fox and Robbins, 1952, with the
negative findings of Bruce et al., 1970), most have employed tests of recall,
recognition, or both (see Aarons, 1976, Table 2).
Recognition may be a more sensitive measure of memory than recall for
events experienced during sleep. As noted earlier, Koukkou and Lehmann
( 1968; see also Lehmann & Koukkou, 1974)found that items whose presenta-
tion during slow wave sleep was followed by an intermediate duration and
level of EEG activation were subsequently recognized but were not spontane-
ously recalled. In addition, Levy, Collidge, and Stabb (1972) observed that,
although subjects were unable to recall Russian-English paired associates that
had been presented during either Stage 1 or Stage 4 sleep, recognition of the
response or target words was slightly but significantly above chance for both
sleep stages. And a study by Johnson, Kahan, and Raye (1984) showed that
the probability of recalling a dream that had been immediately reported upon
awakening fell to below .20 after a 2-week retention interval, whereas the prob-
ability of recognizing the dream was an impressive .80 after the same interval.
Learning During Sleep 97

Why events experienced during sleep should be more readily recognized

than recalled remains to be determined. However, basic memory research in-
volving more conventional materials suggests that how readily information is
comprehended and organized may matter more for recall than for recognition
(see Kintsch, 1970). Conceivably, then, events that occur during sleep may be
particularly difficult to recall either because sleep is not conducive to the co-
herent organization of ongoing events or because the events themselves are
incomprehensible (as often appears to be the case for dreams).

Implicit Versus Explicit Memory

Evidence from several sources suggests that memory for past events can influ-
ence present actions even if one is not aware of remembering the earlier experi-
ences. As an example, prior presentation of a word makes it more likely that
college students can report that word when later it is briefly exposed in a per-
ceptual identification task, regardless of whether or not they recognize the
word as one that had been presented before (Jacoby & Dallas, 1981). Similarly,
amnesic patients reveal the effects of practice in their subsequent performance
of a cognitive, perceptual, or motor skill, even though they cannot remember
ever having practiced that skill (Schacter & Tulving, 1982). These and other
observations suggest that it is possible to distinguish the implicit effects of
memory for prior episodes or experiences on a person’s current behavior from
the person’s explicit awareness that he or she is remembering events of the
past (Richardson-Klavehn & Bjork, 1988; Schacter, 1987; Schacter & Kihl-
strom, 1989).
The point I wish to make is that it may be useful to apply the distinction
between implicit and explicit memory to the question of whether events that
occur during sleep can be registered and retained. As noted previously, most
earlier experiments examining this question have focused on the individual’s
ability to recall or recognize a specific item (e.g., a spoken number, word, or
sentence) as having occurred in a specific situation, namely, while the individ-
ual was superficially or soundly asleep. Memory as measured in these experi-
ments is explicit in that the person must necessarily be aware that he or she is
remembering a particular past event. Given that most people appear pro-
foundly amnesic when tested for the deliberate recall or recognition of events
to which they had been exposed while asleep, it is reasonable to infer that
events that are denied conscious attention are ordinarily not amenable to con-
scious reflection or accessible through explicit forms of remembering. The in-
ference need not be drawn, however, that events occurring during sleep leave
no lasting impression in memory and exert no enduring effect on behavior.
The possibility remains that even though the effects of memory for sleep-expe-
rienced events may not be revealed via tests of recall or even of recognition,
such effects might become evident in situations that do not demand explicit
recollection.
How might this possibility be explored? One way of doing so is suggested
by Jacoby and Witherspoon ( 1982). Participants in their study were 5 univer-
98 ERIC ElCH

sity students and 5 Korsakoff alcoholics, clinically diagnosed as amnesic. In

the first phase of the study, subjects answered questions such as: “Name a
musical instrument that employs a reed.” As implied by this example, the
intent of the question-answering task was to encourage the subjects to encode
homophones, such as reed, in relation to their low-frequency or less common
interpretations.
In the second phase of the study, the subjects were read a list that con-
sisted in part of equivalent numbers of old and new homophones (i.e., words
that either had or had not appeared in the context of biasing questions), and
were asked to spell each word aloud. Jacoby and Witherspoon (1982) reasoned
that if the first presentation of a “biased” homophone is remembered and
influences its later interpretation, more of the old than the new homophones
should be spelled in keeping with their less common interpretations. They
further reasoned that an influence of memory on the spelling of a word would
not necessarily require the subjects to be aware that they were remembering
the first presentation of that word but rather that such explicit recollection
would be required in the test of old/new recognition memory that was given
in the third and final phase. To the extent, then, that the processes or systems
underlying implicit and explicit memory are distinct, performance on a test
(viz., recognition) that requires awareness of earlier events and performance
on a test (viz., spelling) that does not demand explicit recollection should be
independent of one another, in that performance on one type of test should
not be predictable on the basis of performance on the other type.
Evidence of such independence was revealed in two ways. First, in com-
parison with the students, the amnesic patients recognized far fewer old homo-
phones (25%vs. 76%, with neither group of subjects generating any false posi-
tives) but spelled more of these items in line with their less common,
experimentally biased interpretations (63%vs. 49%,with approximately 2 1%
of the new homophones spelled in their low frequency forms by both groups).
Thus, the disadvantage of the amnesic patients was restricted to recognition
memory, an explicit form of remembering. Second, for both patients and stu-
dents, the conditional probability of recognizing an old homophone in the
third phase of the study, given that its spelling had been biased by memory in
the second phase, p(recognition I spelling), did not differ significantly from the
unconditional probability of correct recognition, p(recognition). Thus, for
neither type of subject was recognition memory associated with or enhanced
by an effect of memory on spelling.
In short, it seems that the prior presentation of a word has a substantial
impact on its subsequent interpretation and spelling, regardless of whether or
not the word is correctly classified as old in a later test of recognition memory.
Though this dissociation of spelling and recognition performance was espe-
cially striking among the amnesics (who outperformed the students on the
former test and were themselves outperformed by the students on the latter),
it was demonstrated by the nonamnesics as well, as evidenced by the students’
statistically equivalent values of p(recognition I spelling) and p(recognition).
Learning During Sleep 99

Recognition and spelling thus seem to reflect different forms or functions of

memory, whether intact or organically impaired. Whereas recognition of an
old word requires the recognizer to be aware of its prior presentation, an in-
fluence of memory on the spelling of a word does not necessarily demand
explicit remembering (see Eich, 1984;Jacoby, 1982).
Approached from the standpoint of sleep learning, the idea that recogni-
tion and spelling tap different memory processes or systems raises an interest-
ing question for research. Specifically, suppose that during sleep, subjects are
presented with a series of short, descriptive phrases, each consisting of a homo-
phone and one or two words that bias the homophone’s less common interpre-
tation (e.g., WUY and PEACE; deep SEA). Suppose further that, upon awakening,
the subjects are read a list composed chiefly of old and new homophones on
two successive occasions. On one occasion, the subjects are simply asked to
spell each list item aloud; on the other occasion, the subjects are asked to state
aloud which list items they recognize as having been presented during sleep.
Might the subjects spell significantly more old than new homophones in line
with their less common interpretations and yet fail to discriminate reliably
between the two types of items in the test of recognition memory-a test,
unlike that of spelling, which would presumably require the subject to con-
sciously reflect upon events that had occurred during a state of unconscious-
ness? More broadly stated, is it possible that people implicitly know something
about events that take place during sleep but do not explicitly know that they
know? The answer to this question might be of interest from an applied as well
as a theoretical perspective.

Subject factors

Age
Several authors have speculated about whether the ability to learn during sleep
is dependent on age, but no one has provided any telling data. Svyadoshch
(1962/1968), for example, employed subjects ranging in age from 10 to 60
years in a series of studies concerning the reproduction of stories presented
during sleep. Though Svyadoshch asserted that the majority of his subjects,
regardless of their age, demonstrated a high level of text reproduction (arbi-
trarily defined as 66% or more of the story material), he did not provide a
breakdown of reproduction scores by age group. Svyadoshch also offered no
hard numbers to support a second, seemingly contradictory assertion concern-
ing the relation between sleep learning and age: specifically, that the ability to
assimilate speech during sleep can be acquired artificially by means of sugges-
tions delivered in the context of either deep hypnosis or ordinary wakefulness
and that children and adolescents,being more suggestible by nature than older
adults, are especially adept at developing sleep-learning abilities.
Interestingly, the idea that an optimal period for learning how to learn
during sleep may arise at an early age also occurred to Aarons ( 1976), but for
different and more defensible reasons. These include the observation that (a)
100 ERIC ElCH

even as early as 3 days after birth, the human voice and its fundamental fre-
quency are more effective than other sounds in eliciting behavioral and physi-
ological reactions during alert, relaxed, and somnolent states (Hutt, Hutt, Len-
ard, Bernuth, & Muntjewerff, 1968); (b) children appear to acquire second
languages more readily than do adults, which suggests a greater facility in pho-
netic processing during wakefulness that could conceivably carry over to sleep;
and (c) in comparison with older children, younger children devote more at-
tention to and are more likely to remember auditorily rather than visually
presented information (Hallahan, Kauffman, & Ball, 1974). Though the fore-
going observations are compatible with the developmental hypothesis ad-
vanced by Aarons (1976), more direct evidence is needed.

Health
Given that (a) between 5% and 10% of otherwise healthy medical students
suffer from chronic sleep disturbances that range from mild to moderate in
severity (Johns, Gay, Goodyear, & Masterton, 197I), (b)’emotionalstress dis-
rupts the natural sleep cycles of men and women alike (Breger, Hunter, &
Lane, 197I), and (c) both mentation and physiological processes during sleep
are influenced by menstruation in women (Sheldrake & Cormack, 1974), the
need to screen subjects for sleeplearning research on the basis of specific cri-
teria related to their physical and psychological health seems clear. Yet apart
from the research of Zukhar’ et al. (196511968)in which people with histories
of sleep disturbance were specifically excluded from participation, health-re-
lated variables have not been taken into account in prior studies of sleep learn-
ing. Instead, researchers have simply assumed that their subjects are in gener-
ally good health and have normal hearing. As Aarons (1976) remarked,
information on personal health and sleep habits would help investigators de-
termine the suitability of a particular person to a particular sleep-learning in-
tervention, and it is therefore hoped that the gathering of such information
will become a standard practice in future studies of sleep learning.

Capacity to Learn WhileAwake

According to Simon and Emmons (1955), sleep-learningresearchers would be
well-advised to select as their subjects people who are particularly proficient
at learning in the waking state, since the effect of presenting material during
sleep may be so subtle that its benefits will be evident only in highly intelligent
individuals. In consideration of Simon and Emmon’s (1955) conjecture, four
points are worth making. First, there is no a priori reason to assume that gen-
eral intelligence plays a more prominent role in learning during sleep than it
does in wake instruction (Aarons, 1976). Second, it seems plausible that posi-
tive effects of sleep learning might be more readily demonstrated in individuals
who are deficient rather than proficient in wake-state acquisition, in much the
same manner as the memory-enhancing effects of nootropic drugs, such as
oxiracetam (Itil, Menon, Bozak, & Sangor, 1982), may be more readily dem-
onstrated in memory-impaired patients (e.g., those with senile dementia of
Learning During Sleep 101

the Alzheimer’s type) than in cognitively intact controls. Third, there is no

empirical evidence to support Simon and Emmon’s (1955) position, and
fourth, what little evidence does exist runs counter to Simon and Emmon’s
conjecture. The pertinent evidence comes from an early experiment by Elliott
(1947/1968). All 40 of the subjects in Elliott’s study first learned one list of
words (List A) to criterion in the waking state. Subsequently, a second list (List
B) was presented to 20 subjects while they slept (the experimental group) but
not to the other 20 subjects (the control group). The following morning, all 40
subjects learned List B to criterion. The key finding was that the percentage of
savings in learning List B, that is, S = ( A - B/A) X 100, where A and B desig-
nate the number of trials required by a given subject to learn Lists A and B
and S designates percentage of savings, was significantly greater for experi-
mental than for control subjects, a finding that Elliott interpreted as evidence
of sleep learning. For the present purpose, a more interesting finding concerns
the correlation between the values of A and S for each group of subjects. For
purely statistical reasons, one would expect to observe a positive correlation
between these measures for either group (because if any 2 subjects take the
same number oftrials to learn List B, the one who required more trials to learn
List A will necessarily obtain a higher savings score). However, if Simon and
Emmon’s ( 1955)speculation that the benefits of sleep learning are more likely
to be detected in good than in poor wake-state learners, then one would also
expect to find a smaller correlation between the A and Sscores of experimental
subjects than between those of control subjects. That is to say, good learners
in the experimental group (those who required relatively few trials to master
List A) ought to show more savings in their learning of List B than should
poor learners in the same group (those who required relatively many trials to
learn List A). In fact, the correlation between A and S was somewhat greater
among the experimental subjects (r = .37) than it was among the control sub-
jects (r = .21). (These correlations were calculated from the data presented in
Table I11 of Elliott, 1968,p. 13.) Thus, the advantage of having been presented
with List B during sleep on the later learning of that list appears to have ac-
crued more to the poor than to the good wake-state learners, the opposite of
what would have been anticipated on the basis of Simon and Emmon’s ac-
count. I emphasize the word appears because Elliott’s experiment was not free
of methodological flaws (e.g., he did not continuously monitor sleep using
EEG; see Simon & Emmons, 1955). More rigorous research will need to be
performed before the relation between learning capacitiesin waking and sleep
ing states can be stated with any degree of precision.

Suggestibility:Hypnotic Susceptibility and Learning Set

By several Soviet accounts (e.g., Svyadoshch, 1962/ 1968; Zavalova, Zukhar,
& Petrov, 1964/ 1968;see also Aarons, 1976;Hoskovec, 1966),learning during
sleep is possible provided that the learners are suggestible. As a rule, however,
Russian researchers have been neither clear nor consistent in their usage of
the term suggestible: At times the term appears to imply susceptibilityto hyp-
102 ERIC ElCH

nosis, at other times it refers to a strong waking set that is induced in the
subjects to convince them that sleep learning is a bona fide phenomenon, and
on still other occasions it connotes both of these senses. In addition, the evi-
dence these researchers have presented to support their position cannot be
regarded as compelling.
Consider, for example, the work of Kulikov (1 964/1968). Subjects in his
studies included 21 grade school and 15 college students, all of whom were
highly susceptibleto hypnosis (as tested by the method of hand gripping). The
subjects were separated into three groups of 12, each composed of 7 children
and 5 adults; random assignment of subjects to groups was not specifically
noted by the author.
Subjectsin the first group were repeatedly presented during natural sleep
with a narrative (a Tolstoy story for the children; a description of nervous
system functions for the adults) and were tested for recall of the text when they
awoke. These subjects were not, as Kulikov (1964/1968) put it, “prepared”
for sleep learning; that is, they had received no specific suggestions for assimi-
lation and retention of the text prior to its presentation. Kulikov did not spec-
ify the number of times the text was presented, the precise form of the recall
test (i.e., whether it was spontaneous or prompted), or the duration of the
retention interval. Further, it is not clear from Kulikov’s account exactly when
the text was presented; on procedure, he remarked only that the text was pre-
sented, via tape recorder, at a volume that was below the threshold of hearing
in the waking state and that sleep was monitored by activity records (assessing
the absence of motor movements) and pneumographic tracings (assessing the
absence of marked respiratory reaction). Be that as it may, Kulikov found that
only 1 of the 12 subjects in this group had any waking recollection of the text,
and the 1 exceptional subject was a boy who had taken part in previous studies
in which hypnopaedic suggestions had been delivered.
Kulikov began testing of subjects in the second group by establishing
contact with them while they slept. After the subjects had been sleeping for 1
or 2 hours, tape-recorded suggestions were presented (e.g., “You are sleeping
peacefully, do not wake up” and “Your breathing is becoming deeper and
deeper”). Having made contact with the sleeping subjects in this manner, the
suggestion was given, “Now you will hear a story, listen to what is said, try
and memorize it as much as possible, you will remember this all your life, and
whenever wanted you will be able to relate it.” The text was then presented
(an unspecified number of times) and was followed by additional suggestions
to remember the text and to sleep soundly.
The impact these suggestions had on the subjects’ waking-recall perfor-
mance appears to have been profound. Among the 12 subjects in the second
group who had been prepared with a suggested set to learn while asleep, the
percentage of idea units contained in the text that were recalled averaged 64%
and ranged from 47% to 87%;there was no appreciable difference in the per-
formance of the children and the adults. These figures are remarkably similar
to those yielded by the third group of subjects, who were awake at the time of
text presentation (mean recall of 66%, range from 44% to 80%).
Learning During Sleep 103

Taken at face value, Kulikov’s (1964/1968) data indicate that learning

during sleep is possible in hypnotically suggestible subjects when a suggested
set to register and retain the learning material is involved. Moreover, his data
suggest that the capacity to learn during sleep is comparable to that of the
waking state.
Kulikov’s ( 1964/ 1968)results are not beyond reproach, however. First,
the suggestions (“preparations”) that were imparted to subjects in the second
group were evidently not imparted to subjects in the third group, thus preclud-
ing a valid comparison of the effectivenessof sleep versus wake learning. Sec-
ond, it is possible that the striking difference in recall performance found be-
tween the first and second groups does not demonstrate the importance of
preparing subjects for sleep learning but rather reflects the fact that only the
second group of subjects received any suggestions at all. A more meaningful
contrast would have been between groups receiving suggestions that either
were or were not relevant to the specific learning task at hand.
Though Kulikov’s ( 1964/ 1968)studies have some serious shortcomings,
his contention that sleep learning is possible in hypnotically susceptible sub-
jects who have acquired an appropriate set to learn, which is shared with other
Soviet researchers (see Hoskovec, 1966), finds support in a small study by
Evans (1972), an American-based investigator. Nine of the subjects in Evans’
experiment were people of varying levels of hypnotic susceptibilitywho could
respond while they remained asleep to suggestions for specific motor actions
(e.g., “Whenever I say the word pillow, your pillow will feel uncomfortable
until you move it”) without a presleep “set” to perform these actions; none of
the subjects had any waking recollection of the suggestions. A strong waking
set was then instilled that sleep learning is possible. Specifically, subjects were
told that, unlike most people, they were able to respond to suggestions pre-
sented during sleep and that this made them particularly promising candidates
for sleep learning. Further, subjects were informed about successful Soviet
demonstrations of sleep learning and thus were motivated both by their own
special qualifications and by the competitive aim to duplicate the Russian
results. In addition to these 9 subjects, several others were included who did
not receive the suggested set.
Material of the form “A is for Apple, P is for Palace,” was presented to
the subjects during REM sleep and Stages 2 and 4. Any letter-word pair whose
presentation was accompanied simultaneously by alpha was excluded from
subsequent analyses of retention. Eight target words, each beginning with a
different letter, were presented twice to each subject; at least two words were
presented during each sleep stage.
Waking retention was tested by having subjects check any familiar word
on a list of 10 words beginning with the letter A and again from 10 words
beginning with the letter P, and so forth; two similar “dummy” lists, contain-
ing words that had not been presented during sleep, were also administered.
Thus, the conservative probability of recognizing a target word by guessing
was .10 for each of the eight relevant lists.
104 ERIC ElCH

Three main findings emerged from the recognition test. First, subjects
who had not received the set to learn during sleep recognized none of the
target words from any sleep stage. Second, subjects who had received the set
recognized, on the average, .28, .lo, and .OO of the words that had been pre-
sented during Stages REM, 2, and 4, respectively; none of these subjects ever
claimed to recognize a word that was not a true target. Thus, only those words
that had been presented to suggested-set subjects during REM sleep were rec-
ognized at a better-than-chance level. Third, among the suggested-set subjects,
those who had a relatively high level of hypnotic susceptibility (as indexed by
the Stanford Hypnotic SusceptibilityScale, among other instruments) tended
to recognize more REM stage targets than did subjects who had a relatively
low level (r = .49).
Viewed as a whole, the results of Evans’ (1 972) experiment seem to
square with the Soviet position that sleep learning is possible in hypnotically
susceptible subjects in whom a strong set to learn has been established. As
such, Evans’ results illuminate a number of interesting issues for future re-
search. Consider first the concept of a suggested set. Intuitively, it seems rea-
sonable to suppose that the induction of a set will increase the subjects’ moti-
vation to learn while they sleep. If motivation is indeed one of the keys to
successful sleep learning, then the odds of observing significant sleep-learning
effects should be improved by offering subjects a substantial monetary reward
for good retention performance (e.g., Levy et al., 1972), by ensuring that the
material to be learned during sleep is pertinent to the subjects’ personal needs
or educational goals (e.g., Balkhasov, 1965/1968),or by restricting the subject
sample to individuals who have a strong interest in the research (e.g., Svya-
doshch, 1962/ 1968).
Examination of the role of hypnotic susceptibility in sleep learning, (re-
viewed by Hilgard, 1979)indicates that high hypnotizables are able to process
information outside of conscious awareness more effectively and completely
than are low hypnotizables. A striking example of this “splitting” of conscious-
ness, a process termed dissociation,is shown when a person discovers that he
or she is reacting, in an apparently automatic or involuntary manner, to a
suggestion implanted previously under hypnosis. Owing to their greater dis-
sociative abilities, high hypnotizables may be able to attend and process in-
coming information selectively, without consciousness awareness, after they
have fallen asleep. Lacking this ability, low hypnotizables have to awaken to
process similar information and are therefore incapable of learning while they
sleep. Though this hypothesis is as speculative as it is sketchy, it does seem to
fit with the findings that, in comparison with low hypnotizables, high hypno-
tizables are (a) less likely to awaken either spontaneously or following verbal
stimulation (Evans, 1972), (b) more likely to respond to behavioral sugges-
tions administered during sleep (Evans et al., 1966, 1969), and (c) more adept
at changing their dream experiencesto conform with specific presleep instruc-
tions (Belicki & Bowers, 1982). These findings, in addition to the others men-
tioned earlier in this section, suggest that the relations among hypnotizability,
Learning During Sleep 105

dissociability, and sleep learning comprise an inviting target for future re-
search.

CONCLUSlON
Whether it is possible and practical for people to learn while they sleep is a
question to which Western and Eastern researchers have given different an-
swers. Little, if any, learning has been revealed in most Western studies,
wherein novel verbal material has been presented to randomly selected sub-
jects during a single session of EEG-defined sleep. The learning that has mate-
rialized in these studies has frequently been found to be correlated with both
the duration and the level of EEG wakefulness patterns that coincide with or
closely follow presentation of the learning material. In contrast, evidence of
substantial sleep learning has emerged in numerous Eastern studies, wherein
familiar material was presented to suggestible subjects who had received a
strong presleep instruction set to learn and were willing to participate in a
lengthy training regimen. No attempt has been made in these studies to input
information during deep stages of sleep; instead, presentations have been
timed to correspond with sleep onset, initial sleep, and early morning sleep,
periods in which significant EEG activations are likely to occur. Any improve-
ments in performance obtained under these conditions would thus appear to
reflect a composite of wake and sleep experience and not pure, unadulterated
“sleep learning.”
Although it appears clear that information presented during sleep that is
not accompanied by EEG activation is not retained upon awakening, it would
be interesting to know, for theoretical as well as for applied reasons, whether
there is any substance to the Soviet claim that substantial improvements in
learning can be achieved by a systematic program of combined wake/sleep
instruction. It would also be informative to discover whether such improve-
ments are dependent (a) on the learner’s age, health, capacity to acquire
knowledge in the waking state, susceptibility to hypnosis, and motivation to
learn; (b) on the nature of the learning materials (e.g., whether they are affec-
tively intoned or personally insignificant)and the methods of material presen-
tation (e.g., air- vs. bone-conducted transmission); and (c) on the means by
which retention of the material is measured (e.g., whether the test taps implicit
vs. explicit memory). These are among the many issues that remain to be
settled in future research aimed at investigating both the possibility and the
practicality of learning during sleep.

References
Aarons, L. ( 1976). Sleepassisted instruction. Psychological Bulletin, 83, 1-40.
Adams, J. A. (1980). Learning and memory: An introduction. Homewood, IL: Dorsey
Press.
106 ERIC ElCH

Arkin, A. M., & Antrobus, J. S. ( 1978). The effects of external stimuli applied prior to
and during sleep on sleep experiences. In A. M. Arkin, J. S. Antrobus, & S. J.
Ellman (Eds.), The mind in sleep: Psychology and psychophysiology (pp. 351-
39 1). Hillsdale, NJ: Erlbaum.
Baddeley, A. D. ( 1976). The psychology of memory. New York: Basic Books.
Balkhasov, I. (1968). The rapid teaching of a foreign language by lessons heard during
sleep. In F. Rubin (Ed.), Current research in hypnopaedia (A. Scott, Trans.) (pp.
160-163). New York: Elsevier. (Originally published in 1965)
Bartlett, J. C., & Santrock, J. W. (1979). Affect-dependent episodic memory in young
children. Child Development, 50, 5 13-5 18.
Belicki, K., & Bowers, P. (1982). The role of demand characteristics and hypnotic
ability in dream change following a presleep instruction. Journal of Abnormal
Psychology, 91,426-432.
Bliznitchenko, L. ( 1968). Hypnopaedia and its practice in the USSR. In F. Rubin (Ed.),
Current research in hypnopaedia (pp. 202-209). New York: Elsevier.
Breger, L., Hunter, I., & Lane, R. W. (1971). Theeffectofstress on dreams. New York:
International Universities Press.
Bruce, D. J., Evans, C. R., Fenwick, P. B. C., & Spencer, V. ( 1970). Effects of presenting
novel verbal material during slow-wave sleep. Nature, 225, 873-874.
Dement, W. C., & Meitman, N. (1957). The relation of eye movements during sleep
to dream activity: An objective method for the study of dreaming. Journal of Ex-
perimental Psychology, 53, 339-346.
Eich, E. (1984). Memory for unattended events: Remembering with and without
awareness. Memory and Cognition, 12, 105- 1 1 I .
Eich, E. (1989). Theoretical issues in state dependent memory. In H. L. Roediger 111
& F. I. M. Craik (Eds.), Varieties of memory and consciousness:Essays in honor
of Endel Tulving (pp. 33 1-354). Hillsdale, NJ: Erlbaum.
Eich, J. E., Weingartner, H., Stillman, R. C., & Gillin, J. C. (1975). State-dependent
accessibility of retrieval cues in the retention of a categorized list. Journal of Verbal
Learning and Verbal Behavior, 14,408-4 17.
Elliott, C. R. (1968). Extracts from an experimental study of the retention of auditory
material presented during sleep. In F. Rubin (Ed.), Current research in hypnopae-
dia (pp. 6-27). New York: Elsevier. (Original work published 1947)
Emmons, W. H., & Simon, C. W. (1956). The non-recall of material presented during
sleep. American Journal of Psychology, 69, 76-8 1.
Evans, F. J. (1972). Hypnosis and sleep: Techniques for exploring cognitive activity
during sleep. In E. Fromm & R. E. Shor (Eds.), Hypnosis: Research developments
andperspectives (pp. 43-83). Chicago: AldineJAtherton.
Evans, F. J., Gustafson, L. A., OConnell, D. N., Orne, M. T., & Shor, R. E. (1966).
Response during sleep with intervening amnesia. Science, 152, 666-667.
Evans, F. J., Gustafson, L. A., OConnell, D. N., Orne, M. T., & Shor, R. E. (1969).
Sleepinduced behavioral response: Relationship to susceptibility to hypnosis and
laboratory sleep patterns. Journal of Nervous and Mental Disease, 148,467-476.
Evans, F. J., Gustafson, L. A., OConnell, D. N., Orne, M. T., & Shor, R. E. (1970).
Verbally induced behavioral responses during sleep. Journal of Nervous and Men-
talDisease, 150, 171-187.
Firth, H. (1973). Habituation during sleep. Psychophysiology, 10,43-5 1.
Foulkes, D. (1966). Thepsychology of sleep. New York: Scribners.
Fox, B. H., & Robbins, J. S. (1952). The retention of material presented during sleep.
Journal of Experimental Psychology, 43.75-79.
Freud, S. (1953). The interpretation of dreams. In J. Strachey (Ed. and Trans.), The
standard edition of the complete psychological works of Sigmund Freud (Vol. 4-
5 ) . London: Hogarth Press. (Original work published 1900)
Goodenough, D. R. (1978). Dream recall: History and current status of the field. In
A. M. Arkin, J. S. Antrobus, & S. J. Ellman (Eds.), The mind in sleep: Psychology
and psychophysiology (pp. I 13- 140). Hillsdale, NJ: Erlbaum.
Learning During Sleep 107

Goodwin, D. W., Powell, B., Bremer, D., Hoine, H., & Stem, J. (1969). Alcohol and
recall: State dependent effects in man. Science, 163, 1358- 1360.
Hallahan, D. P., Kauffman, J. M., & Ball, D. W. (1974). Developmental trends in
recall of central and incidental auditory material. Journal of Experimental Child
Psychology, 17,409-42 1.
Hebb, D. 0.( 1949). The organization of behavior: A neuropsychological theory. New
York: Wiley.
Hilgard, E. R. ( 1979). Divided consciousness in hypnosis: The implications of the hid-
den observer. In E. Fromm & R. Shor (Eds.), Hypnosis: Developments in research
and new perspectives. New York: Aldine.
Hoskovec, J. (1966). Hypnopaedia in the Soviet Union: A critical review of recent
major experiments. International Journal of Clinical and Experimental Hypnosis,
14,308-315.
Hutt, S. J., Hutt, C., Lenard, H. G., Bernuth, H., & Muntjewefi, W. J. (1968). Audi-
tory responsivity in the human neonate. Nature, 218, 888-890.
Itil, T. M., Menon, G. N., Bozak, M., & Sangor, A. (1982). The effects of oxiracetam
(ISF 2522) in patients with organic brain syndrome. Drug Development Research,
2,447-46 I .
Jacobson, A., Lales, A., Lehmann, D., & Zweizig, J. (1965). Somnambulism: All night
electroencephalographic studies. Science, 148, 975-977.
Jacoby, L. L. (1982). Knowing and remembering: Some parallels in the behavior of
Korsakoff patients and controls. In L. S. Cermak (Ed.), Memory and amnesia.
Hillsdale, NJ: Erlbaum.
Jacoby, L. L., & Dallas, M. (198 1). On the relationship between autobiographical
memory and perceptual learning. Journal of Experimental Psychology: General,
110,306-340.
Jacoby, L. L., & Witherspoon, D. (1982). Remembering with and without awareness.
Canadian Journal of Psychology, 36, 300-324.
Johns, M. W., Gay, T. J. A., Goodyear, M. D. E., & Masterton, J. P. (1971). Sleep
habits of healthy young adults: Use of sleep questionnaire. British Journal of Pre-
ventative and Social Medicine, 25,236-24 I .
Johnson, M. K., Kahan, T. L., & Raye, C. L. (1984). Dreams and reality monitoring.
Journal of Experimental Psychology: General, 113, 329-344.
Jus, K., & Jus, A. (1972). Experimental studies on memory disturbances in pathologi-
cal and physiological conditions. International Journal of Psychobiology, 2, 205-
208.
Kintsch, W. (1970). Models of free recall and recognition. In D. A. Norman (Ed.),
Models of human memory (pp. 33 1-373). New York: Academic Press.
Koukkou, M., & Lehmann, D. (1968). EEG and memory storage in sleep experiments
with humans. Electroencephalography and Clinical Neurophysiology, 25, 455-
462.
Koukkou, M., & Lehmann, D. (1983). Dreaming: The functional state-shift hypothe-
sis. British Journal of Psychiatry, 142, 221-23 I .
Koulack, D., & Goodenough, D. R. ( 1976). Dream recall and dream recall failure: An
arousal-retrieval model. Psychological Bulletin, 83.97 5-984.
Kulikov, V. N. (1968). The question of hypnopaedia. In F. Rubin (Ed.), Current re-
search in hypnopaedia (A. Scott, Trans.) (pp. 132-144). New York: Elsevier.
(Original work published 1964)
Lehmann, D., & Koukkou, M. (1974). Computer analysis of EEG wakefulness-sleep
patterns during learning of novel and familiar sentences. Electroencephalography
and Clinical Neurophysiology, 37, 73-84.
Leuba, C., & Bateman, D. ( 1 952). Learning during sleep. American Journal ofPsycho1-
Ogy, 65,301-302.
Levy, C. M., Collidge, F. L., & Stabb, L. V. (1972). Paired associate learning during
EEG-defined sleep: A preliminary study. Australian Journal of Psychology, 24,
219-225.
108 ERIC EICH

Minard, J., Loiselle, R., Ingledue, E., & Duatlich, D. ( 1 968). Discriminative electrooc-
ulogram deflections (EGDs) and heart rate (HR) pauses elicited during main-
tained sleep by stimulus significance. Psychophysiology, 5,232.
Okuma, T., Nakamura, K., Hayashi, A., & Fujimori, M. (1966). Psychophysiological
study on the depth of sleep in normal human subjects. Electroencephalography
and Clinical Neurophysiology, 21, 140- 147.
Oltman, P. K., Goodenough, D. R., Koulack, D., M a c h , E., Schroeder, H. R., &
Flanagan, M. J. (1977). Short-term memory during Stage 2 sleep. Psychophysiol-
ogy, 14, 439-444.
Oswald, I., Taylor, A. M., & Treisman, M. ( 1960). Discriminative responses to stimula-
tion during human sleep. Brain, 83. 440-453.
Overton, D. A. (1984). State dependent learning and drug discriminations. In L. L.
Iverson, S. D. Iverson, & S. H. Snyder (Eds.), Handbook ofpsychopharmacology
(Vol. 18, pp. 59-127). New York: Plenum Press.
Peters, R., & McGee, R. ( 1982). Cigarette smoking and state-dependent memory. Psy-
chopharmacology, 76, 232-235.
Prince, M. ( 1 9 10). The mechanism and interpretation of dreams. Journal ofAbnormal
P.yychology, 19, 137- 195.
Richardson-Klavehn, A., & Bjork, R. A. ( 1988). Measures of memory. Annual Review
(?fPsvcholog.v,39, 475-543.
Rubin. F. ( 1968). Current research in hypnopuedia. New York: Elsevier.
Rubin, F. ( 1970). Learning and sleep. Nature, 226,477.
Rubin, F. ( I97 I). Learning and sleep. Bristol, England: Wright.
Schacter, D. L. ( 1987). Implicit memory: History and current status. Journal ofExperi-
mental Psychology: Learning, Memory, and Cognition, 13, 501-5 18.
Schacter, D. L., & Kihlstrom, J. F. (1989). Functional amnesia. In F. Boller & J. Graf-
man (Eds.), Handbook qfneuropsychology. Amsterdam: Elsevier.
Schacter, D. L., & Tulving, E. (1982). Memory, amnesia, and the episodic/semantic
distinction. In R. L. Isaacson & N. E. Spear (Eds.), The expression ofknowledge.
New York: Plenum Press.
Sheldrake, P., & Cormack, M. (1974). Dream recall and the menstrual cycle. Journal
qfPsychosomatic Research, 18,347-350.
Simon, C. W., & Emmons, W. H. (1955). Learning during sleep? Psychological Bulle-
tin, 52. 328-342.
Simon, C. W., & Emmons, W. H. (1956). Responses to material presented during
various levels of sleep. Journal of Experimental Psychology, 5 I , 89-97.
Svyadoshch, A. M. (1968). The assimilation and memorization of speech during natu-
ral sleep. In F. Rubin (Ed.), Current research in hypnopaedia (A. Scott, Trans.)
(pp. 9 1 - I 17). New York: Elsevier. (Original work published 1962)
Tilley, A. J. ( 1979). Sleep learning during Stage 2 and REM sleep. Biological Psychol-
OR^, 9. 155- I6 1.
Watkins, M. J. ( 1979). Engrams as cuegrams and forgetting as cue overload: A cueing
approach to the structure of memory. In C. R. Puff (Ed.), Memory organization
and structure. New York: Academic Press.
Zavalova, N. D., Zukhar, V. P., & Petrov, Y. A. (1968). The question of hypnopaedia
(preliminary communication). In F. Rubin (Ed.), Current research in hypnopae-
dia (A. Scott, Trans.) (pp. 145-151). New York: Elsevier. (Original work pub-
lished 1964)
Zukhar’, V. P., Kaplan, Y. Y., Maksimov, Y. A., & Pushkina, 1. P. (1968). A collective
experiment on hypnopaedia. In F. Rubin (Ed.), Current research in hypnopaedia
(A. Scott, Trans.) (pp. 152- 159). New York: Elsevier. (Original work published
1965)
 CHAPTER 8

LUCID DREAMING:
PSYCHOPHYSIOLOGICAL STUDIES OF
CONSCIOUSNESS DURING REM SLEEP

STEPHEN LABERGE

Although we are usually not explicitly aware that we are dreaming while we are
dreaming, at times a remarkable exception occurs, and we become conscious
enough to realize that fact. “Lucid” dreamers (the term derives from van
Eeden, 1913) report being able to remember the circumstances of waking life
freely, to think clearly, and to act deliberately upon reflection, all while experi-
encing a dream world that seems vividly real (Green, 1968; LaBerge, 1985;
Gackenbach & LaBerge, 1988). This contrasts with the usual characterization
of dreams as states that typically evince no reflective awareness or true volition
(Rechtschaffen, 1978).
Lucid dreaming is normally a rare experience. Though most people re-
port having had a lucid dream at least once in their lives, only about 20% of
the population reports having lucid dreams once a month or more (Snyder &
Gackenbach, 1988).
Although most people have experienced lucid dreams, some theoreti-
cians have considered them impossible and even absurd (e.g., Malcolm, 1959).
In the absence of empirical evidence, most sleep researchers have been in-
clined to accept Hartmann’s (1975) “impression” that lucid dreams are “not
typical parts of dreaming thought, but rather brief arousals” (p. 74; see also
Berger, 1977). Schwartz and Lefebvre (1973) noted that frequent transitory
arousals are common during REM sleep and proposed that these “microawak-
enings” are the physiological basis for lucid dream reports. Although no one
has found any evidence for this mechanism, their proposal has been the pre-
dominant opinion (cf. Foulkes, 1974) until the last few years.
109
110 STEPHEN LABERGE

LUCID DREAMING PHYSIOLOGICALLY VERIFIED

Empirical evidence began to appear in the late 1970s suggesting that lucid
dreams occur during REM sleep. Based on standard sleep recordings of 2 sub-
jects who reported a total of three lucid dreams upon awakening from REM
periods, Ogilvie, Hunt, Sawicki, and McGowan ( 1978) cautiously concluded
that “it may be that lucid dreams begin in REM.” However, no proof was
given that the reported lucid dreams had in fact occurred during the REM
sleep immediately preceding the awakenings and reports. What was needed to
establish the physiological status of lucid dreams unambiguously was a behav-
ioral response that would signal to the experimenter the exact time at which
the lucid dream was taking place.
LaBerge, Nagel, Dement, and Zarcone (198 1) provided the necessary
verification by instructing subjects to signal the onset of lucid dreams with
specific dream actions that would be observable on a polygraph (i.e., eye
movements and fist clenches). Using this approach, they reported that the oc-
currence of lucid dreaming during unequivocal REM sleep had been demon-
strated for 5 subjects. After being instructed in the method of lucid dream
induction (MILD) described by LaBerge (1 980c), subjects were each recorded
from 2 to 20 nights. In the course of the 34 nights of the study, 35 lucid dreams
were reported subsequent to spontaneous awakening from various stages of
sleep as follows: REM sleep, 32 times; non-REM (NREM) Stage 1 sleep, 2
times; the transition from NREM Stage 2 to REM sleep, 1 time. The subjects
reported signaling during 30 of these lucid dreams. After each recording, the
reports mentioning signals were submitted along with the respective polysom-
nograms to a judge uninformed of the times of the reports. In 24 cases (90%),
the judge was able to select the appropriate 30-s epoch on the basis of corre-
spondence between reported and observed signals. All signals associated with
lucid dream reports occurred during epochs of unambiguous REM sleep
scored according to the conventional criteria (Rechtschaffen& Kales, 1968).
A later analysis extending these data with 2 additional subjects and 20
more lucid dreams produced identical results (LaBerge, Nagel, Taylor, De-
ment, & Zarcone, 1981). LaBerge et al. argued that their investigations dem-
onstrated that lucid dreaming usually (though perhaps not exclusively)occurs
during REM sleep. This conclusion was supported by research carried out in
several other laboratories (Dane, 1984; Fenwick et al., 1984; Hearne, 1978;
Ogilvie, Hunt, Kushniruk, & Newman, 1983).
Ogilvie et al. (1983) reported that the physiological state preceding 14
spontaneous lucidity signals was unqualified REM in 12 (86%)of the cases; of
the remaining 2 cases, 1 was reported to be “ambiguous” REM and the other
was reported to be wakefulness. Hearne and Worsley collaborated on a pio-
neering study of lucid dreaming in which the latter spent 50 nonconsecutive

I am grateful to Ronald Hertel, the Institute for Human Development, the Mericos Foun-
dation, and the Reality Foundation for financial support and to Lynne Levitan for editorial
assistance.
Lucid Dreaming 111

nights in the Hull University sleep laboratory while the former monitored the
polygraph. Worsley reported signaling in 8 lucid dreams, all of which were
described by Hearne ( 1978)as having occurred during REM sleep.
However, demonstrations that the signaling of lucid dreams occurs dur-
ing REM sleep raise another kind of question: What exactly do we mean by
the assertion that lucid dreamers are “asleep?” Perhaps these “dreamers” are
not really dreamers, as some have argued in the last century; or perhaps this
“sleep” is not really sleep, as some have argued in this century. How do we
know that lucid dreamers are really asleep when they signal? If we consider
perception of the external world as a criterion of being awake (to the external
world), we can conclude that they are actually asleep (to the external world)
because, although they know they are in the laboratory, this knowledge is a
matter of memory, not perception. Upon awakening, lucid dreamers report
total immersion in the dream world and no sensory contact with the exter-
nal world.
One might object that lucid dreamers are simply not attending to the
environment; rather than being asleep, perhaps they are merely absorbed in
their private fantasy worlds as are those, for example, who are deeply im-
mersed in a novel or daydream. However, according to the reports of lucid
dreamers (LaBerge, 1980b, 1985), if they deliberately attempt to feel the bed-
covers they know they are sleeping in or try to hear the ticking of the clock
they know is beside their bed, they fail to feel or hear anything except what
they find in their dream worlds. Lucid dreamers are conscious of the absence
of sensory input from the external world; therefore, on empirical grounds,
they conclude that they are asleep.
Conversely, if subjects claim to have been awake while showing physio-
logical signs of sleep (or vice versa), we might have cause to doubt their subjec-
tive reports. However, when the subjective accounts and objective physiologi-
cal measures are in clear agreement (as they are here), it is embarrassingly
awkward to assert (as some critics have done) that subjects who report being
certain that they were asleep while showing physiological indications of
unequivocal sleep were actually awake (cf. LaBerge, Nagel, Dement, & Zar-
cone, 1981).
Some critics have suggested that demand characteristics might account
for our results. It is true that.our subjects were under demand to have, signal,
and report lucid dreams, but could demand alone account for their actions if
they were not lucid in the first place? If they merely unconsciously signaled,
we would have found REM periods with signals without subsequent reports
of lucidity, but we did not. If they merely reported having signaled without
actually having done so, we would have found reports without signals, which
we did not. Further, by this account, where would the reported and observed
signals have come from?
The evidence is clear: Lucid dreaming is an experiential and physiologi-
cal reality; though perhaps paradoxical, it is clearly a phenomenon of sleep.
112 STEPHEN LnBERGE

PHYSlOLOGlCAL CHARACTERISTICS OF LUCID DREAMlNG

The preceding studies have shown that lucid dreams typically occur in REM
sleep. However, REM sleep is a heterogeneous state that exhibits considerable
variations in physiological activity, ordinarily distinguished by two distinct
phases. In its most active form, REM is dominated by a striking variety of
irregular and short-lived events such as muscular twitching, including the
rapid eye movements that give the state one of its most common names. This
variety of REM is referred to as phasic, whereas the relatively quiescent state
remaining when rapid eye movements and other phasic events temporarily
subside is referred to as tonic. On first thought, one might expect lucid dreams
to be associated with decreased phasic activity (Pivik, 1986). However, re-
search described later has shown lucid dreaming to be associated instead with
increased phasic activity.
LaBerge, Levitan, and Dement ( 1986)analyzed physiological data from
76 signal-verified lucid dreams (SVLDs)of 13 subjects. The polysomnograms
corresponding to each of the SVLDs were scored for sleep stages, and every
SVLD REM period was divided into 30-s epochs aligned with the lucidity
onset signal. For each epoch, sleep stage was scored, and rapid eye movements
(EMS)were counted; if scalp skin-potential (SP) responses were observable as
artifacts in the electroencephalograph (EEG), these were also counted. Heart
rate (HR) and respiration rate (RR) were determined for SVLDs recorded with
these measures.
For the first lucid epoch, beginning with the initiation of the signal, the
sleep stage was unequivocal REM in 70 cases (92%).The remaining 6 SVLDs
were less than 30-s long and, hence, were technically unscorable according to
standard criteria (Rechtschaffen & Kales, 1968). For these cases, the entire
SVLD was scored as a single epoch; with this modification, all SVLDs quali-
fied as REM. The lucid dream signals were followed by an average of 115 s
(range = 5-490) of uninterrupted REM sleep. Physiological comparisons of
EM, HR, RR, and SP for lucid versus nonlucid epochs revealed that the lucid
epochs of the SVLD REM periods had significantlyhigher levels of physiologi-
cal activation than the preceding epochs of nonlucid REM from the same
REM period. Similarly,H reflex amplitude was lower during lucid than nonlu-
cid REM (Brylowski,Levitan, & LaBerge, 1989).
To study the temporal variations of physiology as they correlated with
the development and initiation of lucidity, for each SVLD REM period, the
physiological variables were converted to standard scores and averaged across
dreams and subjects. Figure 1 shows the resultant mean standard scores for
the 5 min before and the 5 min after the initiation of lucidity. Note the highly
significant increases in physiological activation during the 30 s before and after
lucidity onset.
Physiological data (EM, RR, HR, and SP) were also collected for 61
control nonlucid REM periods, derived from the same 13 subjects, to allow
comparison with SVLDs. Mean values for EM and SP were significantly
higher for REM periods with lucid dreams than nonlucid control REM peri-
ods (RR and HR did not differ).
Lucid Dreaming 113

1.0-

0.0 -

2.0- SKIN POTENTIAL

1.0-

0.0 -
-l.O-, , , , , , , , , ,: , , , , , , , , , ,
-5 -4 -3 -2 - 1 0 1 2 3 4 5

MINUTES
Figure 1 Grand mean z-scores and standard errors for REM density (EM), respira-
tion rate (RR), heart rate (HR), and scalp skin potential responses (SP) during the 5
min before the onset of lucidity (black bars) and the 5 min after the onset of lucidity
(white bars). Epochs are 30 s in length and the dotted line represents the signaled onset
oflucidity. Sample sizes vary with variable and epoch, but all values are averaged across
lucid dreams and subjects.

Given the finding that lucid dreams reliably occur during activated (pha-
sic) REM, measures of central nervous system activation, such as eye move-
ment density, should contribute something to the pattern of lucid dream dis-
tribution. Because it has been observed that eye movement density starts at a
low level at the beginning of REM periods and increases until it reaches a
peak after approximately 5-7 min (Aserinsky, 197I), we (LaBergeet al., 1986)
hypothesized that lucid dream probability should follow a parallel develop-
ment. Accordingly, we found that mean eye movement density correlated pos-
itively and significantly with lucid dream probability ( r = .66, p < .Ol).
Lucid dreams have most commonly been reported to occur late in the
sleep cycle (Green, 1968). LaBerge et al. (1986) tested this hypothesis by first
determining for each of their 12 subjects the time of night that divided their
total REM time into two equal parts. All but 1 of the subjects had more lucid
dreams in the second half of their REM time than in the first half (binomial
114 STEPHEN LABERGE

:
'Ii
Lucid Dreaming 115

test, p < .O 1). For the combined sample, relative lucidity probability was calcu-
lated for REM Periods 1-6 of the night by dividing the total number of lucid
dreams observed in a given REM period by the corresponding total time in
the REM stage for the same REM period. A regression analysis clearly demon-
strated that relative lucidity probability was a linear function of ordinal REM
period number ( r = .98, p < .0001).
Lucid dreams are initiated in two distinct ways. Subjects usually report
having been in the midst of a dream when a bizarre occurrence causes suffi-
cient reflection to yield the realization that they are dreaming. Less frequently,
subjects report having briefly awakened from a dream and then, falling back
asleep, directly entering the dream with no (or very little) break in conscious-
ness (Green, 1968; LaBerge, 1985). Here is an example of a wake-initiated
lucid dream:
I was lying awake in bed late in the morning listening to the sound of
running water in the adjoining bathroom. Presently an image of the
ocean appeared, dim at first like my usual waking imagery. But its
vividness rapidly increased while, at the same time, the sound of running
water diminished; the intensity of the internal image and external sound
seemed to alter inversely (as if one changed a stereo balance control from
one channel to the other). In a few seconds, I found myself at the seashore
standing between my mother and a girl who seemed somehow familiar. I
could no longer hear the sound of the bath water, but only the roar of the
dream sea. (LaBerge, 1980b, p. 85)
Note that the subject was continuously conscious during the transition
from wakefulness to sleep. This fact suggests that Foulkes (1985) was overstat-
ing the case by claiming that it is “a necessary part of the experience we call
‘sleep’ that we lose a directive and reflective self. You can’t fall asleep, or be
asleep, if your waking self is still regulating and reflecting upon your conscious
mental state” (p. 42).
Because lucid dreams initiated in these two ways ought to differ physio-
logically in at least one respect (i.e., an awakening preceding one but not the
other), the SVLDs were dichotomously classified as either wake-initiated
(WILD) or dream-initiated (DILD), depending on whether or not the reports

Figure 2 A typical dream-initiated lucid dream (DILD). Four channels of physio-

logical data (central EEG [C3-Az],left and right eye-movements [LOC and ROC], and
chin muscle tone [EMG]) from the last 8 niin of a 30 min REM period are shown.
Upon awakening, the subject reported having made five eye movement signals (labeled
1-5 in figure). The first signal ( 1, LRLR) marked the onset of lucidity. Note the skin
potential artifacts in the EEG at this point. During the following 90 s the subject “flew
about” exploring his dream world until he believed he had awakened, at which point
he made the signal for awakening (2, LRLRLRLR). After another 90 s, the subject
realized he was still dreaming and signaled (3) with three pairs of eye movements.
Realizing that this was too many, he correctly signaled with two pairs (4). Finally, upon
awakening 100 s later he signaled appropriately (5, LRLRLRLR). Calibrations are 50
p V and 5 s.
116 STEPHEN LnBERGE

>
3I
i

i
5
Lucid Dreaming 117

mentioned a transient awakening in which the subject consciously perceived

the external environment before reentering the dream state.
Fifty-five (72%)of the SVLDs were classified as DILDs, and the remain-
ing 2 1 (28%)were classified as WILDs. For all 13 subjects, DILDs were more
common than WILDs (binomial test, p < .OOO 1). As expected, compared with
DILDs, WILDs were more frequently immediately preceded by physiological
indications of awakening, x2(1, N = 76) = 38.3, p < .0001, establishing the
validity of classifying lucid dreams in this manner. See Figures 2 and 3 for
illustrations of these two types of lucid dream.
As was mentioned earlier, momentary intrusions of wakefulness occur
very commonly during the normal course of REM sleep, and Schwartz and
Lefebvre ( 1973) proposed that lucid dreaming occurs during these microawak-
enings. However, LaBerge, Nagel, Taylor, Dement, & Zarcone’s ( 198 1) and
LaBerge et al.’s ( 1986) data indicate that, while lucid dreams do not take place
during interludes of wakefulness within REM periods, a minority of lucid
dreams (WILDs) are initiated from these moments of transitory arousal and
continue in subsequent undisturbed REM sleep.
To summarize, an elevated level of central nervous system (CNS) activa-
tion seems to be a necessary condition for the occurrence of lucid dreams.
Evidently, the high level of cognitive function involved in lucid dreaming re-
quires a correspondingly high level of neuronal activation. In terms of Antro-
bus’s (1 986) adaptation of Anderson’s ( 1983) ACT* model of cognition to
dreaming, working memory capacity is proportional to cognitive activation,
which in turn is proportional to cortical activation. Becoming lucid requires
an adequate level of working memory to activate the presleep intention to
recognize that one is dreaming. This level of cortical and cognitive activation
is apparently available only during phasic REM.

PSYCHOPHYSIOLOGICAL RELATIONS DURING REM SLEEP

Psychologistsattempting to apply rigorous scientific methodology to the study
of such phenomena as mental imagery, hallucinations, dreaming, and con-
scious processes in general face a major challenge: The most direct account
available of the private events occumng in a person’s mind is his or her own

Figure 3 A typical wake-initiated lucid dream (WILD) following a transient awak-

ening during REM. Six channels of physiological data (left and right temporal EEG
[T3 and T4;C, reference], left and right eye-movements [LOC and ROC], chin muscle
tone [EMG], and electrocardiogram [ECG]) from the last 3 min of a 14 min REM
period are shown. The subject awoke at I and after 40 s returned to REM sleep at 2,
realized he was dreaming I5 s later, and signaled at 3. Next he carried out the agreed-
upon experimental task in his lucid dream, singing between signals 3 and 4, and count-
ing between signals 4 and 5. This allowed comparison of left and right hemisphere
activation during the two tasks (LaBerge & Dement, 1982b). Note the heart-rate accel-
eration-deceleration pattern at awakening (1) and at lucidity onset (3), and the skin
potential artifacts in the EEG (particularly T4)at lucidity onset (3). Calibrations are 50
pV and 5 s.
118 STEPHEN LABERGE

subjective report. Unfortunately, subjective reports are difficult to verify ob-

jectively, and introspection is far from an unbiased and direct process of obser-
vation. Two strategies are likely to increase our confidence in the reliability of
subjective reports: (a) the use of highly trained (and in the context of dream
research, lucid) subjects who are skillful reporters, and (b) the use of the psy-
chophysiological approach, which proposes that the convergent agreement of
physiological measures and subjective reports provides a degree of validation
to the latter (Stoyva & Kamiya, 1968).
Indeed, the psychophysiologicalapproach was responsible for the golden
age of dream research in the decades following the discovery of REM sleep
(Aserinsky & Kleitman, 1953) and the subsequent association of REM with
dreaming (Dement & Kleitman, 1957). Although the psychophysiological
paradigm of dream research has yielded an abundant harvest for many years
(see Arkin, Antrobus, & Ellman, 1978),it possesses a fatal flaw: As long as the
subjects are nonlucid, the researcher has no way of making certain that the
subjects will dream about what the researcher might like to study. Presleep
manipulations producing reliable effects on dream content have not been
highly successful (Tart, 1988).One can only wait and hope that, eventually, a
dream report will unearth what one is looking for. This is really no better than
a shot-in-the-dark approach, and some researchers have proposed abandoning
the psychophysiologicalmethod in favor of a purely psychological approach.
Foulkes ( 1981) wrote that “psychophysiologicalcorrelation research now ap-
pears to offer such a low rate of return for effort expended as not to be a wise
place for dream psychology to continue to commit much of its limited re-
sources” (p. 249). This conclusion may well be justified, but only insofar as it
refers to the psychophysiologicalapproach as it is traditionally practiced, using
nonlucid subjects. The use of lucid dreamers overcomes the basic difficulty of
the old methodology and may revitalize the psychophysiologicalapproach to
dream research.
The fact that lucid dreamers can remember to perform predetermined
actions and can signal to the laboratory suggested to LaBerge (1980b) a new
paradigm for dream research: Lucid dreamers, he proposed, “could carry out
diverse dream experiments marking the exact time of particular dream events,
allowing the derivation of precise psychophysiological correlations and the
methodical testing of hypotheses” (LaBerge, Nagel, Dement, & Zarcone,
1981, p. 727). This strategy has been put into practice in a number of studies
that are summarized later.
How long do dreams take? This question has intrigued humanity for
many centuries. A traditional answer is that dreams take very little or no time
at all, as in the case of Maury’s famous dream in which he somehow got mixed
up in a long series of adventuresduring the French Revolution and finally lost
his head on the guillotine, at which point he awoke to find that the headboard
had fallen on his neck. He therefore supposed that the lengthy dream had been
produced in a flash by the painful stimulus. The idea that dreams occur in the
moment of awakening has found supporters over the years.
Lucid Dreaming 119

We have straightforwardly approached the problem of dream time by

asking subjects to estimate 10-sintervals (by counting “one thousand and one,
one thousand and two,” etc.) during their lucid dreams. Signals marking the
beginning and end of the subjective intervals allowed comparison with objec-
tive time. In all cases, time estimates during the lucid dreams were very close
to the actual time between signals (LaBerge, 1980b, 1985). However, this find-
ing does not rule out the possibility of time distortion effects under some cir-
cumstances.
The data reported by LaBerge, Nagel, Dement, and Zarcone ( 1981) and
LaBerge, Nagel, Taylor, Dement, and Zarcone (198 1) indicate that there is a
very direct and reliable relation between the gaze shift reported in lucid dreams
and the direction of polygraphically recorded eye movements. The results ob-
tained for lucid dreams (see also Dane, 1984; Fenwick et al., 1984; Hearne,
1978; Ogilvie, Hunt, Tyson, Lucescu, & Jeakins, 1982) are much stronger
than the generally weak correlations obtained by previous investigatorstesting
the hypothesis that the dreamer’s eyes move with his or her hallucinated
dream gaze, who relied on the chance occurrence of a highly recognizable eye
movement pattern that was readily matchable to the subject’s reported dream
activity (e.g., Rofbarg, Dement, Muzio, & Fisher, 1962).
LaBerge (1985) reported related experimentsin which 2 subjects tracked
the tip of their fingers moving slowly from left to right during four conditions:
(a) awake, eyes open; (b) awake, eyes closed, mental imagery; (c) lucid dream-
ing; and (d) imagination (“dream eyes closed”) during lucid dreaming. The
subjects showed saccadic eye movements in the two imagination conditions
(b and d) and smooth tracking eye movements during dreamed or actual
tracking (a and c).
In another study, LaBerge and Dement (1982b)demonstrated the possi-
bility of voluntary control of respiration during lucid dreaming. They re-
corded 3 lucid dreamers who were asked either to breathe rapidly or to hold
their breath (in their lucid dreams), marking the interval of altered respiration
with eye movement signals. The subjects reported successfully carrying out
the agreed-upon tasks a total of nine times and, in every case, a judge was able
to predict correctly on the basis of the polygraph recordings which of the two
patterns had been executed (binomial test, p < .002).
Evidence of the voluntary control of other muscle groups during REM
was found by LaBerge, Nagel, Dement, and Zarcone (198 1) while testing a
variety of lucidity signals. They observed that a sequence of left and right
dream-fist clenches resulted in a corresponding sequence of left and right fore-
arm twitches as measured by electromyograph (EMG). However, the ampli-
tude of the twitches bore an unreliable relation to the subjective intensity of
the dreamed action. Because all skeletal muscle groups except those that gov-
ern eye movements and breathing are profoundly inhibited during REM
sleep, it is to be expected that most muscular responses to dreamed move-
ments will be feeble. Nonetheless, these responses faithfully reflect the motor
120 STEPHEN LABERGE

patterns of the original dream. Similar observations were made by Fenwick

et al. (1984).
Following reports of cognitive task dependency of lateralization of EEG
alpha activity in the waking state by many researchers, LaBerge and Dement
(1982a) undertook a pilot study to demonstrate the feasibility of similar inves-
tigations in the lucid dream state. The two tasks selected for comparison were
dreamed singing and dreamed counting, activities expected to result in rela-
tively greater engagement of the subjects’ left and right cerebral hemispheres,
respectively.
Integrated alpha band EEG activity was derived from electrodes placed
over right and left temporal lobes while 4 subjects sang and estimated 10 s by
counting in their lucid dreams (marking the beginning and end of each task
by eye movement signals). The results supported the hypothesized lateraliza-
tion of alpha activity: The right hemisphere was more active than the left dur-
ing singing; during counting, the reverse was true. These shifts were similar
to those observed during actual singing and counting. In contrast, a control
condition with imagined singing and counting showed no significant laterality
shifts. Because of the small number of subjects, the conclusions of this study
must be regarded as suggestive at best.
LaBerge and Dement (1982a, 1982b) noted an important implication of
their results for the interpretation of EEG alpha activity during REM sleep.
Because continuous alpha activity occurs when a subject awakens, sleep re-
searchers have usually assumed that increased alpha activity in the context of
sleep is always a sign of wakefulness or relative cortical activation. The findings
just discussed suggest the contrary: Alpha activity during REM sleep is, as in
waking, inversely related to cortical activation. When a person awakens from
a vivid dream to a dark room, his cortical (at least occipital) activation
has decreased, not increased, with the resultant appearance of elevated
alpha power.
In this view, it is a straightforward prediction that occipital alpha power
during REM sleep will correlate negatively with subsequently reported dream
vividness. This could provide the proper explanation for the findingthat awak-
enings following REM periods with high levels ofalpha activity are more likely
to yield “thinking” reports than awakenings from low-alpha REM periods
which yield more “dreaming” reports (Antrobus, Dement, & Fisher, 1964).
Sexual activity is a rather commonly reported theme of lucid dreams
(LaBerge, 1985; Garfield, 1979). LaBerge, Greenleaf, and Kedzierski (1983)
undertook a pilot study to determine the extent to which subjectively experi-
enced sexual activity during REM lucid dreaming would be reflected in physi-
ological responses.
Sixteen channels of physiological data, including EEG, electrooculo-
gram (EOG), EMG, respiration, skin conductance level (SCL), heart rate, vag-
inal EMG (VEMG), and vaginal pulse amplitude (VPA), were recorded from
a single subject. The experimental protocol called for her to make specific eye
movement signals at the following points: When she realized she was dream-
Lucid Dreaming 121

ing (i.e., the onset of the lucid dream), when she began sexual activity (in the
dream), and when she experienced orgasm.
The subject reported a lucid dream in which she camed out the experi-
mental task exactly as agreed upon. Data analysis revealed a significant corre-
spondence between her subjective report and all but one of the autonomic
measures; during the 15-s orgasm epoch, mean levels for VEMG activity,
VPA, SCL, and respiration rate reached their highest values and were signifi-
cantly elevated compared with means for other REM epochs. Contrary to ex-
pectation, heart rate increased only slightly and nonsignificantly.

IMPLICATIONS FOR RESEARCH ON

SLEEP AND COGNITION
Lucid dreaming presents conceptual difficulties for certain traditional beliefs
about sleep and about the presumed limitations of dream mentation. In a
certain sense, the anomalous appearance of lucid dreaming parallels that of
the state that has been called “paradoxical sleep.” The discovery of REM sleep
required the expansion of our concept of sleep. The evidence we have reviewed
associating lucid dreaming with REM sleep seems to require a similar expan-
sion of our concept of dreaming as well as a clarification of our concept
of sleep.
Fenwick et al. (1984) showed that a subject was able to perceive and
respond to environmental stimuli (electrical shocks) without awakening from
his lucid dream. This result raises a theoretical issue: If we take perception of
the external world to be the essential criterion for wakefulness, then Worsley
must have been at least partially awake. On the other hand, when environmen-
tal stimuli are incorporated into dreams without producing any subjective or
physiological indications of arousal, it seems that the perception must have
occurred during sleep.
Furthermore, it may be possible, as LaBerge (1980a) suggested, for one
sense to remain functional and awake while others fall asleep. Similarly, An-
trobus, Antrobus, and Fisher (1965) argued that “the question-awake or
asleep-is not a particularly useful one. Even though we have two discrete
words-sleep and wakefulness-this does not mean that the behavior associ-
ated with the words can be forced into two discrete categories. . . . not only
do sleeping and waking shade gradually into one another but there is only
limited agreement among the various physiological and subjective operations
that discriminate between sleeping and waking. At any given moment,
all systems of the organism are not necessarily equally asleep or awake.”
(pp. 398-399)
As long as we continue to consider wakefulness and sleep a simple di-
chotomy, we will lie in a Procrustian bed that is bound at times to be most
uncomfortable. There must be degrees ofbeing awake, just as there are degrees
of being asleep (i.e., the conventional sleep stages). Before we find our way out
of this muddle, we will probably need to characterize a wider variety of states
122 STEPHEN LABERGE

of consciousness than the few that are currently distinguished (e.g., dreaming,
sleeping, waking, etc.).
It may be helpful to consider lucidity from a cognitive-developmental
perspective. According to Piaget ( 1926), children pass through three stages of
understanding of the concept dream. In the first stage, they believe that dreams
take place in the same external world as all other experiences. In the second
stage, children treat dreams as if they were partially external and partially in-
ternal. This transitional stage gives way to the third stage in which children
recognize the dream is entirely internal in nature, a purely mental experience.
These foregoing developmental stages refer to how children think about
dreams when they are awake. While asleep and dreaming, children, and also
adults, tend to remain at the first stage, implicitly assuming that the dream
events are external reality. Out-of-body experiences,with a contradictory mix-
ture of material and mental (external and internal), may provide examples of
the second stage (LaBerge, Levitan, Brylowski, & Dement, 1988). In the fully
lucid dream, the dreamer attains the third stage, realizing that the dream world
is distinct from the physical world.
Foulkes (1982, 1985) has emphasized the idea that the growth of the
mind, whether dreaming or awake, shows parallel degrees of development:
“there are ‘stages’ of dream development which individual children reliably
pass through one after the other, and that the precise age at which they reach
a new stage is at least partially predictable from independent measures of their
waking mental development” (1985, p. 137).
In this view, lucid dreaming represents what ought to be a normal ability
in adults. If this is correct, why are lucid dreams so rare, especially in cases
such as nightmares, where lucidity would be extremely helpful and rewarding?
A possible answer may be found by comparing lucid dreaming with another
cognitive skill, language acquisition. All normal adults speak and understand
at least one language. But how many would do so if they were never taught?
Unfortunately, in this culture, with few exceptions, we are not taught to
dream.
LaBerge ( 1 980c) demonstrated that lucid dreaming is a learnable skill
and that there are a variety of techniques available for inducing lucid dreams
(LaBerge, 1985; Price & Cohen, 1988). M e r g e and colleagues have experi-
mented with methods for helping dreamers realize that they are dreaming by
means of external cues applied during REM sleep, which, if incorporated into
dreams, can remind dreamers that they are dreaming (Merge, 1980b). They
have tested a variety of stimuli, including tape recordings of the phrase “this
is a dream” (LaBerge, Owens, Nagel, & Dement, 1981), conditioned tactile
stimuli (Rich, 1985), olfactory stimuli (LaBerge, Brylowski, & Levitan, un-
published data, 1986), and light (LaBerge, Levitan, Rich, & Dement, 1988).
The most promising results so far have resulted from light stimuli.
The psychophysiological studies reviewed here all support the following
picture: During REM dreaming, the events we experience (or seem to experi-
ence) are the results of patterns of CNS activity that produce in turn effects on
our autonomic nervous systems (ANS) and bodies, which are to some extent
Lucid Dreaming 123

modified by the specific conditions of active sleep but are still homomorphic
to the effects that would occur if we were actually to experience the corre-
sponding events while awake.
This conclusion may need further qualification and explanation. Al-
though the events we appear to perceive in dreams are illusory, our feelings in
response to dream content are real. Indeed, most of the events we experience
in dreams are real; when we experience feelings like anxiety or ecstasy in
dreams, we really do feel anxious or ecstatic at the time. When we think in
dreams, we really do think (whether clearly or not is another matter). If we
think in our dreams that Monday comes before Sunday, it is not the case, as
some philosophers (e.g., Malcolm, 1959) assert, that we have only dreamed
we thought; we may have thought incorrectly (to the usual way of thinking),
but we thought nonetheless.
If we vividly imagined a detailed sequence of movements, like walking
around the room, it is probable that motor areas of the brain would be acti-
vated in the same pattern that is activated in actual walking. However, they
would presumably be less activated than when walking. Otherwise, what
would prevent us from actually walking when we imagined doing so?
In REM sleep, a spinal paralysis causes the muscles of locomotion and
vocalization to fail to completely execute the action orders programmed by
the brain. Thus, in REM, unlike the waking state, nothing impedes the brain
from issuing sequences of motor commands at normal levels of activation,
and this probably contributesto the experienced reality of dreamed action.
As for the afferent side of the equation, a great deal of evidence suggests
that imagery uses the same neural systems as perception in the corresponding
sensory mode (e.g., see Farah, 1988; Finke, 1980). In this view, the essential
difference between a perception and a corresponding image is how the identi-
cal neural system acquires sufficient activation to produce a conscious experi-
ence. In the case ofperception, neural excitation (and the resultant experience)
is generated by external input, driving activation of the particular schema to-
be-perceived in a largely bottom-up process. In the case of imagining (likewise,
hallucinating or dreaming), the experienced image is generated internally by
topdown processes activating the appropriate neural network (schema).
Imaginations and perceptions are normally distinguishable by the fact
that images are usually much less vivid than perceptions. Normally, percep-
tions seem real and images seem imaginary. How real something appears de-
pends mainly on its relative vividness, and experienced vividness is probably
a function of the intensity of neural activation. Thus, we may conjecture that
images usually involve a lesser degree of neural activation than the corre-
sponding perceptions, and this results in a lesser degree of experiential reality
for imagination. At least two factors contribute to this state of affairs: One is
that, while we are awake, sensory input produces much higher levels of activa-
tion than imaginary input. Imagination interfereswith perception in the same
modality (Perky, 1910; Segal, 1971), and we may suppose the reverse is true
as well. Another more speculative factor favoring perceptual processes over
imagination in the waking state is the existence of a neural system to inhibit
124 STEPHEN LABERGE

the activation (vividness) of memory images while perception is active. Evolu-

tionary considerations make such a system likely; it would obviously be ex-
tremely maladaptive for an organism to mistake a current perceptual image of
a predator for the memory of one (LaBerge, 1985). Mandell ( 1980)implicated
serotonergic neurons as part of a system that normally inhibits vivid images
(hallucinations) but is itself inhibited in REM sleep, allowing dreamed percep-
tions (i.e., images) to appear as vividly real as perceptions. In REM, sensory
input is also actively suppressed to prevent competition from perceptual pro-
cesses.
Perhaps this explains in part why we are so inclined to mistake our
dreams for reality: To the functional systems of neuronal activity that con-
struct our experiential world (model), dreaming of perceiving or doing some-
thing is equivalent to actually perceiving or doing it.

References
Anderson, J. R. (1983). The architecture of cognition. Cambridge, MA: Harvard Uni-
versity Press.
Antrobus, J. S. (1986). Dreaming: Cortical activation and perceptual thresholds. Jour-
nal of Mind and Behavior, 7, 193-2 12.
Antrobus, J. S., Antrobus, J. S., & Fisher, C. (1965). Discrimination of dreaming and
nondreaming sleep. Archives of General Psychiatry, 12, 395-401.
Antrobus, J. S., Dement, W., & Fisher, C. (1964). Patterns of dreaming and dream
recall: An EEG study. Journal of Abnormal and Social Psychology, 69, 244-252.
Arkin, A., Antrobus, J., & Ellman, S. (Eds.). (1978). The mind in sleep. Hillsdale, NJ:
Erlbaum.
Aserinsky, E. (197 1). Rapid eye movement density and pattern in the sleep of young
adults. Psychophysiology, 8, 36 1-375.
Aserinsky, E., & Kleitman, N. (1953). Regularly occumng periods of eye motility and
concomitant phenomena during sleep. Science, 118,273-274.
Berger, R. ( 1977). Psyclosis: The circularity of experience. San Francisco: Freeman.
Brylowski, A., Levitan, L., & LaBerge, S. (1989). H-reflex suppression and autonomic
activation during lucid REM sleep: A case study. Sleep, 12, 374-378.
Dane, J. (1984). An empirical evaluation of two techniquesfor lucid dream induction.
Unpublished doctoral dissertation, Georgia State University.
Dement, W., & Kleitman, N. (1957). Cyclic variations in EEG during sleep and their
relation to eye movements, body motility, and dreaming. Electroencephalography
and Clinical Neurophysiology, 9, 673-690.
Farah, M. J. (1988). Is visual imagery really visual? Overlooked evidence from neuro-
physiology. Psychological Review, 95, 307-3 17.
Fenwick, P., Schatzman, M., Worsley, A., Adams, J., Stone, S., & Baker, A. (1984).
Lucid dreaming: Correspondence between dreamed and actual events in one sub-
ject during REM sleep. Biological Psychology, 18,243-252.
Finke, R. A. (1980). Levels of equivalence in imagery and perception. Psychological
Review, 87, 1 13- 132.
Foulkes, D. (1974). [Review of Schwartz and Lefebvre ( 1973)]. Sleep Research, 3, 1 13.
Foulkes, D. (1981). Dreams and dream research. In W. Koella (Ed.), Sleep 1980 (pp.
246-257). Basel, Switzerland: Karger.
Lucid Dreaming 125

Foulkes, D. (1982). A cognitive-psychological model of dream production. Sleep, 5,

169-187.
Foulkes, D. ( 1985). Dreaming: A cognitive-psychological analysis. Hillsdale, NJ:
Erlbaum.
Gackenbach, J., & LaBerge, S. (1988). Conscious mind, sleeping brain. New York:
Plenum Press.
Garfield, P. (1979). Pathway to ecstasy. New York: Holt, Rhinehart, &Winston.
Green, C. ( 1968). Lucid dreams. London: Hamish Hamilton.
Hartmann, E. ( 1975). Dreams and other hallucinations: An approach to the underlying
mechanism. In R. K. Siegal & L. J. West (Eds.), Hallucinations (pp. 7 1-79). New
York: Wiley.
Hearne, K. M. T. (1978). Lucid dreams: An electrophysiological and psychological
study. Unpublished doctoral dissertation, University of Liverpool.
LaBerge, S. ( 1 980a). Induction of lucid dreams. Sleep Research, 9, 138.
LaBerge, S. (1980b). Lucid dreaming: An exploratory study of consciousness during
sleep (Doctoral dissertation, Stanford University, 1980). Dissertation Abstracts
International, 41, 05B-1966. (University Microfilms International No. 80-24,
69 1).
LaBerge, S. ( 1980~).Lucid dreaming as a learnable skill: A case study. Perceptual and
Motor Skills, 51, 1039-1042.
LaBerge, S. (1985). Lucid dreaming. Los Angeles: Tarcher.
LaBerge, S., & Dement, W. C. (1982a). Lateralization of alpha activity for dreamed
singing and counting during REM sleep. Psychophysiology, 19,331-332.
LaBerge, S., & Dement, W. C. (1982b). Voluntary control of respiration during REM
sleep. Sleep Research, 11, 107.
LaBerge, S., Greenleaf, W., & Kedzierski, B. (1983). Physiological responses to
dreamed sexual activity during lucid REM sleep. Psychophysiology, 20,454-455.
LaBerge, S., Levitan, L., Brylowski, A., & Dement, W. (1988). “Out-of-body’’ experi-
ences occumng during REM sleep. Sleep Research, 17, 1 15.
LaBerge, S., Levitan, L., & Dement, W. C. (1986). Lucid dreaming: Physiological cor-
relates of consciousness during REM sleep. Journal ofMind and Behavior, 7,25 1-
258.
LaBerge, S., Levitan, L., Rich, R., & Dement, W. (1988). Induction of lucid dreaming
by light stimulation during REM sleep. Sleep Research, 17, 104.
LaBerge, S., Nagel, L., Dement, W. C., & Zarcone, V., Jr. (1981). Lucid dreaming
verified by volitional communication during REM sleep. Perceptual and Motor
Skills, 52, 727-732.
LaBerge, S., Nagel, L., Taylor, W., Dement, W. C., & Zarcone, V., Jr. (198 1). Psycho-
physiological correlates of the initiation of lucid dreaming. Sleep Research, 10,
149.
LaBerge, S., Owens, J., Nagel, L., & Dement, W. (198 1). “This is a dream”: Induction
of lucid dreams by verbal suggestion during REM sleep. Sleep Research, 10, 150.
Malcolm, N. ( 1 959). Dreaming. London: Routledge.
Mandell, A. J. (1980). Toward a psychobiology of transcendence: God in the brain. In
J. M. Davidson & R. J. Davidson (Eds.), The psychobiology of consciousness. New
York: Plenum Press.
Ogilvie, R., Hunt, H., Kushniruk, A., & Newman, J. (1983). Lucid dreams and the
arousal continuum. Sleep Research, 12, 182.
Ogilvie, R., Hunt, H., Sawicki, C., & McGowan, K. ( 1 978). Searching for lucid dreams.
Sleep Research, 7, 165.
Ogilvie, R., Hunt, H., Tyson, P. D., Lucescu, M. L., & Jeakins, D. B. (1982). Lucid
dreaming and alpha activity: A preliminary report. Perceptual and Motor Skills,
55,795-808.
Perky, C. W. (1 9 10). An experimental study of imagination. American Journal of Psy-
chology, 21,422-452.
Piaget, J. (1926). The child’s conception of the world. New York: Harcourt, Brace.
126 STEPHEN LABERGE

Pivik, R. T. (1986). Sleep: Physiology and psychophysiology. In M. G. H. Coles, E.

Donchin, & S. Porges (Eds.), Psychophysiology:Systems, processes, and applica-
tions (pp. 378-406). New York: Guilford Press.
Price, R. F., & Cohen, D. B. (1988). Lucid dream induction: An empirical evaluation.
In J. Gackenbach & S. LaBerge (Eds.), Consciousmind, dreaming brain (pp. 105-
154). New York Plenum Press.
Rechtschaffen, A. (1978). The single-mindedness and isolation of dreams. Sleep, I,
97-109.
Rechtschaffen, A., & Kales, A. (Eds.). (1968). A manual of standardized terminology,
techniques and scoring system for sleep stages of human subjects. Bethesda, MD:
HEW Neurological Information Network.
Rich, R. (1985). The induction of lucid dreams by tactile stimulation during REM
sleep. Unpublished honors thesis.
Rofharg, H., Dement, W. C., Muzio, J., & Fisher, C. (1962). Dream imagery: Rela-
tionship to rapid eye movements of sleep. Archives of General Psychiatry, 7, 235-
238.
Schwartz, B. A., & Lefebvre, A. (1973). Contacts v&lle/P.M.O.: 11. Les P.M.O. morcel-
6es [Conjunction of waking and REM sleep: 11. Fragmented REM periods]. Revue
d’Electroencephalographieet de NeurophysiologieClinique, 3, 165- 176.
Segal, S. J. (197 1). Processing of the stimulus in imagery and perception. In S. J. Segal
(Ed.), Imagery: Current cognitive approaches(pp. 73- 100). New York Academic
Press.
Snyder, T., & Gackenbach, J. (1988). In J. Gackenbach & S. LaBerge (Eds.), Conscious
mind, dreaming brain (pp. 221-259). New York: Plenum Press.
Stoyva, J., & Kamiya, J. (1968). Electrophysiological studies of dreaming as the proto-
type of a new strategy in the study of consciousness. Psychological Review, 75,
192-205.
Tart, C. (1988). From spontaneous event to lucidity: A review of attempts to con-
sciously control nocturnal dreaming. In J. Gackenbach & s. LaBerge (Eds.), Con-
scious mind, dreaming brain (pp. 67-103). New York: Plenum Press.
van Eeden, F. ( 1913). A study of dreams. Proceedings of the Society for Psychical Re-
search, 26,43 1-46 1.
 CHAPTER 9

COGNITIVE PROCESSING AND

GENERAL ANESTHESIA
LES GOLDMANN

In this chapter, I will (a) draw comparisons between sleep and general anesthe-
sia and (b) present evidence for cognitive processing under general anesthesia.
To those familiar with general anesthesia, it may be puzzling that this topic is
included in a conference devoted to sleep and cognition. Although anesthetists
and surgeons speak of putting the patient “to sleep,” the two states are worlds
apart, both physiologicallyand psychologically.
Before noting these differences, let me explain why cognitive processing
under general anesthesia may be of interest to this audience. First, because it
is difficult, if not impossible, to know whether a sleeper or an anesthetized
patient is phenomenally aware, research in each area provides ideas and meth-
ods that are of mutual benefit. Second, many authors have referred (I believe
mistakenly) to instances of the recall of perioperative events as dreams. Al-
though these instances may or may not be dreams, it is possible that they are
influenced in a manner similar to dream recall. And finally, increasing evi-
dence shows that patients under general anesthesia continue to hear, process,
and respond to auditory input. Responses to this input bear comparison to
those displayed by sleeping subjects.

GENERAL ANESTHESIA AND SLEEP

General anesthesiais not sleep. Lunn ( 1982,p. 66) defined sleep as a physiological
coma and general anesthesia as a pharmacological coma. Physiologically, anes-
thesia resembles non-REM (NREM) sleep, with increased activity in the sympa-
thetic nervous system and cortical depression. Electroencephalograph(EEG) and
electromyograph (EMG) recordings during anesthesia bear some resemblances
to NREM sleep. However, EEG recordings are dramatically effected by differing
127
128 LES GOLDMANN

drug combinations,which makes it difficult or impossible both to assessthe depth

of anesthesia and to make consistent comparisons with sleep. A flat EEG, as
might be found with barbiturate induction, for example, does not change the
auditory evoked response to a stimulus at the cortex.
Patients are often given anticholinergicsthat interfere with the transition
from NREM to REM sleep. These factors may explain why patients almost
uniformly report that no time has passed and that REM-like dreams are rare.
In addition to changes in time perception, patients often report feelings of
being another person or being dissociated from their bodies when they
awaken. They do not report feeling rested or having dreamt, and they do not
remember moving. They are usually awakened by someone calling their name
and asking them to open their eyes, which is clearly different from a usual
awakening from sleep. Finally, the reasons for sleep and general anesthesia are
unquestionably different, and the contexts in which they occur are relevant.

AWARENESS UNDER ANESTHESIA

Crawford ( 197 1) defined awareness under anesthesia as the recall of an event
that is neither a dream nor a hallucination. Although this definition appears
to be quite sensible, he did not suggest how instances of awareness might be
differentiated, nor did he consider the possibility of recall via dreams or hallu-
cinations. In addition to these problems, patients often fall asleep after their
operations, if not immediately afterward, usually before they are interviewed
about perioperative events. It is therefore possible that reports of awareness
could be mixed with sleep mentation occurring after the operation. Most peo-
ple are skilled at separating sleep, dreams, daydreams, and normal waking
reality from each other, and these skills are probably carried into the operating
room. Most important, however, is that patients may recollect perioperative
events that they were unaware of at the time they occurred. In addition, the
possibility exists that patients are phenomenally aware under general anesthe-
sia but that anterograde or retrograde amnesia is sufficiently dense to make
recollection improbable.

COGNITIVE PROCESSING UNDER GENERAL ANESTHESIA

The cognitive processing under investigation is the ability of patients, either
during or after their operation, to respond to auditory stimuli, while under the
influence of a properly administered general anesthetic. In the case of verba-
tim recall, the cognitive process is obvious. However, more often than not, the
process is subtle and open to interpretation. The following statements can be
made about information processing under general anesthesia:
1. Patients do process information under general anesthesia. The
results of this processing can be detected by orienting responses,
classical conditioning (in animals), postoperative nonverbal
responses, recognition, and recall.
2. Evidence for the recollection of perioperative events decreases as the
intentionality of retrieval strategies increases.
Anesthesia 129

3. Recall is, in general, a poor indicator of cognitive processing under

general anesthesia.
4. General anesthesia disrupts the encoding and retrieval of memories.
Retrieval can be enhanced through strategies, such as hypnosis, that
supply cues and contextual elements or improve rapport.
5. Recollection occurs with and without phenomenal awareness.
Support for the occurrence of information processing under anesthesia can be
found in stimuli presentations that result in (a) orienting responses (Gold-
mann & Levey, 1986; Levinson, 1965); (b) nonverbal responses that relate to
recovery (Evans & Richardson, 1988); affective responses (Blacher, 1975) and
specific behavioral responses (i.e., ear pulling; Bennett, Davis, & Giannini,
1985; Goldmann, Shah, & Hebden, 1987); (c) conditioning (in animals) dem-
onstrated by response suppression (Weinberger,Gold, & Sternberg, 1984); (d)
the recognition ofwords in a signal detection task (Millar & Watkinson, 1983);
and (e) the recall of operative events through dreamlike processes (Federn,
1953) and verbatim recall (Cheek, 1979; Goldmann, 1986; Goldmann et al.,
1987; Levinson, 1965).
In examining the evidence, the reader should bear in mind that the phe-
nomenon of cognitive processing under anesthesia and subsequent recollec-
tion occurs in a very particular environment, that of the operating room and
the hospital. Patients may not wish to remember or to report. They may feel
constrained by staff-patient relationships and by their desire to appear compli-
ant and grateful. In addition, patients may have been told that they will be
asleep and unconscious and, thus, may have every reason to doubt their per-
ceptions. This doubt is often reinforced, both by the staff and by the patients’
desire to believe that they were not in any way observers at their own opera-
tions. Consequently, recollections that bypass conscious recall or occur out-
side the patient’s phenomenal awareness are more likely to present evidence
for cognitive processing under anesthesia. Unfortunately, these same re-
sponses may be the most difficult to detect.
One anecdote illustrates this latter point: During the induction of anes-
thesia, an anesthetist explained the observed contractions of different muscle
groups to a medical student. He said that it reminded him of writhing bodies
on funeral pyres in India. The next morning, I questioned the patient about
his operation. He had no recall of events, even with the aid of hypnosis. How-
ever, he emphatically stated that he had had a hot, uncomfortable, restless,
and altogether miserable night’s sleep. Whether there is a connection between
these two events is difficult to ascertain. However, it is suggestive of the plausi-
ble connections that could be made between perioperative events and later
nonconscious recollections.

Orienting
The orienting response can be seen as behavioral, perceptual, or physiological
efferent activity that can be elicited by a particular stimulus and occurs in close
130 LES GOLDMANN

temporal proximity to the eliciting stimulus. Only two orienting responses

have been investigated for general anesthetics, the EEG and the galvanic skin
responses (GSR). Studies of orienting responses have not yielded consistent
results. This may be due, in part, to the collative variables (probability, nov-
elty, etc.) of the stimuli. In one successful study, Goldmann and Levey ( 1986)
found that a majority of premedicated, spontaneously breathing patients re-
ceiving a volatile anesthetic agent showed an orienting response (GSR) when
the name Arthur Scargill (the leader of the National Union of Mineworkers)
was mentioned. The study was conducted in the United Kingdom during the
mineworkers’ strike, and a similar response may not have occurred had the
strike not been in effect at that time. Names other than Scargill’s, including
the patient’s own name, did not elicit orienting responses. This result differs
from the result found with sleepers. In addition, conversation pertinent to the
operation did not elicit an orienting response as readily as unusual conversa-
tion, such as a joke. Salience appears to be a key variable, and studies that do
not provide meaningful or significant stimuli are unlikely to present evidence
supportive of awareness.

Postoperative Nonverbal Responses

Several researchers (Bonke, Schmitz, Verhage, & Zwaveling, 1986; Bonnett,
1966; Evans & Richardson, 1988; Pearson, 196 1, 1965) have investigated the
influence of intraoperative suggestions on the patient’s postoperative course.
Evans and Richardson’s study is most notable for its striking results, including
significant decreases in hospital stay, pyrexia, and gastrointestinal problems.
Other authors have concentrated on highly specific postoperativebehav-
ioral responses. Bennett et al. (1985) found that patients instructed under gen-
eral anesthesia to pull on their ears during a postoperative interview did so
significantly more often than a control group. Goldmann et al. (1987) found
similar results with cardiac patients who were asked to touch their chins. None
of the patients in either study recalled the suggestion, even with the aid of
hypnosis. Goldmann et al. (1987) enhanced the salience of the message by
recording the patient’s name at the beginning and included additional sugges-
tions for an easy and comfortable recovery. All patients were seen preopera-
tively and were told that they might hear a tape during their operation. The
preoperative interview as well as the specific nature of the intraoperative sug-
gestions were designed to enhance expectancy and patient motivation.

Recognition
Several researchers have examined the recognition of words or phrases played
under anesthesia, with differing results. Dubovsky and Trustman (1976) gave
obstetric patients letter-word pairs to remember under anesthesia (e.g., “G is
for game”). The experimental group did not do significantly better than the
control group on simple correct pairings. However, in a more sophisticated
study, Millar and Watkinson ( 1983) found that when the “false alarm” rate
was taken into account, significant differences occurred on recognition of the
Anesthesia 131

target words. Loftus, Schooler, Loftus, and Glauber (1985) conducted a sim-
ilar (if slightly less rigorous) experiment with only a single, highly knowledge-
able and motivated subject (one of the authors). No significant results were
found on three recognition tests. As an afterthought, the authors tested one of
the doctors present at the operation and found that her recognition scores also
did not differ from chance. The authors concluded that the results “still leave
open the possibility that information presented under anesthesia leaves some
lasting impression that cannot be revealed in tests of retention that require
remembering to be deliberate or intentional.”
Goldmann (1986) gave two groups of patients a preoperative “general
information” questionnaire. This included questions such as “What is the
blood pressure of an octopus?’ The experimental group listened to the an-
swers during anesthesia. Both groups received suggestionsfor an easy recovery
as well as a suggestion of the patient’s own choosing. None of the patients
recalled hearing the answers, but the experimental group performed signifi-
cantly better on a postoperative recognition test. As with the source amnesia
experienced with hypnosis, patients acquired new knowledge but were uncer-
tain about where or how they had learned it. Unlike hypnotic source amnesia,
neither hypnosis nor a posthypnotic cue aided the patients in recalling how
they had learned the new facts.

Recall
Reports of recall in the anesthetic literature are rare. Retrospective studies
have often reported an incidence of recall around 2% (Hutchinson, 1960).
Recall has been associated with the absence of premedication and volatile an-
esthetics, light anesthesia, equipment failures, poor anesthetic technique
(Breckenridge& Aitkenhead, 1983), and highly salient events. For many rea-
sons, cases have been difficult to verify. As with dreaming, patients may feel
as if something happened but have no specific recall of the event.
If evidence of dreaming were based solely on recall, we might conclude
that dreams are rare. However, most people believe that we dream far more
than we can remember and that the failure to recall does not suggest that
dreaming did not occur. The failure is consideredto be one of retrieval. Several
factors associated with dream recall are when and how a subject is awakened;
the setting of the awakening; the motivation, personality, and cognitive style
ofthe subject;the demand characteristicsofthe interview; the rapport between
subject and interviewer; and the characteristicsof the dream itself. The charac-
teristics of the dream include aspects of dream content that relate to salience
and intrapsychic conflict.
If factors that influence the recall of dreams also influence the recall of
awareness under anesthesia, then perioperative events will most likely be re-
called if the patient is awakened quickly after the operation, after hearing
something highly salient (that will not cause intrapsychic conflict) during the
operation, achieves good rapport with the interviewer, and is both motivated
and interested in reporting.
132 LES GOLDMANN

When prospective studies have been conducted (with the exception of

Levinson, 1965), intraoperative stimuli have not been highly salient. Levin-
son's study is perhaps the most famous in the literature: Levinson created a
mock perioperative crisis, telling patients under deep ether anesthesia that he
did not like their color and that something was wrong with their breathing.
Under hypnosis, 40% had specific recall, and 40% became highly anxious.
Other researchers, for ethical reasons, have used much less salient material.
For example, McIntyre (1966) repeatedly played a tape about a cleaning
woman who had been robbed of $15 to women undergoing gynecological or
obstetric operations. Patients had no recall of the tape. Other authors have
used music (Brice, Hetherington, & Utting, 1970)or words (Eisele, Weinreich,
& Bartle, 1976; Trustman, Dubovsky, & Titley, 1977) and have found little
evidence supportive of recall. The use of highly salient stimuli are likely to
create intrapsychic conflict, anxiety, or an unwillingness to report, thus ren-
dering recall more difficult.
Because recall is more affected by drugs, context, mood, and the conduct
of the interview than forms of remembering that are less intentional, such as
recognition (Godden & Baddeley, 1975),evidence for awareness based on re-
call is slim. Two studies (Goldmann, 1986; Goldmann et al., 1987) provide
additional support for the notion that evidence for recall, even when efforts
are made to maximize its occurrence, is both difficult to find and may be un-
covered at the risk of psychological distress to the patient.
Returning to the study of cardiac patients mentioned earlier (Goldmann
et al., 1987), 7 of the 30 patients (23%) reported recall. Three reports (10%)
included verbatim recall of conversations that were later corroborated. High
postoperative anxiety was significantly associated with patients reporting re-
call. These patients had not been more anxious than the rest when tested pre-
operatively. All patients who reported recall became anxious during their in-
terviews and believed that they had been conscious during the incidents they
reported. Memories were both verbal and kinesthetic. Several were enhanced
with hypnosis.
In a double blind trial with premedicated spontaneously breathing pa-
tients, little eyidence for recall was found when intraoperative stimuli of low
and moderate salience were played to the patients. Information thought to be
salient to the patient was gleaned during the preoperative interview and was
used in making the intraoperative tape. Of 25 patients, 3 reported images un-
der hypnosis that could be associated with the taped message. Two of these
incidents follow:
Taped message: You have a lovely garden.
Interviewer: Does anything else come to mind?
Patient: Tomatoes and lettuce.
Interviewer: Anything about tomatoes and lettuce?
Patient: We are trying to grow them in our garden.
Taped message;You left nursing to have a lovely little girl.
Anesthesia 133

Interviewer: Does anything come to mind that was said to you or about
you in theater [the operating room]?
Patient: No. I am just thinking about my family.
These messages were designed not to stress the patient. Because the patient’s
hearing was not completely occluded, intraoperative events that were of
greater salience may have overshadowed the prerecorded messages.
Finally, in a study designed to maximize recall, the experimenter was
not blind to the intraoperative stimulus. All patients were women, and all were
good hypnotic subjects. The same suggestion was used throughout: “When
you awaken from your anesthetic, you will believe for a moment that you
have green hair.” The statements of 3 patients are suggestive of the taped mes-
sage. Only the excerpt from the first interview occurred under hypnosis.
Patient A
Interviewer: Anything at all. Anything come to mind?
Patient: Well, I dreamt of green strands.
Patient B
Interviewer: Does anything come to mind when you think about your
operation?
Patient: No.
Interviewer: What about now. Anything at all?
Patient: I keep thinking about green things.
Patient C
Interviewer: How are you this morning?
Patient: Pretty good.
Interviewer: Everything went well?
Patient: Great. I’m leaving this morning. I can’t wait to wash my hair.
Even the use of an informed interviewer,good hypnotic subjects, and hypnosis
does not supply incontrovertible evidence for recall. The connections, how-
ever, are intuitive and plausible.

CONCLUSlON
It is clear from the previous evidence that recollection of perioperative events
decreases as the intentionality of retrieval strategies increases. Eich ( 1984)
wrote that “events that occur while a person is asleep, anesthetized, or selec-
tively attending to other ongoing events . . . are rarely revealed in tests of
retention that require remembering to be deliberate or intentional.” Similar
findings have been found with Korsakoff patients (Graf, Squire, & Mandler,
1984). Evidence for learning can be found where otherwise it might not have
been when methods such as spelling, puzzle completion, or word recognition
are used. These methods supply cues and contextual elements. Hypnosis may
enhance remembering by reinstating contextual elements, by allowing the pa-
tient to supply particular affective cues, and by improving rapport. However,
134 LES GOLDMANN

even with the aid of hypnosis, only a tiny fraction of the events that occur
under anesthesia may be available for recall, although they may perhaps be
remembered in another way. It is possible that cognitive processing under gen-
eral anesthesia occurs outside conscious awareness.
Evidence from subliminal perception, blindsight, and dichotic listening
as well as research with amnesic patients (Dixon, 1981;Eich, 1984;Marcel,
1983)suggests that information processing occurs both consciously and non-
consciously. Nonconsciously processed events are not available for recall, al-
though they may influence behavior (Mattis & Kovner, 1984,p. l 18). Conse-
quently, a careless remark by a surgeon may result in a longer hospital stay
or slower healing. Conversely, a patient's recovery may be improved by the
judicious use of suggestions. As with hypnosis, the suggestion may not be re-
called, but its influence on the patient's subsequent behavior may nevertheless
be important and valuable as a tool in the aid of healing.

References
Bennett, H. L., Davis, H. S., &Giannini, J. A. (1985). Nonverbal response to intraoper-
ative conversation. British Journal ofAnaesthesia, 57, 174- 179.
Blacher, R. S. (1975). On awakening paralyzed during surgery: A syndrome of trau-
matic neurosis. Journal of the American Medical Association, 234,67-68.
Bonke, B., Schmitz, P. I. M., Verhage, F., & Zwaveling, A. (1986). Clinical study of
so-called unconscious perception during general anaesthesia. British Journal of
Anaesthesia, 58, 957-964.
Bonnett, 0. T. (1966). Effects of positive suggestions on surgical patients. Pacific Jour-
nal of Medicine and Surgery, 297-300.
Breckenridge, J. L., & Aitkenhead, A. R. (1983). Awareness during anaesthesia: A re-
view. Annals of the Royal College of Surgeons, 65,93-96.
Brice, D. D., Hetherington, R.R., & Utting, J. E. (1970). A simple study of awareness
and dreaming during anaesthesia. British Journal ofAnaesthesia, 42, 535-54 1.
Cheek, D. B. (1962). Importance of recognizing that surgical patients behave as though
hypnotized. American Journal of Clinical Hypnosis, 4,227-236.
Cheek, D. B. (1964). Surgical memory and reaction to careless conversation. American
Journal of Clinical Hypnosis, 6,237-240.
Cheek, D. B. (1979, November). Awareness of meaningful sounds under general anes-
thesia: Considerations and a review of the literature 1959-1979. Paper presented
at the Annual Scientific Meeting of the American Society of Clinical Hypnosis.
Crawford, J. S. (197 1). Awareness during operative obstetrics during general anaesthe-
sia. British Journal ofAnaesthesia, 43, 179- 182.
Dixon, N. ( 198 1). Preconsciousprocessing. New York: Wiley.
Dubovsky, E. L., & Trustman, R. (1976). Absence of recall after general anesthesia.
Anesthesia and Analgesia, 55, 696-67 1.
Eich, E. (1984). Memory for unattended events: Remembering with and without
awareness. Memory and Cognition, 12(2), 105-1 1 1.
Eisele, V., Weinreich, A., & Bartle, S. (1976). Perioperative awareness and recall. Anes-
thesia and Analgesia, 55, 5 13-5 18.
Evans, C., & Richardson, P. H. (1988). Improved recovery and reduced postoperative
stay after therapeutic suggestions during general anaesthesia. Lancet II: 49 1-493.
Anesthesia 135

Federn, P. (1953). A dream under general anesthesia. In P. Federn (Ed.), Egopsychol-

ogy and the psychoses (pp. 100-103). London: Littlemore.
Godden, D. R., & Baddeley, A. D. (1975). Context dependent memory in two natural
environments: On land and underwater. British Journal of Psychology, 66, 325-
331.
Goldmann, L. ( 1986). Awareness under general anaesthesia. Unpublished doctoral
dissertation, Cambridge University, London, U.K.
Goldmann, L., & Levey, A. B. (1986). Orienting under anaesthesia. Anaesthesia, 41,
1056-1057 (letter).
Goldmann, L., Shah, M. V., & Hebden, M. W. (1987). Memory of cardiac anaesthesia:
Psychological sequelae in cardiac patients of intra-operative suggestions and oper-
ating room conversation. Anaesthesia, 42, 596-603.
Graf, P., Squire, L. R., & Mandler, G. (1984). The information that amnesic patients
do not forget. Journal of Experimental Psychology, 10, 164- 178.
Huchinson, R. ( 1960). Awareness during surgery: A study of its incidence. British Jour-
nal of Anaesthesia, 33,463-469.
Levinson, B. W. ( 1965). States of awareness during general anaesthesia. British Journal
of Anaesthesia, 3 7, 544-546.
Loftus, E., Schooler, J., Loftus, G., & Glauber, D. (1 985). Memory for events occurring
under anaesthesia. Acta Psychology, 59, 123- 128.
Lunn, J. N. (1 982). Lecture notes on anaesthetics (2nd ed.). Blackwell: Oxford.
Marcel, A. J. (1 983). Conscious and unconscious perception: An approach to the re-
lations between phenomenal experience and perceptual processes. Cognitive Psy-
Chology, 15, 238-300.
Mattis, S., & Kovner (1984). Amnesia is as amnesia does: Toward another definition
of anterograde amnesias. In L. Squire & N. Butters (Eds.), Neuropsychology of
memory (pp. 1 15- I2 1). London: Guilford Press.
McIntyre, J. W. R. (1966). Awareness during general anesthesia: Preliminary observa-
tions. Canadian Anaesthetists Society Journal, 13, 495-499.
Millar, K., & Watkinson, N. (1983). Recognition of words presented during general
anaesthesia. Ergonomics, 26, 585-594.
Pearson, R. E. ( 196 1 ). Response to suggestionsgiven under general anaesthesia. Ameri-
can Journal of Clinical Hypnosis, 4, 106- 1 14.
Pearson, R. E. ( 1965). Response to suggestions given under general anaesthesia. British
Journal ofAnaesthesia, 37, 544-546.
Trustman, R., Dubovsky, S., & Titley, R. (1977). Auditory perception during general
anesthesia-myth or fact? International Journal of Clinical and Experimental
Hypnosis, 25, 88-105.
Weinberger, N. M., Gold, P. E., & Sternberg, D. B. (1984). Epinephrine enables Pav-
lovian fear conditioning under anaesthesia. Science, 223,605-606.
 CHAPTER 10

INSOMNIA:
THE PATIENT AND THE PILL

WALLACE B. MENDELSON

In one sense, we have learned a great deal about patients with chronically
disturbed sleep; in another sense, we know very little. A major advance in the
last decade has been the growing recognition that poor sleep is often a symp-
tom of a series of diagnosable, treatable conditions. Complaints of poor qual-
ity sleep may result, for instance, from sleep apnea, nocturnal myoclonus,
depressive illness, and many other causes (Mendelson, 1987). On the other
hand, there remains a core of patients with chronically disturbed sleep in
whom none of these disorders are found. In this article, I shall use the term
insomniacs to refer to these unhappy individuals. I would like to suggest that
the study of insomniacs leads to knowledge of how hypnotics act and that,
conversely, the study of hypnotics provides insight into the nature of insom-
nia. In summary, I will argue that the clinical effects of hypnotics cannot be
well-understood if they are assessed only in terms of enhancing sleep defined
by the electroencephalogram (EEG). On the contrary, hypnotic efficacy may
involve alterations in the patient’s perception of his or her state of conscious-
ness. Similarly, the development of tolerance may be influenced not only by
pharmacologic issues, but also by a learned process. Before we look at these
issues, however, let us briefly review some general thoughts about insomnia.
Insomnia is a complaint, a symptom and not an illness. In the Associa-
tion of Sleep Disorder Centers nosology (1979), a variety of causes of poor
sleep are listed. After eliminating the many physiological (e.g., sleep apnea),
psychiatric (e.g., depression) and chronobiologic (e.g., phase lag) problems, at
least two major categories remain. The first of these is subjective disorder of
initiating and maintaining sleep (DIMS) without objective findings. Although
the exact terminology for this condition will probably change in forthcoming
139
140 WALLACE B. MENDELSON

revisions, the thought remains the same, that is, that many insomniacs com-
plain strongly about poor sleep while showing minimal disturbances during
polysomnography. Thus, there can be a substantial disparity between patients’
subjective reports of their sleep and objective measures from the same night.
A second form of insomnia is persistent psychophysiological DIMS.
In these patients, internal cues (ruminative worries about not sleeping) and
external cues (the bedroom environment) lead to poor sleep. Their EEG mea-
sures of sleep may be relatively normal or disturbed. In either event, however,
patients’ morning estimates of their sleep are relatively consistent with the
EEG findings. In clinical situations, the distinction between subjective and
persistent psychophysiologicalinsomnia is often unclear, because conditioned
elements may be superimposed on the subjective disorder. Nonetheless, the
basic principle that insomnia may involve a dissociation of subjective experi-
ence from objective measures, as well as conditioned responses, sets the back-
ground for the discussion of hypnotics. Let us now look at sleep studies of
insomnia and at data on possible physiological differences between insomni-
acs and control subjects.

NOCTURNAL SLEEP AND DAYTIME WAKEFULNESS

OF INSOMNIACS
From the 1950s through the late 1970s, many laboratories reported sleep data
from insomniacs (Institute of Medicine, 1979). However, these studies are
difficult to assess because patients were selected on the basis of history and not
on the basis of complete polysomnography to rule out other disorders (e.g.,
sleep apnea). Not surprisingly, later studies that did include polygraphic
screening found objective sleep disturbance in insomniacs to be relatively
small. One study of 10 insomniacs reported that they had increased intermit-
tent waking time and lower sleep efficiency compared with control subjects
(Mendelson, Garnett, Gillin, & Weingartner, 1984). The only significant
difference in another study was more early morning waking time for insomni-
acs, with trends toward lower sleep efficiency and less total sleep (Mendelson,
James, Garnett, Sack, & Rosenthal, 1986). Gillin and Mendelson (1981) re-
viewed 25 hypnotic efficacy studies and found that baseline total sleep was
383 min in insomniacs. This is similar to normal control values of 389 min
and 383 min in the two previously cited studies. Similarly, Reynolds et al.
( 1984) reported that patients with persistent psychophysiological DIMS had a
sleep efficiency of 8 1.5% and total sleep of 359.8 min.
Many studies of insomnia have reported decreased slow wave sleep (e.g.,
Gillin, Duncan, Pettigrew, Frankel, & Snyder, 1979), which has been sug-
gested to play a crucial role in insomnia (Gaillaird, 1978). Although this is
possible, such a theory would have to encompass the many other disorders
characterized by decreased slow wave sleep, such as depression and schizo-
phrenia (Mendelson, 1987). Moreover, it is difficult to rationalize this view
with the observation that benzodiazepine hypnotics decrease slow wave sleep
and yet result in improved subjective reports of sleep. Finally, electronic analy-
Insomnia 141

ses of sleep EEGs in insomniacs have not confirmed a decrease in either slow
wave or total EEG power (Mendelson, in press). In summary, EEG sleep stud-
ies of insomnia have found relatively few changes in the sleep of insomniacs.
This may reflect the measures that we use; perhaps more subtle measures
would reveal differences. On the other hand, this may resemble the Sherlock
Holmes story in which a mystery was solved on the basis of the absence of a
dog’s bark. The important point might be the absence of substantial differ-
ences between the EEG sleep of insomniacs and normal subjects.
Insomniacs traditionally complain not only of disturbed sleep but also
of daytime fatigue or sleepiness.An objective measure of daytime wakefulness
may be obtained from the Multiple Sleep Latency Test (MSLT), which records
the mean latency to sleep during four or five 20-min periods in which a patient
is placed in bed in a darkened room (Mendelson, 1987). Interestingly, MSLT
studies in both young and older insomniacs have shown normal daytime
wakefulness (Dement, Seidel, & Carskadon, 1982, 1984; Mendelson et al.,
1984, 1986; Stepanski, Zorick, Roehrs, Young, & Roth, 1988). There are at
least two possible interpretations of these observations. First, given the rela-
tively mild sleep disturbance shown in EEG studies, the normal MSLTs may
confirm that insomniacs are not really sleep deprived. And second, insomni-
acs may indeed suffer from some degree of chronic mild sleep deprivation, but
they may respond differently from normals in terms of how this affects their
daytime wakefulness. At least two observations support this latter view. One
comes from a study by Stepanski, Lamphere, Badia, Zorick, and Roth (1984),
which assessed the relation of nocturnal sleep disturbance and daytime wake-
fulness in patients with sleep apnea, nocturnal myoclonus, and insomnia. In
the first two groups, the number of nocturnal arousals correlated with de-
creased daytime wakefulness. In normal control subjects (who had few arous-
als), there was no significant relation. In the insomniacs, however, the number
of nocturnal arousals was positively associated with increased daytime wake-
fulness. The type of arousal was also critical: Very brief arousals (increased
EEG frequency and electromyograph [EMG] amplitude) were associated
with daytime wakefulness, whereas more substantial awakenings were asso-
ciated with sleepiness. A later study by Stepanski et al. (1988) also found
that shorter sleep times in insomniacs correlated with increased daytime
wakefulness.
A second line of evidence suggesting that insomniacs may respond to
sleep deprivation differently than normal subjects comes from Bonnet ( 1986),
who compared recovery sleep after 64 hours of total sleep loss in older insom-
niacs and control subjects. In both groups, sleep efficiency was very high
(about 97%)on the first recovery night. In the normal subjects, values returned
toward baseline by the second recovery night, whereas for the insomniacs this
did not take place until the fourth night. Thus, the issue of whether insom-
niacs are not sleep deprived or whether they suffer from some form of mild
chronic sleep deprivation to which they respond abnormally remains to
be settled.
142 WALLACE 8. MENDELSON

THE PERcEPTioN OF SLEEP

Another question that arises is whether insomniacs differ from normals in their
perception of the experience of sleeping. This idea was first expressed by Recht-
schaffen (1968), who awakened poor sleepers and control subjects 10 min after
the first sleep spindle and asked them whether they had been awake or asleep.
The normal subjects tended to report that they had been asleep; the insomniacs
reported that they had been awake. Coates et al. (1983) reported a similar phe-
nomenon. Mendelson et al. ( 1984, 1986) found no differences in responses be-
tween insomniacs and control subjects when they were awakened by an 80 dB
tone but did find a disparity after arousal by a tone of gradually increasing inten-
sity. In the periods between the forced awakenings, the insomniacs and control
subjects fell asleep at about the same rate and had equal amounts of sleep. The
subjective estimates of the insomniacs, however, were only about half as long
as the estimates of the control subjects. These data, taken together, raise the
possibility that an important aspect of insomnia might be a misperception of
the insomniac’s state of consciousness vis-&vis the EEG.

THE ACTIONS OF HYPNOTICS

Having reviewed the sleep and wakefulness of insomniacs, we will now exam-
ine the effects of hypnotics. Clinically used hypnotics comprise many different
pharmacologic classes; indeed, an intriguing mystery is how drugs of such dis-
parate biochemical qualities can have such similar effects. We will consider
hypnotic actions on both individual sleep stages and on total sleep.
Until the 1960s, the dominant clinical hypnotics were the barbiturates,
which greatly suppress REM sleep. They were reported variously to not affect
(Baekeland, 1967; Kay, Jasinski, Eisenstein, & Kelly, 1972) or to decrease
(Feinberg, Wender, Koresko, Gottlieb, & Piehuta, 1969) slow wave sleep.
Since that time, benzodiazepines have become the most widely prescribed
hypnotics. In general, they have minimal effects on REM sleep but profoundly
decrease slow wave sleep (Mendelson, 1987). Because it has traditionally been
thought that slow wave sleep is particularly deep and restful, it seems some-
what paradoxical that drugs which suppress slow wave sleep result in subjec-
tively improved sleep. This dilemma may be resolved if slow wave sleep is not
necessarily associated with a subjective sense of deep or good quality sleep.
When arousal thresholds to relatively meaningless stimuli such as electronic
tones are measured, slow wave sleep does indeed seem to be the “deepest”
stage. On the other hand, the amount of Stage 4 sleep in normal volunteers has
been positively associated with morning reports of light sleep (Saletu, 1975).
Volunteers reported having previously been awake in 56% of forced awaken-
ings from Stage 2 sleep compared with 77% of forced awakenings from slow
wave sleep (Sewitch, 1984). In both insomniacs and control subjects, slow
wave sleep has been associated with retrospectively poorer quality sleep on a
morning questionnaire (Mendelson et al., 1986). However, data also suggest
the opposite. Healthy older women (but not men) have been found to associ-
ate slow wave sleep with reported soundness of sleep (Hoch et al., 1987). The
Insomnia 143

issue remains open, but it seems possible that slow wave sleep is not necessarily
experienced subjectively as the deepest or most restful sleep.
Studies of the effects of hypnotics on total sleep mirror the data on base-
line sleep among insomniacs: Just as the deficit in total sleep in insomniacs
is small, the enhancement of sleep by hypnotics is also modest. Gillin and
Mendelson ( 198 l), in a review of hypnotic efficacy studies, found that the
mean increase in total sleep time was 35 min. An often-cited study of fluraze-
pam (which at the time was the most widely prescribed hypnotic) reported an
increase in total sleep of only 6-8% over 1 month (Kales, Kales, Bixler, &
Scharf, 1979, with sleep latency significantly decreased only during Nights
1 1 - 1 3. Thirty milligrams of flurazepam given to insomniacs over 5 weeks en-
hanced total sleep by a mean of 2 1 min (Mitler, Seidel, Van Den Hoed, Green-
blatt, & Dement, 1984). Roehrs, Zorick, Kaffeman, Sicklesteel, and Roth
(1982) reported an increase in total sleep of 29 min in a study using 15 mg of
flurazepam. In spite of these relatively modest improvements in EEG-defined
sleep, approximately 20 million prescriptions are written for hypnotics every
year, and another 3%or 4%ofthe population takes over-the-counter sedatives.
Thus, it seems that many people feel they are helped by these agents. Studies
of subjective responses to hypnotics bear this out. Rickels (1983) found that,
in six of seven studies of flurazepam, patients reported that the drug was more
beneficial than a placebo. Significantlyimproved quality of sleep was also sub-
jectively described in five of the seven studies. One possible interpretation is
that insomniacs suffer from a long-term form of sleep deprivation and that
the modest increase of total sleep from hypnotics corrects this problem. An
alternative view is that the subjective benefit of hypnotics is very dispropor-
tionate to the small hypnotic-induced improvement in sleep (in much the
same way that the subjective distress of insomniacs is disproportionately
greater than their modest decrements in sleep). If the latter is the case, several
interesting hypotheses need to be considered:

1 . Hypnotics improve sleep in polygraphic measures that are more

subtle than those used to date. Evidence increasingly suggests that
sleep continuity, rather than total sleep, may be more important for
the restorative value of sleep. On the other hand, careful studies of
sleep disruption in insomniacs have shown less discontinuity than
might have been expected. Stepanski et al. (1984) found that, in
three of four categories of sleep interruption, insomniacs actually
had fewer arousals than patients with sleep apnea or myoclonus and,
indeed, had amounts similar to those of control subjects. Insomniacs
reported more disturbance only in shifts to a lighter stage of sleep.
Insofar as the sleep of insomniacs may not be substantially more
disrupted than that of control subjects, the argument that hypnotic
therapeutic effects derive from improved continuity is weakened.
2. Hypnotics alter a physiological process not measured by clinical
polysomnography. Such a measure might be core temperature. The
144 WALLACE B. MENDELSON

early studies of Monroe (1967) and more recent work (Mendelson et

al., 1984)have reported that insomniacs have higher mean rectal
temperatures at sleep onset and throughout the night. It seems
unlikely that this merely reflects increased waking time during the
night. In the first 2 hr after sleep onset, insomniacs had the same or
less waking time than control subjects, yet their temperatures
remained higher. Hypnotics are known to lower core temperature
(Pleuvry, Maddison, Odeh, & Dodson, 1980).Slow wave sleep can
be enhanced by increasing core temperature (Home, Moore, Reid,
& Shackell, 1985). One might speculate that the benzodiazepine-
induced decrease in slow wave sleep results from drug-induced
changes in temperature. This view differs from the speculation that
the minimal slow wave sleep among insomniacs results from their
relative lack of decline in core temperature after sleep onset
(Sewitch, 1987). Another confounding variable in studying this
system is that sleep is very sensitive to ambient temperature
(Haskell, Palca, Walker, Berger, & Heller, 1981; Muzet, Libert, &
Candas, 1984). Exploration of this area must continue.
3. Hypnotics alter processes not susceptible to physiological
measurement. It is conceivable that hypnotics change the
insomniac’s memory of sleep. This might occur through two types of
mechanisms. First, a direct pharmacologic action might affect
episodic memory. Benzodiazepinescan decrease the ability to learn
new material, although opinions differ on whether the ability to
retrieve and use previously learned information is affected
(Mendelson, Weingartner, Greenblatt, Garnett, & Gillin, 1982;
Roehrs, Kribbs, Zorick, & Roth, 1986). Thus, hypnotics may alter
the patient’s ability to remember his or her experience. There is also
increasing evidence that an anterograde amnesia occurs for the time
immediately before sleep onset. These observations have been used
to explain the complaints of memory trouble made by narcoleptics.
Similarly, Carskadon et al. (1976) found that, although insomniacs
exaggerated most aspects of their sleep compared with EEG data,
they actually underreported the number of awakenings throughout
the night. These awakenings may not have been remembered as
clearly because of an anterograde amnesia after the return to sleep.
This hypothesis must be assessed very carefully. For instance, word
lists learned at night by subjects receiving 0.25 mg triazolam are
remembered better the next morning than those by subjects
receiving a placebo (Mendelson et al., 1989). Such an effect may
result from the drug-induced sedation that minimizes the amount of
subsequent distracting material.
As we have just mentioned, hypnotics affect memory processes and the
ability to perform a variety of tasks. These effects have been studied exten-
sively because of the concern that patients might be more susceptibleto auto-
Insomnia 145

mobile accidents (Betts & Birtle, 1983; Binnie, 1983;Linnoila, 1978). Indeed,
it could be argued that no clinical hypnotic exists that does not produce cogni-
tive alterations. Although cognitive changes have traditionally been viewed as
side effects, they may be a daytime manifestation of the therapeutic effect, that
is, alterations in the patients’ experiences of whether they are awake or asleep.
A number of studies have shown that benzodiazepinesalter auditory arousal
during sleep. Preliminary evidence suggests that, during forced awakenings 10
min after the first sleep spindle, insomniacs are more likely to report having
been asleep after receiving flurazepam than placebo (Mendelson et al., 1987).
Certainly this area appears to be ripe for exploration.

TOLERANCE TO HYPNOTICS
We have previously described how conditioned responses might result in the
complaint of insomnia, either in persistent psychophysiologicalDIMS or as a
complication to subjective insomnia. Conditioned responses may also play a
role in clinical tolerance to hypnotics. Just as the efficacy of hypnotics must
be measured both physiologically and subjectively,tolerance may be a combi-
nation of traditional pharmacologicissues (i.e., pharmacokineticsand dynam-
ics) and learned responses. Classical conditioning theory could be applied to
suggest that environmental cues associated with drug administration may re-
sult in conditioned responses that are opposite to the unconditioned effect of
the drug (Siegel, 1975; Spielman, Caruso, & Glovinsky, 1987). These “com-
pensatory” responses, which contrast with those resembling the uncondi-
tioned drug response, have been described in a number of studies. King, Bou-
ton, and Musty (1987), for instance, gave the sedative benzodiazepine
midazolam to rats at 48-hr intervals until tolerance was established. When
they gave saline to the rats in the same environment used for drug injection,
animals tested 2 min after administration became hyperactive. Similar com-
pensatory responses have been demonstratedfor motor behavior after chlordi-
azepoxide (File, 1982) and for the hypothermic effect of ethanol (Melchior
& Tabakoff, 1985). These observations have a number of interesting clinical
implications. As Spielman et al. (1987) pointed out, the conditioning theory
of tolerance predicts that the occasional administration of placebo (which is
in effect an extinction trial) would be superior to continuous drug administra-
tion in maintaining efficacy. Occasional placebos might also be more benefi-
cial than drug holidays. The clinical utility of this principle has not yet been
established. However, it exemplifiesonce again the complexity of the relation
between the patient and the pill.

References
Association of Sleep Disorder Centers. ( 1979). Diagnostic classification of sleep and
arousal disorders. Sleep, 2, 1 - 137.
146 WALLACE B. MENDELSON

Baekeland, F. ( 1967). Phenobarbital and dextroamphetamine sulfate: Effects on the

sleep cycle in man. Psychopharmacologia, I I, 388-396.
Betts, T. A., & Birtle, J. (1983). Effect of two hypnotic drugs on actual driving perfor-
mance next morning. British Medical Journal, 285,852.
Binnie, G. A. (1983). Psychotropic drugs and accidents in a general practice. British
Medical Journal, 287, 1349- 1350.
Bonnet, M. H. (1986). Effect of 64 hours of sleep deprivation upon sleep in geriatric
normals and insomniacs. Neurobiology ofAging, 7, 89-96.
Carskadon, M. A., Dement, W. C., Mitler, M. M., Guilleminault, C., Zarcone, V. P.,
& Spiegel, R. ( 1976). Self-reports versus sleep laboratory findings in 122 drug-free
subjects with complaints of chronic insomnia. American Journal of Psychiatry,
133, 1382-1383.
Coates, T. J., Killen, J. D., Silberman, S., Marchini, J., Hamilton, S., & Thoresen,
C. E. (1983). Cognitive activity, sleep disturbance, and stage specific differences
between recorded and reported sleep. Psychopharmacology, 20,243.
Dement, W. C., Seidel, W., & Carskadon, M. A. (1982). Daytime alertness, insomnia
and benzodiazepines. Sleep, 5, 528-545.
Dement, W. C., Seidel, W., & Carskadon, M. A. (1984). Issues in the diagnosis and
treatment of insomnia. Psychopharmacology, 2(Suppl.), 1 1-43.
Feinberg, I., Wender, P. H., Koresko, R. L., Gottlieb, F., & Piehuta, J. (1969). A differ-
ential effect of chlorpromazine and phenobarbital on EEG sleep patterns. Journal
ofpsychiatric Research, 7, 101-109.
File, S. E. (1982). Development and retention of tolerance to the sedative effects of
chlordiazepoxide: Role of apparatus cues. European Journal of Pharmacology, 81,
637-643.
Gaillaird, J. M. ( 1978). Chronic primary insomnia: Possible physiopathological in-
volvement ofslow wave sleep deficiency. Sleep, 1, 133-147.
Gillin, J. C., Duncan, W., Pettigrew, K. D., Frankel, B. L., & Snyder, F. (1979). Suc-
cessful separation of depressed, normal, and insomniac subjects by EEG sleep
data. Archives of General Psychiatry, 36, 85-90.
Gillin, J. C., & Mendelson, W. B. ( 1 98 1). Sleeping pills: For whom? When? How long?
In G. C. Palmer (Ed.), Neuropharmacology of central nervous system and behav-
ioral disorders (pp. 285-3 16). New York: Academic Press.
Haskell, E. H., Palca, J. W., Walker, J. M., Berger, R. J., & Heller, H. C. (1981). The
effects of high and low ambient temperature on human sleep stages. Electroen-
cephalography and Clinical Neurophysiology, 51,494-50 I.
Hoch, C. C., Reynolds, C. F., Kupfer, D. J., Berman, S. R., Houck, P. R., & Stack,
J. A. ( 1987). Self-report versus recorded sleep in healthy seniors. Psychophysiol-
ogy, 24,293-299.
Horne, J. A., Moore, V. J., Reid, A. J., & Shackell, B. S. (1985). Waking body tempera-
ture manipulation and subsequent sleep (SWS). Sleep Research, 14, 15.
Institute of Medicine. ( 1979). Sleeping pills, insomnia and medical practice (Publica-
tion No. 10M-79-04).Washington, DC: National Academy of Sciences.
Kales, A,, Kales, J. D., Bixler, E. O., & Scharf, M. B. ( 1975). Effectiveness of hypnotic
drugs with prolonged use: Flurazepam and pentobarbital. Clinical Pharmacology
and Therapeutics, 18(3), 356-363.
Kay, D. C., Jasinski, D. R., Eisenstein, R. B., & Kelly, 0.A. (1972). Quantified human
sleep after pentobarbital. Clinical Pharmacology and Therapeutics, 13, 22 1-23 1 .
King, D. A., Bouton, M. E., & Musty, R. E. (1987). Associative control of tolerance to
the sedative effects of a short-acting benzodiazepine. Behavioral Neuroscience,
101, 104-1 14.
Linnoila, M. (1978). Psychomotor effects of drugs and alcohol on healthy volunteers and
psychiatric patients. Advances in Pharmacology and Therapeutics,8,235-249.
Melchior, C. L., & Tabakoff, B. ( 1 985). Features of environment-dependent tolerance
to ethanol. Psychopharmacology, 87,94- 100.
Mendelson, W. B. (1987). Human sleep: Research and clinical care. New York: Ple-
num Press.
insomnia 147

Mendelson, W. B. (in press). Do studies of sedative/hypnotics suggest the nature of

chronic insomnia? In R. Godbout & J. Montplaisir (Eds.), Sleep and biological
rhythms: Application to psychiatry. New York: Oxford University Press.
Mendelson, W. B., Garnett, D., Gillin, J. C., & Weingartner, H. (1984). The experience
of insomnia and daytime and nighttime functioning. Psychiatry Research, 12, 235-
250.
Mendelson, W. B., James, S. P., Garnett, D., Sack, D. A., & Rosenthal, N. E. (1986).
A psychophysiological study of insomnia. Psychiatry Research, 19, 267-284.
Mendelson, W. B., Sicuro, F., Starz, K. E., Gupwartz, K., Dorothea, M., and Weingart-
ner, H. ( 1 989). Effects of two preparations of triazolam on cognition and perfor-
mance in normal volunteers. Sleep Research, 18, 63.
Mendelson, W. B., Stephens, H., Giesen, H., & James, S. P. (1987). Effects of fluraze-
pam on sleep, arousal threshold, and the perception of being asleep. Psychophar-
macology, 95, 258.
Mendelson, W. B., Weingartner, H., Greenblatt, D. J., Garnett, D., & Gillin, J. C.
(1982). A clinical study of flurazepman. Sleep, 5, 350-360.
Mitler, M. M., Seidel, W. F., Van Den Hoed, J., Greenblatt, D. J., & Dement, W. C.
( 1984). Comparative hypnotic effects of flurazepman, triazolam and placebo: A
long-term simultaneous nighttime and daytime study. Journal of Clinical Psycho-
pharmacology, 4, 2-1 3 .
Monroe, L. J. (1967). Psychological and physiological differences between good and
poor sleepers. Journal of Abnormal Psychology, 72,255-264.
Montlaisir, J., & Godbout, R. (Eds.). (in press). Sleep and biological rhythms: Applica-
tion to psychiatry. New York: Oxford University Press.
Muzet, A., Libert, J.-P., & Candas, V. (1984). Ambient temperature and human sleep.
Experientia, 40, 42 5-429.
Pleuvry, B. J., Maddison, S. E., Odeh, R. B., & Dodson, M. E. (1980). Respiratory and
psychological effects of oral temazepam in volunteers. British Journal of Anaes-
thesiology, 52, 90 1.
Rechtschaffen, A. (1968). Polygraphic aspects of insomnia. In Gastaut, H., Lugaresi,
L., Berti, G., and Coccagno (Eds.), The abnormalities of sleep in man (pp. 109-
125). Bologna, Italy: Gaggi.
Reynolds, C. F., Taska, L. S., Sewitch, D. E., Restifo, K., Coble, P. A., & Kupfer, D. J.
( 1984). Persistent psychophysiologic insomnia: Preliminary research diagnostic
criteria and EEG sleep data. American Journal ofPsychiatry, 141, 804-805.
Rickels, K. ( 1983). Clinical trials of hypnotics. Journal of Clinical Psychopharmacol-
Ogy, 3, 133-142.
Roehrs, T., Kribbs, N., Zorick, F., & Roth, T. (1986). Hypnotic residual effects of
benzodiazepines with repeated administration. Sleep, 9, 194-1 99.
Roehrs, T., Zorick, F., Kaffeman, M., Sicklesteel, B. A., & Roth, T. (1982). Flurazepam
for short-term treatment of complaints of insomnia. Journal of Clinical Pharma-
cology, 22, 290-296.
Saletu, B. (1975). Is the subjectively experienced quality of sleep related to objective
sleep parameters? Behavioral Biology, 13,433.
Sewitch, D. E. (1984). NREM sleep continuity and the sense of having slept in normal
sleepers. Sleep, 7, 147.
Sewitch, D. E. (1987). Slow wave sleep deficiency insomnia: A problem in thermo-
downregulation at sleep onset. Psychophysiology, 24, 200-2 15.
Siegel, S. (1975). Evidence from rats that morphine tolerance is a learned response.
Journal of Comparative Physiology and Psychology, 89,498-506.
Spielman, A. J., Caruso, L. S., & Glovinsky, P. B. (1987). A behavioral perspective on
insomnia treatment. Sleep Disorders, 10, 54 1-553.
Stepanski, E., Lamphere, J., Badia, P., Zorick, F., & Roth, T. (1984). Sleep fragmenta-
tion and daytime sleepiness. Sleep, 7, 18-26.
Stepanski, E., Zorick, F., Roehrs, T., Young, D., & Roth, T. (1988). Daytime alertness
in patients with chronic insomnia compared with asymptomatic control subjects.
Sleep, 1I, 54-60.
 CHAPTER 1 1

THE PERCEPTION OF SLEEP ONSET

IN I N S O M N I A C S AND
N O R M A L SLEEPERS
MICHAEL H. BONNET

The ability to measure the onset of sleep reliably is clearly a centraI component
of modern sleep research. Because behavioral quiescence, muscular relax-
ation, alpha drop, respiratory decrease, and auditory threshold increase gener-
ally show strong agreement, sleep onset may be considered to be well-defined.
Unfortunately, subjective reports of sleep onset frequently do not agree well
with physiological measures. It is not unusual for people to report having been
awake when they are awakened from sleep (Anch, McCoy, & Somerest, 1980;
Campbell & Webb, 198 1;Rechtschaffen, 1968). Some patients, usually those
who are depressed, claim to have had no sleep during nights in which physio-
logical indicators suggest that they had several hours of sleep. One of the most
persistent areas of disagreement exists between objective and subjective re-
ports (from patients with reported difficulty falling asleep) of the length of time
it initially takes to fall asleep at night. At least nine studies have documented
median electroencephalograph(EEG) sleep onset latencies of 51 min com-
pared with subjective median estimates of 82 min in such patients (Carskadon
et al., 1976; Frankel, Coursey, Buchbinder, & Snyder, 1976; Hauri & Cohen,
1977; Mendelson, Garnett, Gillin, & Weingartner, 1984; Monroe, 1967;
Nease, Monroe, Bonnet, & Kramer, 198 1;Olmstead, Hauri, Percy, & Hellek-
son, 1980; Spielman et al., 1980; Tietz, Zorick, Roth, & Kaffeman, 1980).

SLEEP ONSET ESTIMATES IN NORMALS

We initially began our examination of sleep onset by observingcarefully delin-
eated periods of sleep in normal young adults. In all studies reported here, we
140
Sleep Onset 149

I 1 I I I I I 1
2 4 8 16 32 64 128 256
TIME AFTER SLEEP ONSET (MINUTES)

figure 1 Plot of percentage of subjects subjectively reporting themselves to be asleep

at awakening as a function oftime after electroencephalograph (EEG) sleep onset. (The
open circle corresponds to the initial spindle awakening. Used with permission from
Bonnet & Moore, 1982.)

timed EEG intervals from the appearance of the first sleep spindle rather than
from the point of alpha drop because alpha may decrease for reasons other
than sleep onset (like opening the eyes) and because a substantial proportion
of normal young adults do not produce significant alpha at any time. In the
initial study (Bonnet & Moore, 1982), 12 subjects spent 5 nights each in the
laboratory. The subjectswere awakened at specified intervals ranging from the
first sleep spindle and 1 min after the first spindle to 140 min after the first
spindle. When the subjects were awakened, they were asked if they had been
asleep at that point. If they answered no, they were asked how long it had been
since the technician had told them good night. If they answered yes, they were
asked how long it had taken to fall asleep and how long they had been asleep.
Figure 1 plots the percentage of subjectswho reported themselves to be asleep
as a function of how long they had been asleep when awakened. It can be seen
from the figure that, whereas only 40% of the subjects reported being asleep 1
min after the first sleep spindle, 100%of the subjects reported being asleep 25
min after the first sleep spindle. By classical perceptual criteria, the threshold
of sleep onset (i.e., the point at which 50% of the subjects reported being

This work was supported by the Veterans Administrationand by the Sleepwake Disorders
Research Institute. The author would like to acknowledge the many sleepless nights of Sarah
E. Moore and Ralph Downey 111, whose respective master’s thesis and doctoral dissertation are
described at length in this chapter.
150 MICHAEL H. BONNET

.
.. .
301
0

U
20

l o t
I I 1 I I I I I
2 4 8 16 32 64 128 256
TIME AFTER SLEEP ONSET (MINUTES)

Figure 2 Plot of auditory arousal threshold level in the awake state and after various
intervals of sleep. (The open circle corresponds to the initial spindle awakening. Used
with permission from Bonnet & Moore, 1982.)

asleep) would fall between 4 min and 8 min. Subjects were always awakened
by an audiometer using an ascending method of limits procedure. Figure 2
plots auditory arousal thresholds for the same awakenings displayed in Figure
1. It can be seen that auditory thresholds reached 70% of their stable Stage 2
sleep value after 1 min of sleep. The data reinforce the notion that physiologi-
cal changes such as muscle relaxation and auditory threshold increase occur
rapidly at sleep onset but that cognitive-perceptual changes occur more slowly
and that little subjective certainty of sleep occurs for at least the first 8 min of
sleep, even in normal young adults. These young adults could clearly differen-
tiate longer periods of sleep, and many studies have shown that normal young
adults can estimate their nocturnal sleep latencies with relative accuracy
(Baekeland & Hoy, 197 1; Frankel et al., 1976; Monroe, 1967).

SLEEP ONSET ESTIMATES IN INSOMNIACS

In our next studies, we sought to compare the abilities of control subjects and
sleep-onset insomniacs to estimate their sleep latencies (Moore, 198 1; Moore,
Bonnet, Warm, & Kramer, 1980). One problem of estimating sleep latency in
most studies is that estimates are generally collected in the morning about 8
hr after the actual event. Many things can happen during those 8 hr, including
multiple awakenings, movements, and various forms of disturbance. In addi-
tion, it can be very difficult for some people to report accurately an event that
happened that long ago. Therefore, the first study examined the abilities of an
Sleep Onset 151

age-matched ( 1 8-35 years) group of 12 normal and 12 insomniac subjects to

estimate their latency to sleep onset both at the normal early morning time
and when they were awakened at their first sleep spindle on counterbalanced
nights. All the subjects had previously had a screening night in the laboratory
to rule out other sleep disorders as a cause of the insomnia. Several other vari-
ables, such as oral temperature and anxiety level, that are potentially related
to sleep-related estimates, were measured or controlled. The subjects were
drug free for 2 weeks prior to the study and were allowed no naps, alcohol, or
caffeine during the study. Watches and clocks were not allowed in the sleeping
rooms. All estimates were made from the time when the technician told the
subject good night over the intercom system. This time was also marked by
the technician as the beginning of the sleep recording. Oral temperature was
measured immediately prior to the beginning of the recording each night and
immediately after each estimate of sleep latency. Anxiety levels were measured
each evening and each morning. For all sleep latency measures, the values to
be presented are geometric means (inverse of the mean of the log transform)
because sleep latency estimates are not normally distributed.
Interestingly, it was found that overall EEG-determined sleep latencies
did not differ significantly in the two groups of subjects (1 8 min vs. 26 min).
However, as expected, normal subjects' subjective estimates of their sleep la-
tency were significantly shorter than those of insomniac subjects (14 min vs.
38 rnin), F( 1,22) = 17.64, p < .O 1 . Of primary interest was the lack of a signifi-
cant interaction between time of estimate (sleep spindle and morning) and
experimental group; that is, the insomniacs overestimated their sleep onset
latency both in the morning and at the first sleep spindle at night. It was found,
however, that both groups estimated that their sleep onset latencies were
longer when they made the estimates in the morning. This probably resulted
because few of the subjects thought they had actually been asleep yet when
they were awakened for their nocturnal sleep spindle estimate. However, it is
also possible that there is a circadian rhythm to time perception in general and
that insomniacs differ from normal subjects in either the amplitude or phase
of that rhythm.
If the circadian rhythm explanation were valid, then one might expect
to find body temperature differences in the two groups. Significant differences
between insomniac and normal subjects in oral temperature were not found
at any test point. In fact, body temperature at lights out was 98.4"Fin both
groups and was actually 0.2"F lower in insomniacs at the first sleep spindle.
Two previous studies have found a significantly higher oral temperature in
insomniac than in normal subjects at sleep onset (Monroe, 1967) and two
studies have not (Johns, Masterton, & Bruce, 1971; Mendelson et al., 1984).
The oral temperature findings are relevant because Monroe and others have
hypothesized that insomniacs have increased physiological activation at sleep
onset and that it is the activation (most easily indexed by temperature) that is
related to the overestimation of sleep latency. Although the time perception
literature also suggests that there is a positive correlation between body tem-
152 MICHAEL H. BONNET

perature and magnitude of temporal estimate (Pfaff, 1968),those findings can-

not explain the data in the current study.
Significant differences in anxiety were found. A main effect was found
showing insomniacs to be more anxious than normal sleepers, F( 1,22) =
10.11, p < .O 1. Therefore, anxiety may have played a significant role in the
estimates of sleep onset, as suggested by other temporal perception studies
involving anxiety (Falk & Bindra, 1954; Langer, Wapner, & Werner, 1961).
A second experiment was performed to determine whether insomniacs
are simply poor estimators of time. If general anxiety level were the major
factor in determining overestimation, then one might suppose that insomni-
acs would overestimate all intervals of time. However, it is also possible that
insomniacs might simply become more anxious when they are left alone for
relatively long periods of time (ie., the 20 min before falling asleep) and might
therefore overestimate only longer intervals. Several investigators have shown
circadian rhythms in time perception (Aschoff, Geidke, Poppel, & Wever,
1972; Masao, 1978). Because the circadian curve of time perception may be
different for insomniac than normal sleepers and because these rhythm differ-
ences may have accounted for the results of the initial experiment, we found
it important to examine the time estimation of insomniacs at several different
times during the day. Alternatively, insomniacs may simply become more
anxious in bedroom settings and therefore may overestimate intervals only
when they occur in a bedroom setting. Finally, insomniacs may react only to
the sleep-specific demand situation and may therefore overestimate only when
they are under the pressure to attempt to fall asleep. These possibilities became
the experimental conditions for the second study (Moore, 1981; Moore, Bon-
net, &Warm, 1982).
The screening, age range, and subject requirements that were specified
in the previous experiment were carried over to the second experiment, and
12 new insomniac and 12 new normal sleepers were recruited. The subjects
participated for 1 day. During that day, subjects estimated short intervals (5-
35 s) and a long interval (19 min) five times (at 7:OO a.m., 11:OO a.m., 3:OO
p.m., 7:OO p.m., and 11:OO p.m.). The subjects also took a nap at noon and
were given 19 min to fall asleep. Half of the subjects did all of their time esti-
mating in a neutral laboratory setting, whereas the remaining subjects did all
of their time estimating in a bedroom setting. All intervals that were estimated
were unfilled (i.e., the beginning and end of each interval were marked by a
burst of white noise, but there were no intervening events). All estimates were
made individually on log sheets that were analogous to those used in the first
study. Estimating procedures were practiced by all subjects within 48 hr of
participation in the study. To replicate and extend other findings in the first
study, anxiety and oral temperature were measured at the beginning and end
of each testing session and nap across the day.
The results of this study with comparison data from the first experiment
for nocturnal sleep onset estimates are displayed in Figure 3. For all estimates,
the ratio of subjective estimate to objective interval length was calculated. The
Sleep Onset 153

W
I-
la60 1 a Insomniac
0N o r m a l
<
z
I-
u)
w
w
2
!-
V
W
7
m
?
W
2
I-
V
w
-2
m
3
u)

Sleep Onset Nap Long Intervals Short Intervals

CONDITION

Figure 3 Subjective to objective sleep onset latency ratios in groups of insomniac

and normal sleepers at sleep onset, at daytime nap onset, for long intervals during the
day, and for short intervals during the day. (Data from Moore, 1981; Moore, Bonnet,
& Warm, 1982; Moore, Bonnet, Warm, & Kramer, 1981.)

resulting values were used in an analysis of variance (ANOVA) with terms for
group (normal vs. insomniac), place of estimate (bedroom vs. nonbedroom),
and time of test. For the short intervals and long intervals, no significanteffects
were found for group, place, or time of day. For oral temperature, there was
no significant effect for group (F< 1.OO),but there was the expected effect for
time of day, F(4,80) = 15.18, p < .01. As in the first study, it was found that
the insomniacs scored significantly higher on the State Anxiety scale of the
Spielberger State-Trait Anxiety Inventory (Spielberger, Gorusch, and Lu-
shene, 1968). The significant effect was the main effect, F( 1,20) = 8.22, p <
.O1. The anxiety ANOVA effects for place of estimate, for time of day, and for
interactions were not significant. In the afternoon nap attempt, both insom-
niac and normal sleepers estimated relatively well. The insomniacs slightly
overestimated the interval, and the normal sleepers estimated the interval al-
most correctly. No significant difference was shown in the subjective to objec-
tive sleep latency ratio for normal and insomniac sleepers, t(22) = 0.72. As
expected from the other data, no napassociated temperature differenceswere
found, but anxiety again differentiatedthe groups, t(22) = 2.54,p < .01.
Sufficient observations of time estimates, oral temperature measures,
and anxiety scores were available so that Spearman rank-order correlations
could be calculated for each subject across the day. The correlations were then
154 MICHAEL H. BONNET

tested for direction (to determine if a positive or negative relation between

pairs of variables existed for either insomniac or normal sleepers). No indica-
tion of a significant association between subjective time estimates and either
state anxiety or body temperature were found in either group.
The two studies together suggest that insomniacs do overestimate their
sleep onset latency but that this overestimation is directly linked to events that
take place specifically during the first 20-30 min in bed at night as the insom-
niac attempts to fall asleep. No evidence was found to suggest that insomniacs,
as compared with normal sleepers, misperceive any time interval during the
day. This implies (a) that insomniacs do not simply misperceive time, (b) that
insomniacs do not misperceive long intervals of time, (c) that insomniacs do
not become more anxious or misperceive time in the presence of sleep-associ-
ated stimuli (i.e., in a bedroom), and (d) that insomniacs do not perceive a
nap sleep onset as similar to a nocturnal sleep onset even though the setting
contains both the bedroom stimulus and the demand to sleep. This also im-
plies that the nap is viewed as a unique sleep situation without the usual de-
mands (e.g., “If I do not fall asleep quickly, I will feel terrible in the morning”),
and this view is supported by the fact that the participating subjects reported
that they normally did not take naps.
The data also seem to imply that neither oral temperature nor anxiety is
directly related to the misperception of nocturnal sleep onset in insomni-
acs. Although the insomniacs were significantly more anxious than the normal
sleepers, they were more anxious than the normal sleepers at all measuring
points. This means that the insomniacs were more anxious than normal sleep-
ers at many points at which their perception of time was just as accurate as
that of the normal subjects. These data leave open the possibility of a complex
interaction in which insomniacs are stably more anxious than normal sleepers
but react differently only when the lights go out and they are faced with the
nocturnal sleep period. Conversely, insomniacs may simply ruminate more at
their nocturnal sleep onset in a way that is not directly measured by a state
anxiety scale. Finally, it is possible that insomniacs have a poor wake/sleep
perceptual interface and simply do not know when they actually enter the
sleep state. At the subjective level, at least, insomniacs may slide slowly into
sleep rather than fall asleep.

SLEEP ONSET INSOMNIA AS A PERCEPTUAL PROBLEM

Recently, we have begun to consider sleep onset problems in insomniacs with-
out other sleep disorders or significant psychopathology as more of a cogni-
tive-perceptual difficulty. Depending on the extent of this cognitive-percep-
tual problem, we may be able to teach these insomniacs to identify more
accurately the time of their own sleep onset and thereby to estimate more
accurately their sleep latency. Furthermore, if this perceptual ability can be
taught, it may by definition eliminate the insomnia complaint because the
long subjective sleep latencies will no longer exist. A current study (Downey
& Bonnet, 1989) is directed toward improving perception of the sleep/wake
interface.
Sleep Onset 155

As in the previous studies, participants were carefully screened to elimi-

nate other sleep disorders such as periodic leg movements or sleep apnea. Pa-
tients with significant psychopathology, as measured by the Minnesota Multi-
phasic Personality Inventory (MMPI) were also eliminated. In the current
study, insomniacs were between 18 and 40 years of age and had reported in-
somnia for more than a year. On a sleep laboratory screening night, subjects
were required to overestimate their objective sleep latency by at least 50% but
to have relatively normal sleep efficiency (>85%) once they had fallen asleep.
The final criterion was used in an attempt to define the insomnia clearly as
sleep onset rather than sleep maintenance.
After an adaptation night, subjects were randomly assigned to either a
control or a treatment condition. The next laboratory night was an undis-
turbed baseline night. On the following laboratory night, subjects in both con-
trol and treatment conditions were awakened 27 times during the night ac-
cording to criteria to be described. The control and treatment conditions
differed in two respects. First, in the treatment condition, subjects were shown
their sleep recording from the adaptation night, and various sleep events were
defined for them. At the same time, typical mental content at the various sleep
times was described. For example, at a representative period of alpha, the sub-
jects were told that alpha represented relaxed wakefulness and that when peo-
ple are "awakened" at that point, they usually report thinking about things
and being in general control of their thoughts. The initial sleep spindle was
differentiated as a state in which thoughts may be drifting, without direct con-
trol, or no longer completely linked together with earlier thoughts or concerns.
Thoughts 5 min after the initial sleep spindle were characterized as showing
more loss ofcontrol or contact. Subjectswere told that occasionally these latter
awakenings may not be accompanied by a clear thought or may seem cogni-
tively unusual. Second, subjects in the treatment condition were given appro-
priate feedback concerning their sleep/wake state from the experimenter after
each awakening, whereas subjects in the control condition were not.
The awakeningswere designed to occur at three times around sleep onset
to give the subjects three consistent cognitive surrounds as described pre-
viously and to help them clarify their own sleep onset process. The three awak-
ening possibilities were labeled A, B, and C. Point A was preceded by a minute
of clear alpha; Point B was the first well-defined sleep spindle; and Point C was
5 min after the first well-defined sleep spindle. Trials A, B, and C were block
randomized. Nine trials were given early in the night, nine trials were given in
the middle of the night, and the remaining nine trials were given early in the
morning.
After all laboratory nights, subjects completed standard morning ques-
tionnaires about their night of sleep and provided estimates of their sleep onset
latency. Following the treatment/control night, subjects were allowed a nor-
mal night at home to recover from any sleep disturbance effects from the
awakening procedure. After the home night, subjects returned to the sleep
laboratory for an undisturbed night of sleep followed by normal morning esti-
156 MICHAEL H. BONNET

mates. Sleep latency estimates from this last night were used to determine
changes in perceived sleep latency. Subjects who received the control condi-
tion first received the training condition at least 1 week later. Subjects who
received the treatment condition first could not be scheduled for a control
condition (because they had already been trained to identify the cognitive sur-
round of sleep onset).
To date, 4 subjects have completed both the control and the treatment
conditions, and 1 subject has completed the treatment condition only. The
data indicate that insomniacs can learn to differentiate the three sleep condi-
tions. Subjects improved their correct sleep state differentiation from 46% on
the control week to 93% on treatment weeks. The 93% correct on treatment
weeks implies that a very good discrimination can be made. Of course, the
more important issue is the subjective to objective sleep latency ratio. This
ratio was also decreased. The mean ratio was 4.0 after the final laboratory
night in the control condition and 1.8 after the final laboratory night in the
treatment condition, F( 1,4) = 13.26, p < .05. Although this seems to imply
that patients had greatly improved estimates, the subjective to objective ratio
on the baseline night directly preceding treatment was 2.1, and that was not
significantly different from the 1.8 value following treatment. The numbers
should stabilize with a greater number of subjects.
Of additional interest, subjects were asked to estimate the probability
with which they thought they would be able to fall asleep in less than 30 min
on a given night across the course of the study. Those estimates ranged from
18-28% prior to the treatment night. On the final laboratory night, the esti-
mate was 59%, F( 1,4) = 22.62, p < .01. This finding corresponds to reports
from subjects that they learned to feel more in control of their sleep behavior.
One subject had a long-term ( 5 month) follow-up. That subject reported that
sleep remained improved and that, despite some bad nights, nights were not
all bad as they previously had been. When the subject slept in the sleep labora-
tory once again, she could still identify the three conditions with 85% accu-
racy. In addition, at 5-month follow-up, that subject had a subjective to objec-
tive latency ratio of 1.5 compared with a ratio of 3.3 at the end of the
experiment proper.

SUMMARY
If you ask yourself how long it took to fall asleep last night, you will probably
be faced with a definition based primarily on absence (i.e., “I do not remember
much happening after I went to bed, so I probably fell asleep quickly”). It is
difficult to make accurate retrospective judgments of periods without clear
markers, and it is surprising how well most people can make such judgments.
Our studies have shown that when normal subjects are given clear markers
and immediate judgments, there is still a 4-8 min period around sleep onset
when cognition is changing and determination of sleep state is uncertain. In
most people, this error has little impact. In patients with insomnia, errors in
defining sleep state are much greater, and such errors in perception may form
Sleep Onset 157

the basis for the long-term use of hypnotics or other medication. Thus, the
problem of sleep onset in patients with insomnia is real.
The current work was based on a small and select group of insomniacs
who appeared to suffer primarily from a problem of misperception of sleep
onset. These patients almost certainly had a problem linked to events that
occurred during their sleep onset, and the present data suggest that those
events were probably cognitive. It is likely that a cognitively based treatment,
perhaps the one described here, can be of benefit. However, some cautions are
in order. There is no evidence that the results described herein will apply to
patients with major depression or with a long history of medication use. It is
also unlikely that a cognitive treatment program will be helpful for patients
with disorders that significantly fragment sleep, because each brief arousal
may bring a return of conscious events that may become mixed with the sleep
onset definition. Rather, it is hoped that individuals can begin to apply objec-
tive behavioral methods to diagnose and treat some forms of insomnia and
that these methods will serve as a starting point for the development of behav-
ioral treatments for more difficult insomnias.

References
Anch, A. M., McCoy, J. S., & Somerest,J. S. (1980). The relationship between EEG
alpha activity and perceived awakening. Sleep Research, 9, 22.
Aschoff, J., Geidke,H., Poppel, E., & Wever, R. ( I 972). The influence of sleep interrug
tion and sleep deprivation on circadian rhythms in human performance.In W. P.
Colquhoun (Ed.), Aspects of human efficiency (pp. 135- 150). London: English
University Press.
Baekland, F., & Hoy, P. (197 1). Reported vs. recorded sleep characteristics.Archives
of General Psychiatry, 24, 548-55 1.
Bonnet, M. H., & Moore, S. E. (1982). The threshold of sleep: Perception of sleep as a
functionof time asleep and auditory threshold. Sleep, 5,267-276.
Campbell, S. S., & Webb, W. B. (198 1). The perception of wakefulness within sleep.
Sleep, 4, 177- 183.
Carskadon, M. A., Dement, W. C., Mitler, M. M., Guilleminault,C., Zarcone, V. P.,
& Spiegel, R. (1 976). Self-reportsvs. sleep laboratory findings in 122 drug-free
subjects with complaints of chronic insomnia. American Journal of Psychiatry,
133, 1382-1388.
Downey, R. D., &Bonnet,M. H. (1989, June). Sleep-wakediscrimination in subjective
insomnia improves as a function of sleep onset feedback. Paper presented at the
meeting of the Association of Professional Sleep Societies, Washington, DC.
Falk, J. L., & Bindra, D. (1954). Judgments of time as a function of serial position and
stress. Journal of Experimental Psychology, 47,279-284.
Frankel, B., Coursey, R., Buchbinder, R., & Snyder, F. (1976). Recorded and reported
sleep in chronic primary insomnia. Archives of General Psychiatry, 33, 6 15-623.
Hauri, P., & Cohen, S. (1977). The treatment of insomnia with biofeedback: Final
report of Study 1 . Sleep Research, 6, 136.
Johns, M. W., Masterton, J. P., & Bruce, D. W. (1971). Relationship between sleep
habits, adrenocorticalactivityand personality.Psychosomatic Medicine, 33,499-
508.
158 MICHAEL H. BONNET

Langer, J., Wapner, S., & Werner, H. ( 1 96 1). The effect of danger upon the experience
of time. American Journal of Psychology, 74, 94-97.
Masao, K. (1978). Performance and time of day: Diurnal variations in physiological
and psychological tests. Journal of Child Development, 14, 16-24.
Mendelson, W. B., Garnett, D., Gillin, J. C., & Weingartner, H. (1984). The experience
of insomnia and daytime and nighttime functioning. Psychiatry Research, 12,
235-250.
Monroe, L. J. ( 1967). Psychological and physiological differences between good and
poor sleepers. Journal ofAbnorma1 Psychology, 72,255-264.
Moore, S. E. (198 1). Estimates of sleep latency and of other temporal intervals in insom-
niac and normal sleepers. Unpublished master's thesis, University of Cincinnati.
Moore, S. E., Bonnet, M. H., & Warm, J. S. (1982). Time estimation in normal and
insomniac subjects. Sleep Research, 'I 1, 16 1.
Moore, S. E., Bonnet, M. H., Warm, J. S., & Kramer, M. (1980). Estimates of sleep
latency in the morning and at sleep onset in insomniac and normal subjects. Sleep
Research, 10, 219.
Nease, N. F., Monroe, S. E., Bonnet, M. H., & Kramer, M. (198 1). Perception of sleep
latency across nights and conditions in insomniacs. Sleep Research, 10, 220.
Olmstead, E., Hauri, P., Percy, L., & Hellekson, C. (1 980). Subjective vs. objective
estimation of sleep onset in normal sleepers and in insomniacs. Sleep Research,
9,216.
Pfaff, D. ( 1968). Effects of temperature and time of day on time judgments. Journal of
Experimental Psychology, 76,4 19-422.
Rechtschaffen, A. (1968). Polygraphic aspects of insomnia. In H. Gastaut, E. Lugaresi,
G. Berti, G. Ceroni, & G. Coccagna (Eds.), The abnormalities of sleep in man.
Bologna, Italy: Gaggi.
Spielberger, C. D., Gorusch, R. L., & Lushene (1968). The state-trait anxiety inven-
tory. Preliminary test manual for form X . Tallahassee, FL:Florida State Univer-
sity.
Spielman, A. J., Tannenbaum, R., Adler, J., Saskin, P., Pollak, C. P., RofFwarg, H. P.,
& Weitzman, E. D. (1980). A comparison between objective and subjective sleep
latency in patients with chronic insomnia. Sleep Research, 9,225.
Tietz, E. I., Zorick, F. J., Roth, T., & Kaffeman, M. E. (1980). Objective and subjective
sleep parameters in documented and undocumented insomniacs. Sleep Research.
9, 222.
 CHAPTER 1 2

ARE YOU AWAKE? COGNITIVE

PERFORMANCE AND REVERIE
DURING THE HYPNOPOMPIC STATE
DAVID F. DINGES

Kleitman (1 963) observed that “immediately after getting up, irrespective of

the hour, one is not at one’s best” (p. 124). It is a paradoxical phenomenon-
being more impaired upon awakening from sleep than upon going to sleep-
that has been documented for a wide array of behavioral tasks. Typically, such
impairment is modest and short-lived due to gradual awakening or to a slow
transition out of the hypnopompic state. It can be dramatic, however, if the
arousal from sleep is abrupt, regardless of whether the sleep occurs at night or
during a daytime nap (Dinges, Orne, Evans, & Ome, 1981). Further, the inten-
sity of the hypnopompic state as evidenced by the severity of cognitive impair-
ment can be most profound if the awakening occurs during the first half of the
night or if the person has been awake for a protracted period of time and is
aroused after only an hour or two of recovery sleep. In such situations, the
sleepiness evident in the hypnopompic state results in vastly impaired perfor-
mance compared with that seen in a sleep-deprived subject (Dinges, Orne, &
Orne, 1985b).
The dramatic nature of hypnopompic disorientation, confusion, and
performance impairment has been described by a variety of names, including
“sleep drunkenness” (Broughton, 1968, 1973), “postdormital sleepiness” (As-
sociation of Sleep Disorders Centers, 1979), and “sleep inertia” (Lubin, Hord,
Tracy, & Johnson, 1976). The former terms are now used primarily for sleep
pathology, whereas the latter is widely applied to describe transient awakening
(i.e., hypnopompic) impairment in healthy persons. This chapter is concerned
with hypnopompic cognition during intense sleep inertia and the processes
that might account for it.
159
160 DAVID F. DINGES

SLEEP INERTIA AND PERFORMANCE DURING

THE HYPNOPOMPIC PERIOD
The studies cited in Table 1 represent much of what is known about hypno-
pompic cognition and the awakening process as it pertains to performance
capability. Table 1 does not include a separate body of literature on dream
reports at awakening. Virtually all the work conducted on performance during
the hypnopompic state has derived from two theoretical perspectives, with
slightly different emphases. The first reflects an interest in the functional
differences between REM and non-REM (NREM) sleep stages in information
processing potential; this is the performance analogue of the studies of sleep
stage-dependent dream mentation. The approach typically taken is to awaken
subjects from different stages of sleep and evaluate performance on a parame-
ter of theoretical import, such as perception (e.g., Lavie & Sutter, 1975) or
memory (e.g., Bonnet, 1983;Stones, 1977).The second approach derives from
studies of human performance during periods of prolonged quasi-continuous
wakefulness, when the adverse effects of sleep loss must be weighed against
the adverse effects of sleep inertia upon abrupt arousal from sleep due to an
emergency (e.g., Hartman & Langdon, 1965; Haslam, 1982).In these studies,
emphasis is placed not on the preawakening stage of sleep b i t rather on the
magnitude of sleep inertia, its duration, and the range of performances affected
by it.
At the heart of sleep inertia is the nature of hypnopompic cognition and
biobehavioral functioning on arousal from sleep. Although much more has
been written about the hypnagogic state (Mavromatis, 1987; Schacter, 1976)
than the hypnopompic state, the phenomenon of sleep inertia and its accom-
panying cognitive processes are ubiquitous. Because of this ubiquity and be-
cause it is typically a transient phenomenon (lasting between 1-20 min) dur-
ing which cognition and performance can be grossly altered relative to other
times, the hypnopompic period is often ignored in many studies on the effects
of sleep on human functioning.
It is, however, precisely because of the dramatically altered cognitive per-
formance, reverie, and subsequent amnesia of the hypnopompic period that
sleep inertia is worthy of increased attention. At the very least, the phenome-
non offers a window to the changes in cognition from sleep to waking. In its
extreme form in an otherwise healthy individual, sleep inertia affords a model
for cognitive impairment of a kind rarely seen in either experimental (e.g.,
sleep deprivation) or clinical (e.g., insomnia) studies. What follows is an espe-
cially dramatic example of sleep inertia that illustrates how profound the phe-
nomenon can become if the appropriate paradigm is used. Following the ex-

The research and substantive evaluation upon which this manuscript is based were s u p
ported in part by Office of Naval Research Contract N00014-80-C-0380, in part by National
Institute of Mental Health Grants MH-19156 and MH-44193, and in part by a grant from the
Institute for Experimental Psychiatry Research Foundation. I am grateful to Kelly A. Gillen for
help in preparing the manuscript.
Hypnopompic Cognitive Performance 161

ample is a discussion of factors that contribute to it, including a review of

evidence that sleep depth more than the preawakening stage of sleep is the
crucial variable that influences hypnopompic cognition. We conclude with a
theoretical framework that posits a common process (sleep pressure) underly-
ing hypnopompic, hypnagogic, and sleeprelated waking reverie.

A PARADIGM TO STUDY SLEEP INERTIA

To study the role of sleep pressure and various aspects of sleep infrastructure
on hypnopompic performance, we conducted two studies of human cognitive
functioning upon abrupt awakening from naps (Dinges et al., 1981; Dinges,
Ome, & Ome, 1985a, 1985b). Like the studies cited in Table 1, the goal was
to use a cognitive task as a probe to investigate the processes underlying h y p
nopompic cognition. Relatively short periods of sleep ( 1- 120 min) provide
a theoretically important way of studying cognitive performance during the
hypnopompic state without confounding circadian variation and sleep infra-
structure changes, which result when awakening occurs from longer periods
of sleep (4-8 hr). Thus, studies of naps under 2 hr, at different phases of the
circadian cycle and following varying amounts of prior sleep loss, permit as-
sessment of specific aspects of sleep infrastructure and depth in relation to
performance upon awakening.
In both studies, performance on a 3-min descending subtraction task
(DST)was used as a cognitive probe on abrupt awakening. Details of the task
have been published elsewhere (Dingeset al., 1985.a). Briefly, it was developed
to permit the subject to perform it while lying on a bed in the dark, thereby
allowing assessment within a few seconds of sleep offset. The task requires an
oral rather than a nonverbal response from the subject, which permits reverie
intrusionsto be observed. The subtractionsare done silently, and the answers
are said aloud. Both speed and accuracy are emphasized,and because the sub-
trahend and minuend change after each response, a considerableload is placed
on memory. If a subject does not give a response on the DST for 20 s, the
experimenter then says, “Please continue, guess if you have to,” and thereafter
pushes the subject every few seconds to respond. (This was necessary often
during the hypnopompic period but was necessary only infrequently during
baseline, presleep, or sleep deprivation phases of the studies.) Thus, no subject
could score poorly on the task merely by not responding for the bulk of the
allotted 3 min.
In our studies, DST performance was assessed repeatedly before each
nap (to ensure that performance was asymptotic) and immediately following
a motor response (answering a telephone) and an affirmative oral response (to
the question “Are you awake?’), as well as repeatedly thereafter. In all 198
awakening nap protocols we have run to date, sleep was polysomnographically
recorded, and the EEG was monitored during hypnopompic performance.
Naps were either 1 hr (Dinges et al., 1981) or 2 hr (Dinges et al., 1985a) in
duration, although subjects were not told how long they would be permitted
to nap. They were aware, however, that they would be expected to perform
162 DAVID F. DINGES

Table 1
STUDIES OF PERFORMANCE UPON AWAKENING
Study Task

Dinges, Orne, Evans, and Orne (1981) Simple reactiontime

Dinges. Ome, and Orne (1985a)
Okuma, Nakamura, Hayashi, and Fujimori (1966)
Rosa, Bonnet, and Warm (1983)
Webb and Agnew (1964)
Wilkinson and Stretton (1971)
Williams, Morlock, and Morlock (1966)

Feltin and Broughton(1968) Complex reactiontime

Goodenough, Lewis, Shapiro. Jaret, and Sleser (1965)
Scott (1969)
Seminara and Shavelson (1969)

Jeanneret and Webb (1963) Grip strength

Tebbs and Foulkes (1966)

Wilkinson and Stretton (1971) Steadiness/coordination

Hartman and Langdon (1965) Complex simulation

Hartman, Langdon, and McKenzie (1965)
Langdon and Hartman (1961)
Seminara and Shavelson (1969)

Fort and Mills (1972) Letter cancellation

Haslam (1982) Logical reasoning

Akerstedt and Gillberg (1979) Memory tasks
Bonnet (1983)
Gastaut and Broughton(1965)
Grosvenor and Lack (1984)
Stones (1977)

Dinges et al. (1981) Mental arithmetic

Dinges et al. (1985a)
Pritchett (1969)
Wilkinson and Stretton (1971)
Scott (1969) Clock reversal
Hypnopompic Cognitive Performance 163

Table 1 (continued)
Study Task

Carlson, Feinberg, and Goodenough (1978) Time estimates

Koulack and Schultz (1974) Vigilance, trailmaking

Lavie and Giora (1973) Visual-perceptual

Lavie and Sutter (1975)
Scott (1969)
Scott and Snyder (1968)

the DST immediately upon awakening from sleep. Awakening was done audi-
torily by a telephone, which rang continuously for 1 min and, if not answered,
rang on and off every 2 s for another minute. If the subject still did not respond,
his or her name was spoken over the intercom until a verbal response was
elicited. After answering the phone, the subject was asked how much time had
elapsed since he or she had last spoken with the experimenter (immediately
prenap) and was asked to estimate his or her sleepiness on a 10-point analogue
scale. Following this, the subject was instructed to hang up the phone, lie back
down (in the dark), and perform the DST. The polygraph was kept running
throughout awakening performance, which was audio tape-recorded. After
electrode removal, DST performance was again assessed repeatedly.
In one major study (Dinges, 1986; Dinges, Ome, Whitehouse, & Ome,
1987; Dinges, Whitehouse, Ome, & Ome, 1988), the amount of continuous
wakefulness (sleep loss) prior to a 2-hr nap opportunity was varied from 6 hr
to 52 hr. This had the effect of producing marked differencesin nap sleep-stage
infrastructure and amount, which permitted an assessment of hypnopompic
cognition as a function of varying sleep depths or intensities. At the most ex-
treme intensity, sleep inertia was profound, and intrusions of hypnopompic
reverie occurred during the DST at awakening. The following example illus-
trates the power of our experimental protocol to produce a dramatic hypno-
pompic condition that is characterized by social interaction with simultaneous
performance impairment, hypnopompic reverie, misjudgment of sleepiness,
and a dissociation between the electroencephalogram (EEG) and behavior.

SOCIALLY AWAKE YET FUNCTIONALLY ASLEEP

The nature of the cognitive impairment in DST performance that was evident
at awakening, especially after subjects had been sleep-deprived and therefore
had slept very deeply, was such that it was often accompanied by dramatic
intrusions of reverie as subjects attempted to say the answers aloud. Table 2
provides a transcript of the interaction between one of our sleep-deprived sub-
jects (1 8-year old healthy man) and the experimenter after the subject had
been awake for 52 consecutive hours in the sleep-deprivation protocol, during
164 DAVID F. DINGES

Table 2
DIALOGUE BETWEEN SUBJECT AND EXPERIMENTER DURING
PRESLEEP PERFORMANCE OF DESCENDING SUBTRACTION TASK
AND UPON AWAKENING FROM SLEEP
~~ ~

Minute Speaker Content

Presleep Performance

0 Experimenter: Iwould like you to do the subtractiontask again. Remember to

work as fast and accurately as you can. Your starting number
is 931.
1-3 Subject: 931 9P2 914 907 901 895 891 888 886
869 862 . . . . . . . . . 509 503.
[75correct, 2 errors]
4-5 Experimenter: That’s good. On a scale from 1 to 10,where 1 is very wide
awake and 10 is very sleepy, how do you feel now?
Subject: 8
Experimenter: Please lie quietly with your eyes closed, but stay awake, until I
tell you that you can go to sleep.
Subject: Okay.
Experimenter: Remember that the end of the nap will be signalled by the
telephone ringing, which you should answer as quickly as
possible. You will then be asked to do the subtraction task.
Okay, you can go to sleep now.

Postsleep Performance

128-129 Experimenter: Are you awake?

Subject: Yes!
Experimenter: Can you hear me okay?
Subject: Yeah.
Experimenter: How long since I spoke to you last?
Subject: Um . . . um . . .90minutes.
Experimenter: On a scale from 1 to 10,where 1 is very wide awake and 10 is
very sleepy, how do you feel now?
Subject: Um . . . about um . . .6.
Experimenter: Hang the phone up and lie back. Iwould like you to do the
subtractiontask.
[S having difficulty hanging up telephone.]
Experimenter: Can you see to hang it up?
Subject: There, got it.
Experimenter: Remember to work as fast and accurately as you can. Your
starting number is 648.
Hypnopompic Cognitive Performance 165

Table 2 (contlnued)
Minute Speaker Content
130 Subject: 648 64. . ah 63. . ummm 500 and ah. . (sigh)
Experimenter: Continue!
Subject: 500 and um . . let’s see. . um . .696 685 640, ah,
um. . . 632 637 637
Experimenter: Continue!
132 Subject: What if people ran faster than normal people run h o m e t h a n
the normal person runs faster than the square root of two
times. . . (mumbles incoherently). . . and normal quote-
unquote people take up two derivatives of normal people in
skin brackets. . .
[l correct, 4 errors]
133 Experimenter: That’s good.

Note: After the presleep test, the subject, who had been awake for 52 hr, took a 2-hr nap. AAer
the nap, the subject did not respond to the telephone bell but picked up the phone when his name
was called.

which the DST was repeatedly performed. The script in Table 2 begins as the
subject is lying on a bed, in a dark quiet room, just before being allowed to
sleep. His presleep (postsleep-loss)DST performance for 3 min of 75 correct
answers and 2 errors was below his presleep-loss levels of 82-99 correct an-
swers and 0-2 errors; this net loss in number completed is characteristicof the
cognitive slowing on subject-paced tasks that typically occurs with sleep loss
(Dinges, 1989b).
Followingthis performance, as is typical for intensely sleepy persons, the
subject fell asleep (Stage I ) within 30 s of being permitted to do so and had his
first epoch of slow wave sleep (SWS)4 min later (Dinges, 1986).He remained
asleep for the entire 2-hr period. He had no REM sleep but accumulated 82.5
min (69%of total sleep time) of SWS (70 min of Stage 4 sleep),which is nearly
as much SWS as healthy young adults of his age acquire in an average 8-hr
nocturnal sleep (cf. R. L. Williams, Karacan, & Hursch, 1974)! He was in
Stage 3 sleep at the time of the awakening bell. Although he did not answer
the awakening telephone call until his name was spoken, which indicates an
intense sleep depth, he interacted immediately thereafter, affirming that he
was awake, estimating the time since he had last spoken to the experimenter,
and providing a rating of his sleepiness.
Curiously, he estimated his sleepiness at a score of 6 at awakening, indi-
cating that he was less sleepy than he had been before the nap (rating of 8).
This was a clear dissociationbetween his self-report and his DST performance,
which was far worse at awakening than it was after 52 hr of wakefulness, just
prior to the nap. We have analyzed enough data to show that this hypnopom-
pic misjudgment of sleepinessis common in subjectswhose sleep has followed
166 DAVID F. DINGES

a sustained period of wakefulness beyond 18 hr, but that it does not occur if
the nap is taken before sleep deprivation (Dinges, 1988). The intensity of the
sleep inertia seems to make it difficult to estimate how sleepy one feels. This
finding is consistent with Sewitch’s (1984) report that the length of continuous
NREM sleep affects the normal sleeper’s perception of having been awake or
asleep, such that awakenings from SWS or after prolonged periods of NREM
sleep most often result in subjects being least able to identify accurately
whether they had been awake or asleep.
Following his time and sleepiness estimates and his interaction with the
experimenter over the phone, the subject was asked to perform the DST. He
was totally unable to execute subtractions correctly (only his repeat of the
starting number was correct; his four subtraction errors are in italics in Table
2), and he could not remember where he was in the sequence, despite two
prods to continue from the experimenter.
He also had considerable difficulty preventing hypnopompic reverie
from intruding into his oral output. The spoken reverie that occurred 2 min
into the awakening DST performance (1 32 min in Table 2) began after the
experimenter prompted him a second time to continue. It is noteworthy that
the apparent evocation of this spoken reverie by the experimenter is consistent
with the observations of Broughton (1 968, 1982) that parasomnic episodes
such as sleep walking and enuresis can be triggered by external stimuli and,
therefore, can be considered disorders of arousal or of partial arousal. In fact,
the awakening performance behavior of this subject is reminiscent of the con-
fusional arousals from SWS described by Broughton (1968), which he sug-
gested were due to “impaired cerebral responsiveness or of functional deaffer-
entation” (p. 1074). The actual reverie of our subject at 2-min postawakening
is illustrative of the semicoherent material we have observed in the awakening
reverie of other subjects. Neologistic phrases such as “normal people in skin
brackets” are referred to, along with phenomenological material appropriate
to the individual such as “the square root of two times” (this subject was a
math major and, indeed, the reverie has the quality of a statement of a mathe-
matical problem).
Interestingly, the subject indicated that he had dreamed during the nap;
this response was made in a questionnaire booklet he completed while elec-
trodes were being removed, 15 rnin after awakening. No dream content was
asked for and, consequently, it is unknown whether the dream related to his
hypnopompic reverie. This was not the only reverie he experienced, however.
Despite coherent social interaction and the completion of performance tasks,
he experienced other reverie intrusions into his cognitive performance oral
output at 6 min and 35 min postawakening. Although reverie was less dra-
matic in many other subjects to the extent that it lasted only a few seconds
rather than a full minute, it was apparent in most who had undergone inten-
sive sleep loss in the form of one- to five-word intrusions into oral performance
output. Such intrusions were most common in the hypnopompic period.
Hypnopompic Cognitive Performance 167

Figure 7 Vertex (Cz-AIA2) EEG recordings from a healthy young adult with eyes
closed lying on a bed in a dark quiet bedroom, taken at eight different times. Record
(A) after 6 hr ofwakefulness-instructed to lie awake quietly. (B) after 52 hr ofwakeful-
ness-minute 1 after being instructed to lie awake quietly. (C) 2 rnin after B-still
attempting to-stay awake. (D) 1 rnin after C and after being instructed to go to sleep.
(E) 2 min after D (asleep by polysomnographic criteria). (F) after 2-hr nap-minute 1
of DST performance at awakening (see text). (G) 2 rnin after F-during reverie of final
minute on DST (see Table 2). (H) 2 rnin after completion of DST at awakening-
instructed to remain awake, but S fell asleep.

FUNCTIONALLY ASLEEP YET

ELECTROENCEPHALOGRAPHICALLY AWAKE
The EEG recording obtained during the awakening reverie of our subject was
compared with recordings made under comparable conditions (e.g., eyes
closed, prone in bed) prior to and following sleep deprivation, during wakeful-
ness and sleep onset, in an effort to determine whether the reverie was accom-
panied by a clear change in EEG. Increased theta and delta activity in particu-
lar is characteristic of a period of reverie, and one would expect the EEG to
be visibly different during reverie than during the prenap DST performance,
when the subject was intensely sleepy but able to perform 75 correct subtrac-
tions in 3 min.
Figure 1 displays vertex EEG recordings made at eight times for this
subject. The presleep-deprivation record (Record A) taken after only 6 hr
awake is notably different from all others because it contains less slow wave
activity (delta and theta) and more beta activity. All records (Records B-H)
168 DAVID F. DINGES

from the sleepdeprivation period show increased slow wave activity. Of par-
ticular interest is the comparison between Records C and G. Although no
EEG recording was made during the presleep DST performance, Record C (at
52 hr and 2 rnin of sleep loss) was obtained 2 rnin into the 3-min presleep
wake, eyes-closed baseline (i.e., 1 rnin before the subject was allowed to sleep).
Record G was obtained 2 rnin into the postsleep awakening DST perfor-
mance, when the reverie at the bottom of Table 2 was elicited. There is no
obvious difference between these records, and they are clearly different from
the high-amplitude, slow-frequencywaveforms apparent in the sleep records
at 1 rnin and 2 rnin of nap onset (Records D-E) and when the subject fell
asleep after completing the awakening DST (Record H). Despite the lack of
EEG differences between Records C and G, differences at these times in DST
performance and cognitive coherency are profound (Table 2).

SLEEP DEPTH AND SLEEP INERTIA

The depth of sleep achieved by our subject during the 2-hr (recovery) nap
following over 2 days without sleep was intense, as it has been in every subject
we have examined who was sleep deprived. Recovery sleep following sleep loss
has long been known to involve a greater depth of sleep. Consistent with other
studies, the greater depth of sleep was evidenced by a decreased latency to
SWS (Dinges, 1986),an exceptionally high amount of SWS, especially during
the first NREM cycle of sleep (Borbely, Baumann, Brandeis, Strauch, & Leh-
mann, 1981; Dinges, 1986; Feinberg, Floyd, & March, 1987; Hume & Mills,
1977; Webb & Agnew, 1967, 1971), decreased body movement during sleep
(Naitoh, Muzet, Johnson, & Moses, 1973),a failure to respond to the awaken-
ing bell (Rechtschaffen, Hauri, & Zeitlin, 1966; Rosa & Bonnet, 1985; Wil-
liams, Hammack, Daly, Dement, & Lubin, 1964), and a rapid return to sleep
following awakening (Bonnet, 1978).All of these parameters provide an index
of sleep depth. Ironically, sleep depth often has not been considered in studies
of performance upon awakening.
There has been considerable interest in determining which aspects of
sleep account for sleep inertia or the performance impairment evident in the
hypnopompic period. The typical experimental paradigm used to study it has
consisted of abrupt, forced awakening from REM or NREM sleep stages.
Thus, when performance during the hypnopompic period has been investi-
gated (Table I), the paradigm for studies of sleep-stage-relatedmentation has
been most commonly used. In these studies, it is often assumed that the sever-
ity of performance sleep inertia is directly related to the preawakening stage of
sleep, such that awakenings from SWS yield the most dramatic hypnopompic
phenomena, whereas those from REM sleep yield more wakelike perfor-
mances. Although the results of most studies that have investigated arousal
from different sleep stages support this view, the bias of evaluating only the
preawakening stage of sleep is so pervasive that other aspects of the sleep infra-
structure are rarely examined to determine whether they are more consistently
associated with the magnitude and nature of hypnopompic phenomena.
HypnopompicCognitive Performance 169

There is reason to hypothesize, however, that the intensity of hypno-

pompic reverie and the extent to which cognitive performance is impaired
during awakeningare most accurately characterized as being a function of the
depth of sleep, of which the preawakening stage of sleep is but one facet
(Dinges et al., 1985a, 1985b).This is evidenced in three ways: (a) awakenings
exclusively from Stage 2 NREM sleep yield performance decrements that vary
as a function of time of night (circadian phase) (Rosa, Bonnet, & Warm,
1983); (b) awakenings from recovery sleep following prolonged wakefulness
yield decrements greater than those found for awakeningsfrom the same stage
of sleep prior to deprivation (Akerstedt & Gillberg, 1979; Fort & Mills, 1972;
Rosa et al., 1983); and (c) abrupt awakenings from naps yield increasingly
more severe cognitive performance decrements as the amount of NREM
(Stages 2 + 3 + 4) sleep increases (Dinges et al., 1981) and as the amount of
wakefulnessprior to sleep increases, regardless of the stage of sleep from which
subjects are awakened (Dinges et al., 1985a).
How can we account for the fact that most of the studies in Table 1 have
observed greater hypnopompic impairment on awakeningsfrom SWS relative
to Stage 2 or REM sleep? This may have resulted because SWS is consistently
associated with greater sleep depth, especiallyin the first NREM cycle (Dinges
1986; Feinberg et al., 1987).Generally, the preawakening depth of sleep may
be more important than the depth of sleep at other times in the sleep period
(Bonnet, 1983), making the preawakening stage of sleep a salient variable for
hypnopompic cognition. On the other hand, at least one study has reported
that the behavioral performance variables at repeated awakenings from recov-
ery sleep following (40 hr and 64 hr of) sleep loss are generally more sensitive
than sleep stages to different amounts of prior wakefulness (Rosa & Bon-
net, 1985).
In fact, it is not clear that SWS is essential for severe hypnopompic per-
formance impairment to occur in response to sleep depth. Bonnet (1985) ob-
served that hypnopompic disorientation could be profound in subjects who
were permitted to sleep but who were denied most of their SWS and REM
sleep during repeated nights of experimentally induced sleep disruption every
minute. In his study, auditory arousal thresholds increased dramatically by
the second night of sleep disruption, indicating increased sleep depth, while
subjectsbegan to become confused on awakening:
They often could not give ratings. One subject later recounted that at
awakening she could hear the technician talking but his words did not
seem to make sense.Other subjects could not perform simple tasks such
as being able to respond with “a” when prompted for the letter that
precedes “b”.One explanation for this behavior is sleep drunkenness
(i.e., arousal from very deep sleep resulting in confusion). (p. 18)

SLEEP PRESSURE AND THERMOREGULATION

It is my contention that sleep pressure, or the probability of transition from
wakefulnessto sleep, (Pr[W --+ S]) underlies sleep depth, SWS, and the magni-
170 DAVID F. DINGES

tude of sleep inertia effects during the hypnopompic period. The simplest way
to manipulate sleep pressure is through prior wakefulness. But which physio-
logical process underlies sleep pressure and correlates with manifestations of
sleep depth, especially pressure for SWS? Although biochemical changes must
underlie any such process, and candidates have been proposed (e.g., see Chap-
ter 2 by Hobson in this volume), there is reason to suggest that body core
temperature and, presumably, brain metabolic activity may be an essential
link in the hypnopompic process. Body core temperature shows a circadian
rhythm, but sleep (especially the first SWS period of the night) also has the
(evoked) effect of thermal down-regulation (Gillberg & Akerstedt, 1982).
There is evidence that this effect may be enhanced whenever the pressure for
sleep is increased. Aschoff, Giedke, Poppel, and Wever ( 1972) observed that
“after two days without sleep, one night of uninterrupted sleep results in an
exaggerated drop in rectal temperature” (p. 144). A significant covariation
between oral temperature and performance upon awakening has been re-
ported (Rosa & Bonnet, 1985).But the relation between SWS and temperature
may go both ways (Sewitch, 1987). Experimentally induced lowering of rectal
temperature at sleep onset has been associated with an increase in Stage 4 sleep
and a lengthening of the first NREM/REM cycle (Sewitch, Kittrell, Kupfer, &
Reynolds, 1986). Passive afternoon body heating has been observed to result
in an enhanced drop in rectal temperature during the first 2-4 hr of subse-
quent nocturnal sleep (Horne & Staff, 1983).
One possible mechanism underlying the increasingly impaired perfor-
mance at awakening of subjects who have experienced intense pressure for
sleep is a decline in cerebral metabolism resulting from thermal down-regula-
tion exacerbated by sleep pressure. The purpose of such down-regulation is
unclear, although speculation is easy (e.g., the increased need for protein syn-
thesis, which is favored during sleep, especially during periods of lowest basal
metabolic level). Whatever its purpose, such basal drops may have a longer
time course than the EEG manifestation of specificsleep stages and may make
coherent cortical function impossible until metabolic or specific biochemical
activity has been increased through the passage of time (e.g., circadian varia-
tion), a change in sleep stage (e.g., accumulation of REM sleep), or increased
physical activity at awakening (e.g., getting out of bed)-all of which are inter-
correlated such that REM sleep is more likely to occur as body core tempera-
ture is rising, and awakenings are more likely to occur from REM sleep
(Dinges, 1989a). Unfortunately, despite many studies of hypnopompic perfor-
mance, none have assessed body core temperature as a covariant of sleep iner-
tia beyond looking at circadian variability.
The hypothesized widespread metabolic decline in cortical activity that
would covary with pressure (and therefore with depth) of sleep would probably
make it exceedingly difficult for a person to perform well if aroused abruptly
from sleep when basal metabolic levels are low. Virtually every type of perfor-
mance, especially all cognitive performances involving memory and atten-
tion, would be adversely affected. There is no reason to believe, however, that
Hypnopompic Cognitive Performance 171

the subject could not interact socially at some simple level such as name ac-
knowledgment; indeed, this is what appeared to be the case with our subject.
On the other hand, when performance is demanded, reverie can intrude, espe-
cially in response to a stimulus (e.g., the experimenter says “continue”). The
nature of such semicoherent reverie may be determined by the extent to which
the “functional deafferentation” suggested by Broughton ( 1968) has taken
place. The reverie would, therefore, be the result of insufficient neural metabo-
lism for coherent mentation. The inhibition needed for directed cognition
would be lacking.

THEORETICAL MODEL OF REVERIE

BASED ON SLEEPINESS
The concept of pressure for sleep, which is expressed as the probability of
transition from wakefulness to sleep, (Pr[W + S]), is in effect a definition of
sleep tendency or of sleepiness. To the extent that sleep-relatedreverie occurs
without exception during sleep stages, especially during Stage 1 sleep, then a
sleepiness model can account for it. That is, as the probability of a transition
to sleep increases, the probability of reverie will increase: Increased pressure
for a thermoregulatory down-regulation may underlie this phenomenon.
Whatever its physiologic basis, it can account for the “semi-dreaming’’ (Kleit-
man, 1963, p. 22 1) of sleep-deprived subjects, which was observed in the earli-
est human sleep loss study (Patrick & Gilbert, 1896)as well as for both hypna-
gogic and hypnopompic reverie.
Figure 2 displays the manner in which sleep pressure manifests itself in
behavior and physiology, depending on the context. If the context is one in
which the subject is attempting to remain awake, as in sleep-deprivation stud-
ies, then increased microsleeps (Stage 1 intrusions), cognitive slowing, re-
sponse blocks or lapses, and response habituation will be increased. Semi-
dreaming or dreaming while awake during tasks that require oral output will
be more likely. If the subject is attempting to go to sleep, then increased sleep
pressure will be evident in decreased sleep latencies (e.g., as seen on the multi-
ple sleep latency test) and hypnagogic reverie will be more likely to occur
sooner. If the context is one of attempting to awaken from sleep, then in-
creased sleep pressure will result in increased sleep depth, and the probability
of hypnopompic reverie will be increased. There is, therefore, no reason to
posit that the reverie resulting from these three instances is qualitatively
different. The various effects of sleep pressure in Figure 2 are intercorrelated.
The word used to describe the resulting reverie is more aptly thought of as a
descriptor of the context in which each occurs, not of any unique characteristic
of the mentation.
Differences can occur, of course, in the severity of reverie intrusions (i.e.,
pressure for sleep). Hypnopompic mentation may be the most dramatic form
of reverie because it involves transition from sleep to wake (sleep inertia),
whereas waking and hypnagogic mentation involve transition from wake to
sleep (“wake inertia”). In all three cases, the process underlying the emanation
172 DAVID F. DINGES

PRESSURE FOR SLEEP

P r (W - S)
J
1microsleeps + latency to stage 1 4 amount of sws
4 slowing & lapsing + latency to sleep onset 4 duration of sleep
4 habituation 1 latency to SWS 4 awakening threshold
(Performance) (MSLT & variants) (Sleep depth)

REVERIE

“WAKE HYPNAGOGIC H YPNOPOMPIC

(Dreaming Awake)

Figure 2 Theoretical model of how sleep pressure (i.e., probability of a transition

from wake to sleep) can manifest in sleeprelated behavior and physiology, depending
on the context, and result in reverie. To the left is the sleepdeprivation context, in
which the subject is attempting to remain awake but microsleep intrusions result in
“semi-dreaming” reverie. Center is the sleep onset context, in which the subject is
attempting to go to sleep and intense sleepiness leads to rapid sleep onset and hypnogo-
gic reverie. To the right is the awakening context, in which the subject is attempting to
transition out of sleep but sleep inertia resulting from increased sleep depth leads to
hypnopompic reverie. In all three cases, the resulting reverie is the product of increased
physiological sleepiness.

of reverie and the nature of cognition in general may well be the same. What
happens to this process as REM sleep (and circadian time) accumulate re-
mains to be determined.

References
Akerstedt, T., & Gillberg, M. (1 979). Effects of sleep deprivation on memory and sleep
latencies in connection with repeated awakenings from sleep. Psychophysiology,
16,49-52.
Aschoff, J., Giedke, H., Poppel, E., & Wever, R. (1972).The influence of sleep intermp
tion and sleep deprivation on circadian rhythms in human performance. In W.P.
Colquhoun (Ed.), Aspects of human efficiency (pp. 135- 150). London: English
Universities Press.
Hypnopompic Cognitive Performance 173

Association of Sleep Disorders Centers. (1979). Diagnostic classification of sleep and

arousal disorders, first edition, prepared by the Sleep Disorders Classification
Committee, H. P. Rofiarg, Chairman. Sleep, 2, 1-137.
Bonnet, M. H. (1978). The reliability of depth of sleep and the effects of flurazepam,
pentobarbital and caffeine on depth of sleep. Dissertation Abstracts International,
38, 5632.
Bonnet, M. H. (1983). Memory for events occurring during arousal from sleep. Psycho-
physiology, 20, 8 1-87.
Bonnet, M. H. (1985). Effect of sleep disruption on sleep, performance, and mood.
Sleep, 8, 11-19.
Borbely, A. A., Baumann, F., Brandeis, D., Strauch, I., & Lehmann, D. (198 1). Sleep
deprivation: Effect on sleep stages and EEG power density in man. Electroenceph-
alography and Clinical Neurophysiology, 51,483-493.
Broughton, R. J. (1968). Sleep disorders: Disorders of arousal? Science, 159, 1070-
1078.
Broughton, R. J. ( 1973). Confusional sleep disorders: Interrelationship with memory
consolidation and retrieval in sleep. In T. J. Boag & D. Campbell (Eds.), A triune
concept of the brain and behavior (pp. 1 15- 127). Toronto: University of Toronto
Press.
Broughton, R. J. (1982). Human consciousnessand sleep/wake rhythms: A review and
some neuropsychologicalconsiderations. Journal of Clinical Neuropsychology, 4,
193-2 18.
Carlson, V. R., Feinberg, I., & Goodenough, D. R.(1978). Perception of the duration
of sleep intervals as a function of EEG sleep stage. Physiological Psychology, 6,
497-500.
Dinges, D. F. ( 1986). Differential effects of prior wakefulness and circadian phase on
nap sleep. Electroencephalographyand Clinical Neurophysiology, 64,224-227.
Dinges, D. F. (1988). When we can and cannot judge our sleepiness upon awakening.
Sleep Research, 17, 83.
Dinges, D. F. ( 1989a). The influence of the human circadian timekeeping system on
sleep. In M. Kryger, W. Dement, & T. Roth (Eds.), Principles andpractice ofsleep
medicine (pp. 153-162). Philadelphia, PA: Saunders.
Dinges, D. F. (1989b). The nature of sleepiness: Causes, contexts and consequences.
In A. Stunkard & A. Baum (Eds.), Perspectives in behavioral medicine: Eating,
sleeping, and sex (pp. 147- 179). Hillsdale, NJ: Erlbaum.
Dinges, D. F., Ome, E. C., Evans, F. J., & Ome, M. T. (198 1). Performance after naps
in sleep-conducive and alerting environments. In L. C. Johnson, D. I. Tepas,
W. P. Colquhoun, & M.J. Colligan (Eds.), Biological rhythms, sleep and shift
work:Advances in sleep research (Vol. 7, pp. 539-552). New York: Spectrum.
Dinges, D. F., Ome, M. T., & Ome, E. C. (1985a). Assessing performance upon abrupt
awakening from naps during quasi-continuous operations. Behavior Research
Methods, Instruments, and Computers, 17, 37-45.
Dinges, D. F., Ome, M. T., & Ome, E. C. (1985b). Sleepdepth and other factors associ-
ated with performance upon abrupt awakening. Sleep Research, 14.92.
Dinges, D. F., Ome, M. T., Whitehouse, W. G., & Ome, E. C. (1987). Temporal place-
ment of a nap for alertness: Contributions of circadian phase and prior wakeful-
ness. Sleep, 10, 3 13-329.
Dinges, D. F., Whitehouse, W. G., Ome, E. C., & Ome, M. T. (1988). The benefits of
a nap during prolonged work and wakefulness. Work and Stress, 2, 139- 153.
Feinberg, I., Floyd, T. C., & March, J. D. (1987). Effects of sleep loss on delta (0.3-3
Hz) EEG and eye movement density: New observations and hypotheses. Electro-
encephalography and Clinical Neurophysiology, 6 7,2 17-22 1.
Feltin, M., & Broughton, R. ( 1968). Differential effects of arousal from slow wave ver-
sus REM sleep [Abstract].Psychophysiology, 5, 23 1.
174 DAVID F. DINGES

Fort, A., & Mills, J. N. (1972). Influence of sleep, lack of sleep and circadian rhythm
on short psychometric tests. In W. P. Colquhoun (Ed.), Aspects of human efi-
ciency (pp. 1 15- 127). London: English University Press.
Gastaut, H., & Broughton, R. J. (1965). A clinical and polygraphic study of episodic
phenomena during sleep. In J. Wortis (Ed.), Recent advances in biologicalpsychia-
try (Vol7, pp. 197-22 1). New York: Plenum Press.
Gillberg, M., & Akerstedt, T. (1982). Body temperature and sleep at different times of
day. Sleep, 5, 378-388.
Goodenough, D. R., Lewis, H. B., Shapiro, A., Jaret, L., & Sleser, I. (1965). Dream
report following abrupt awakening from different kinds of sleep. Journal of Per-
sonality and Social Psychology, 2, 170- 179.
Grosvenor, A., & Lack, L. C. (1984). The effect of sleep before or after learning on
memory. Sleep, 7, 155-167.
Hartman, B. O., & Langdon, D. E. ( 1965).A second study on performance upon sudden
awakening (School of Aerospace Medicine Report No. TR-65-6 I). Brooks AFB,
TX: USAF.
Hartman, B. O., Langdon, D. E., & McKenzie, R. E. (1965). A third study on perfor-
mance upon sudden awakening (School of Aerospace Medicine Report No. TR-
65-63). Brooks, TX: U.S. Air Force.
Haslam, D. R. (1982). Sleep loss, recovery sleep, and military performance. Ergonom-
ics, 25, 163- 178.
Horne, J. A., & Staff, L. H. E. (1983). Exercise and sleep: Body-heating effects. Sleep,
6, 36-46.
Hume, K. I., & Mills, J. N. (1977). Rhythms of REM and slow wave sleep in subjects
living on abnormal time schedules. Waking and Sleeping, 3,29 1-296.
Jeanneret, P. R., & Webb, W. B. (1963). Strength of grip on arousal from full night’s
sleep. Perceptual and Motor Skills, 17, 759-76 1.
Kleitman, N. ( 1963). Sleep and wakefulness. Chicago: University of Chicago Press.
Koulack, D., & Schultz, K. J. (1974). Task performance after awakenings from differ-
ent sleep stages. Perceptual and Motor Skills, 39, 792-794.
Langdon, D. E., & Hartman, B. 0. (1961). Performance upon sudden awakening
(School of Aerospace Medicine Report No. 62-17). Brooks, TX: U.S. Air Force.
Lavie, P., & Giora, Z. (1973). Spiral aftereffect durations following awakening from
REM and non-REM sleep. Perception and Psychophysics, 14, 19-20.
Lavie, P., & Sutter, D. (1975). Differential responding to the beta movement following
awakening from REM and NONREM sleep. American Journal of Psychology, 88,
595-603.
Lubin, A., Hord, D., Tracy, M. L., &Johnson, L. C. (1976). Effects ofexercise, bedrest
and napping on performance decrement during 40 hours. Psychophysiology, 13,
334-339.
Mavromatis, A. (1987). Hypnagogia. London: Routledge & Kegan Paul.
Naitoh, P., Muzet, A., Johnson, L. C., & Moses, L. (1973). Body movements during
sleep after sleep loss. Psychophysiology, 10, 363-368.
Okuma, T., Nakamura, K., Hayashi, A., & Fujimori, M. (1966). Psychophysiological
study on the depth of sleep in normal human subjects. Electroencephalography
and Clinical Neurophysiology, 21, 140-147.
Patrick, G. T. W., & Gilbert, J. A. (1896). On the effects of loss of sleep. Psychology
Review, 3, 469-483.
Pritchett, T. P. (1969). An investigation of sudden arousal from rest: A study of im-
paired performance on an addition task (Doctoral dissertation, University of Ken-
tucky, 1964). Dissertation Abstracts International, 30, 2934B. (University Micro-
films No. 69-20,443)
Rechtschaffen, A., Hauri, P., & Zeitlin, M. (1966). Auditory awakening thresholds in
REM and NREM sleep stages. Perceptual and Motor Skills, 22,927-942.
Rosa, R. R., & Bonnet, M. H. (1985). Sleep stages, auditory arousal threshold, and
body temperature as predictors of behavior upon awakening. International Jour-
nal ofNeuroscience, 27, 73-83.
Hypnopompic Cognitive Performance 175

Rosa, R. R., Bonnet, M. H., & Warm, J. S. (1983). Recovery of performance during
sleep following sleep deprivation. Psychophysiology, 20, 152- 159.
Schacter, D. L. (1976). The hypnagogic state. A critical review ofthe literature. Psycho-
logical Bulletin, 83, 452-48 1.
Scott, J. (1969). Performance after abrupt arousal from sleep: Comparison of a simple
motor, a visual-perceptual, and a cognitive task. Proceedings of the 77th Annual
Convention of the American Psychological Association, 4,225-226.
Scott, J., & Snyder, F. (1968). “Critical Reactivity” (Pieron) after abrupt awakenings
in relation to EEG stages of sleep. Psychophysiology, 4, 370.
Seminara, J. L., & Shavelson, R. J. (1969). Effectiveness of space crew performance
subsequent to sudden sleep arousal. Aerospace Medicine, 40,723-727.
Sewitch, D. E. (1984). NREM sleep continuity and the sense of having slept in normal
sleepers. Sleep, 7, 147- 154.
Sewitch, D. E. (1987). Slow wave sleep deficiency insomnia: A problem of thermo-
downregulation at sleep onset. Psychophysiology, 24,200-2 16.
Sewitch, D. E., Kittrell, E. M. W., Kupfer, D. J., & Reynolds, C. F. (1986). Body tem-
perature and sleep architecture in response to a mild cold stress in women. Physi-
ology and Behavior, 36,95 1-957.
Stones, M. J. (1977). Memory performance after arousal from different sleep stages.
British Journal of Psychology, 68, 177- 181.
Tebbs, R. B., & Foulkes, D. ( 1966). Strength of grip following different stages of sleep.
Perceptual and Motor Skills, 23, 827-834.
Webb, W. B., & Agnew, Jr., H. (1964). Reaction time and serial response efficiency on
arousal from sleep. Perceptual and Motor Skills, 18, 783-784.
Webb, W. B., & Agnew, Jr., H. (1967). Sleep cycling within 24-hr periods. Journal of
Experimental Psychology, 74, 158-1 60.
Webb, W. B., & Agnew, Jr., H. (197 1). Stage 4 sleep: Influence of time course variables.
Science, 174, 1354-1357.
Wilkinson, R. T., & Stretton, M. (197 1). Performance after awakening at different
times of night. Psychonomic Science, 23,283-285.
Williams, H. L., Hammack, J. T., Daly, R. L., Dement, W. C., & Lubin, A. (1964).
Responses to auditory stimulation, sleep loss, and the EEG stages of sleep. Electro-
encephalography and Clinical Neurophysiology, 16,269-278.
Williams, H. L., Morlock, Jr., H. C., & Morlock, J. V. (1966). Instrumental behavior
during sleep. Psychophysiology, 2,208-2 16.
Williams, R. L., Karacan, I., & Hursch, C. J. (1974). EEG of human sleep: Clinical
applications. New York: Wiley.
 CHAPTER 13

A NETWORK MODEL OF DREAMS

ROSALIND CARTWRIGHT

BACKGROUND AND HISTORY

Given that we first learned the technology almost 40 years ago that permitted
us to tune in on the dream channel to identify when dreams occur, how long
they last, and (roughly)how exciting they are, and given the enormous public
interest in dreams, the progress we have made on understanding this part of
our regular mental life has been very slow. There are several obvious reasons
for this. The working conditions are temble. Dreams must be collected from
the sleeping human, who typically sleeps long hours at an inconvenient time,
at night. Sleepers are affected by laboratory sleep circumstances and, thus,
must be accommodated over a period of nights before typical dream data can
be collected. Dreams have not yielded easily to manipulative studies in which,
prior to sleep, subjects were deprived of or satiated with water, social stimula-
tion, and so forth. Nor have clear incorporationscome from exposing subjects
to specific stimuli before or during sleep (Webb & Cartwright, 1978). All in
all, the area has been a tough one for the experimentalist. Subjects tend to
ignore our best efforts and to go along dreaming their own dreams, providing
very little evidence of incorporation (Berger, 1963).
We do not start from zero, however. A good deal of descriptivework has
been conducted and replicated to broadly outline the landscape of the night
mind. Clear differencesexist between REM and non-REM (NREM) reports,
and blind judges can make these discriminationswith 90%accuracy (Monroe,
Rechtschaffen, Foulkes, & Jensen, 1965). However, an examination of audi-
tory awakening thresholds has shown that light sleepers have more NREM
fantasy reports than do deep sleepers (Zimmerman, 1970).Within REM peri-
ods, reports from the beginning are less dreamlike than reports from the end
(Foulkes, 1966). The dreamlike character of the reports is also related to the
179
180 ROSALIND CARTWRIGHT

eye movement density of the epoch preceding the awakening. Low eye move-
ment density leads to no recall or to recall that is “thoughtlike.” High density
is related to more storylike content (Cartwright, Stephenson, Kravitz, & East-
man, 1987). The REM reports also become more dreamlike as the night pro-
gresses, and because eye movement density also increases with time of night,
these two measures can be somewhat confounded.
Generally, the progression across the night shows movement from short,
reality-oriented thought in the first half of the night, when more NREM sleep
is present, to longer, perceptually vivid, nonreality tested experiences in the
second half of the night, when REM sleep is proportionately higher.
The field has been troubled not only by how to grasp dream content
experimentally but also by how to define dreaming itselE Should the definition
focus on its psychological properties whenever and wherever it occurs (e.g., at
sleep onset and in NREM sleep) or on only the mental content that accompa-
nies REM periods? This question remains debatable. In my own work, I prefer
to sample REM periods and to define dreaming as the content of these periods,
which may be scaled as more or less dreamlike psychologically. This decision
was made because so much theorizing about dreaming starts from the proper-
ties of REM sleep, which is highly activated and closely associated with the
PGO spiking that, in man, appears to be related to the REMs that, in turn, are
associated with the dreamlike quality of mentation. Not that dreaming only
takes place in REM periods: Several good studies have shown that dreaming
occurs in NREM periods (Brown & Cartwright, 1978). However, I chose to
examine the REM stage because dreaming is most often associated with REM.
In fact, dreaming probably occurs cyclically throughout the 24-hr cycle, but it
is only sustained long enough to produce the narrative continuity necessary
for these experiences to be recalled when a highly aroused brain state persists
in the absence of demands for attention from external sources (Koulack &
Goodenough, 1976).

QUESTIONS ABOUT DREAMING

Given that we can collect distinctive psychological material by sampling
REM, and given that this material is rather resistant to manipulation, how can
we best proceed with our study? We must answer four questions:
1. Do dreams have intrinsic meaning, or is this added from waking
associations?
2. Does dreaming serve a psychological function or functions, or
are these epiphenomena to the physiological functions of acti-
vated sleep?
3. How do dreams relate to the prior waking state and to personality
traits? Are these relations consistently or alternately, compensatory
or continuous?
4. Do dreams have an impact on the subsequent waking state or on the
more enduring personality traits?
Network Model 181

Table 1
CURRENT MODELS OF DREAM MEANING AND FUNCTION
Crick-Mitchison Hobson-McCarley Breger Freud
Variable (1983) (1977) (1969) (1955)

Meaning None Degraded Emotional Hidden

Function Purging Synthetic Assimilation Safety valve

The theories of dream meaning and function are many, but they can be
grouped into four rough categories (see Table 1).
The Crick and Mitchison (1983) theory relates dreaming to previous
waking but views content as accidental and useless material to be purged to
make room in memory for better data. The Hobson and McCarley (1977)
theory, at least as originally published, postulates that the images originate
randomly from the pons but grants that associated meaning is then synthe-
sized by the cortex to express the current concerns camed over from waking
into sleep. Breger is one of many modem theorists who believe dream meaning
depends on the structure of the individual’s memory networks, particularly
those related to affect, and believe the function of dreaming is assimilative,
“integrating contemporary material into the adaptive solutions worked out in
the past” (Breger, Hunter, 8z Lane, 1971, p. 188). This position relates dreams
both to the immediate and recent presleep experience and to related memory
networks from the past. It implies no immediate impact on subsequent waking
behavior unless the material becomes conscious and is worked on to accom-
modate a change in outlook or behavior on the basis of an insight from a
dream. In this view, dreams have meaning and function but exert no effect on
postsleep states or traits without further integrative work. These theorists are
thus closer to the Freudians, who see the work of dreams as the gratification
of hidden impulses. Dream work is regularly done at night; only with a lot of
additional work during waking, to make the unconscious conscious, will this
effect any change in the waking personality.
Breger’s information-processing hypothesis has been taken one step fur-
ther by Greenberg, Pillard, and Pearlman (1972), who see the assimilation
work of dreams as adaptive for waking functioning in the original stress-pro-
ducing situation. According to these authors, behavior after a night of dream-
ing should be more productive (or at least less distressed) and should contrib-
ute to better coping in the situation that originally provoked anxiety. Jones
(1962), also associated with this position, took one more theoretical step to
relate dreams to postsleep waking by suggestingthat dreams preserve and pro-
tect the waking self-image. In other words, Jones believed that dreams serve
not just a stress-related adaptive function but a full-timejob.
To go beyond the descriptive level and test some of these theories has
been enormously difficult. The controlled studies have largely been failures.
Even the dream deprivation studies have run afoul because of the amount of
182 ROSALIND CARTWRIGHT

individual variability that occurs in studies with small sample sizes. The mean-
ing of the situation to the subject, including both the experiment itself and the
target stimulus; the subjects’ long-term traits and coping styles; and the sub-
jects’ own transformation language as it relates to what is in the memory net-
work of like experiences must be taken into account before much headway
can be made in understanding dream meaning and function.
How then can we best proceed? One good method is to take advantage
of anything promising from past work and to build better designs to clarify
helpful insights. I focused on work that has tested the assimilative function as
that starting point, particularly on the two studies reported by Breger, Hunter,
and Lane in “The effect of stress on dreams” (197 l), which represents a land-
mark in this area. Breger et al. took advantage of the laboratory to collect
dreams through sleep monitoring and, thus, avoided all the problems associ-
ated with home-recalled dreams. However, they used not experimenter-de-
signed manipulations to produce affect but genuine experiments in nature.
One consisted of presleep group therapy sessions and the other consisted of
dream collections prior to and following elective surgery. The studies were not
tightly controlled, but they are important because they took real-life events
that have tremendous personal meaning and traced their impact on dream
content. The studies were conducted in 1964- 1966, more than 20 years ago,
and they stand today as the best in the field.
The thinking is clear: External input is replaced in REM sleep by inter-
nally generated images, bits of memories generated by systems activated from
the presleep period, which becomes increasingly distant in the course of the
night. By looking at these images, we can make inferences about the memory
systems. To do this, we put the system under stress, generate good strong
affect, and watch the work of assimilation take place through dreams. This
can best be done by finding a distressing event that will threaten the currently
organized self-structure in some basic way so that the work of dreams and
their relevance to the specific emotional turmoil can be highlighted and exam-
ined. For this reason, I undertook the study of the dreams of people undergo-
ing divorce. Divorce is a stressful event. No matter how strongly one wishes
to end a bad relationship, no one gets out scot-free. One partner may retaliate
and cause so much trouble that both partners become stressed beyond mea-
sure for a considerable time before they reach a new reorganization as self-
respecting singles. Divorce involves a lot of loss, not only the loss of love but
also the loss of physical security (a home), of economic security (a standard of
living), of a social network of friends and shared relatives, and (most of all) of
self-esteem.
Although bereavement is credited with creating more stress on most re-
cent events scales, many divorcing people have told me that life would be so
much easier if only their spouse would “drop dead.” Divorce provides a good
natural stressor to work with. There is also a large pool of subjects to draw
from (there were 1.5 million divorces last year). The distress lasts a longtime-
most divorces take 1 to 2 years because the courts are crowded-and people
Network Model 183

vary in how well they cope. Some people are initially upset but then take back
their investment in the shared life and reach a new internal reorganization.
Others proceed from the state of stress to focus on the loss, shame, and guilt
of failure and become clinically depressed. This gives us some ability to stratify
a sample into those who are coping well with the stress of a major life change
and those who are not. Depression rates are high among those who are divorc-
ing (approximately 37%according to recent work).
What does it take to get through this life event successfully?Time, good
friends, good genes, good luck, and a good dream system was my hypothesis.
We know that stress disrupts sleep and that stress that culminates in the
hopelessness of depression can create a type of sleep disruption that affects
REM sleep. The work by Kupfer (1976) and Reynolds, Shaw, Newton, Cable,
and Kupfer (1983), which has been replicated by Rush, Giles, Rofiarg, and
Parker (1982); by Gillin and Borbely (1985); and by Cartwright (1983), shows
major affective disorder to have some usual (although not universal) effects on
the first REM period of the night and some additional effects on the distribu-
tion of REM across the night (see Table 2).
If REM sleep is disrupted, what happens to dreams? Do they malfunc-
tion in some way and fail to assimilate affective experience into memory net-
works? Is that why depressed patients feel at their worst first thing in the morn-
ing? If dreaming has an assimilative function, can we see this at work in
subjects who are going through divorce without becoming depressed? If good
copers under stress show productive dreaming, is there a threshold of affect
above which dreaming is overloaded and breaks down? Since we have meth-
ods of indexing depression in both the waking and sleep states (high eye move-
ment density and reduced REM latency), we have measures to help identify
those with good and poor dream functioning.

THE DlVORCE STUDY

This study builds on a previous study conducted with a small sample of 29
divorcees, all women, diagnosed only on the Beck Depression Inventory
(Beck, 1967). (No psychiatric diagnoses were obtained). Nonetheless, there
were clear sleep differences in REM latency and eye movement density (Cart-
Wright, 1983; see Figures 1 and 2). There were also some clear dream differ-
ences between the depressed and nondepressed women (Cartwright, 1986; see
Table 3). Traditional women who had more invested in the homemaker/wife/
mother role were more often depressed; showed reduced REM latency; and
had dreams that were short, past-oriented, negative in motivation, and fea-
tured a self-character who was inadequate to cope in the dream. The nonde-
pressed women, who often chose alternate roles, had longer and more compli-
cated dreams with wider time frames and positive motives. In addition, not
the self but some other character was inadequate in the dream (Trenholme,
Cartwright, & Greenberg, 1984).
The present study was designed to look at a larger sample of people (in-
cluding both sexes) undergoing divorce who had been better diagnosed for
184 ROSALIND CARMRIGHT

Table 2
DIFFERENCE BETWEEN REM SLEEP IN DEPRESSED
AND NORMAL PERSONS
Variable Normal Group Depressed Group

First REM
Latency 80-1 20 min <65 min
Eye movement density 4.5 >1.5 per min
Duration 8-1 2 min 20 min
All night

REM distributionin first half of night 113 112

Eye movement distribution Successively higher with each Sequence irregular
REM stage

depression. The study is still in progress. The target sample contains 300 peo-
ple ( 150 men and 150 women) who are examined at the time of marital separa-
tion when the decision to divorce from a first marriage has been made. All are
unmedicated volunteers who have agreed to return in 1 year for a follow-up.
So far, 170 subjects have been seen (85 men, 85 women). Of these, 87 (40
men, 47 women) have returned for follow-up. Of the original sample, 42 sub-
jects ( 19 men and 23 women) were currently depressed according to the con-
joint criterion of meeting a Schedule of Affective Disorders and Schizophrenia
(SADS) (Endicott & Spitzer, 1978) Research Diagnostic Criteria (RDC)
(Spitzer, Endicott, & Robbins, 1978) diagnosis and yielding a Beck score
above 15. On SADS RDC criteria alone, the figures were much higher, with
47% of the men and 5 1% of the women classed as currently in an episode of
major affective disorder. Most people get over this. At follow-up, only 7.5% of
the men and 6.3% of the women remained depressed.
The original evaluation was used to select 90 subjects (45 men and 45
women) for the sleep laboratory study to examine the effects of divorce on the
sleep and dreams of 30 depressed and 15 nondepressed subjects of each sex
at the initiation of the break-up and a year later. To date, 54 subjects have
been studied for 3 nights, and 36 subjects have been restudied at a 1-year
follow-up.
The study of sleep showed that 18 subjects had short REM latencies (less
than 65 min when averaging Nights 2 and 3), and 13 subjects had high eye
movement density at the first REM stage (greater than 1.5 on the Kupfer
scale), but only 3 subjects had both markers. Both markers were relatively
stable over a 1-year period. The product-moment correlation for REM latency
was .53 and for eye movement density was .55. The factor most heavily con-
tributing to this stability was family history of depression. For those with fam-
ily history positive (n = 21), the correlation between the two REM latency
measures (at the breakup point and 1 year later) was .73. For those without a
Network Model 185

0 FIRST HALF NIGHT

SECOND HALF NIGHT

LOW
B€CK D€PRESSfON SCORES
Figure 1 Eye movement density in first and second half of night in relation to Beck
depression scores.

family history for affective disorder (n = 13), the correlation was .06. Not all
subjects who showed the markers of depression, however, were currently de-
pressed on their own self-reports. Twelve of 18 subjects with short REM laten-
cies had Beck scores above 145, and 8 of 1 3 subjects with high eye movement
density had Beck scores above 15. For those with a SADS diagnosis and self-
reported current depression, short REM latency was associated with higher
dreamlike fantasy ratings (on a 1-5 scale) and higher affect scores (1 = no
efect. 2 = mild,and 3 = strong) for the dream collected from the first REM
than those with normal REM latencies. In contrast, high eye movement den-
sity was associated with lower than normal dreamlike fantasy scores and neu-
tral affect in the first REM period (see Table 4).
Both sleep markers of depression were stable over 1 year, but they
seemed to represent two different subgroups. Short REM latency subjects had
earlier, more affective and storylike dreams, whereas high eye movement den-
sity subjects showed poor dream formation and reduced dream affect. The
numbers are too small for formal testing of an interaction effect at this time,
but the presence of an affective disorder seems to heighten the differences in
dreaming that are represented in these sleep groups.
Things are never as straightforward as one would wish. The so-called
depression markers are not statelike but rather are more traitlike. None of the
3 subjects who showed both markers were currently depressed, but all 3 had
past divorce-related depression episodes by RDC. The two markers seem to
represent two different sleep sub-groups: Those with early REM are affect re-
186 ROSALIND CARTWRIGHT

LOW MODERATE HIGH LO\ MODERATE

IN1 TIAL DEPRESSION FOLLOW- UP 2 YEARS LATER OF
nIGnLr DEPRESSED WOMEN
Figure 2 REM latency at time of divorce and 2 years later in relation to Beck depres-
sion scores.

sponsive in terms of generating more and better fantasies, and those with
higher than normal phasic events at the beginning of the night have disrupted
dreaming. This means we need a large sample size to examine how stable these
findings are over time and what happens to these groups psychologically.
What does the sequential dream content look like in the laboratory for
those subjects who are coping well and those who are not? What theory do
they match, if any? Are Hobson and McCarley (1977) right in stating that,
because REM is turned on by the unthinking pons, the images are not selected
to tell a story but rather give the cortex the work of trying to interpret them as
best it can in the light of its own agenda of emotional concerns? Or are images
selected from a memory network that has been stimulated in response to some
particular type and level of affect aroused in connection with a recent experi-
ence, and do they relate the issues to specific memory components?
My best guess leads me to accept the latter premise. Because dreams
from the beginning to the end of the night often have repeated images, mem-
ory networks seem to be activated not randomly but in response to some per-
sistent issues. Dreams look like creative constructions, drawing on memory
systems that are activated by persistent emotional concerns. Memories are
encoded as unique bits but are organized into networks of related materials
and are decoded according to our needs, without our conscious attention or
direction. Emotional connections are made via earlier memories. These are
then lit up selectively when the brain is activated in REM sleep. Dream mean-
ing relates to the status of our ongoing needs, and dreams function to assimi-
Network Model 187

Table 3
DIFFERENCES IN DREAM CHARACTERISTICS BETWEEN
DEPRESSED AND NONDEPRESSED WOMEN (N = 29)
Dream Characteristic Depressed Group NondepressedGroup

Length Short Long

Time reference Past Wider range of time
Anxiety Low Higher
Self Inadequate Other inadequate
Motivation Abasement, avoidance(negative) Affiliation, achievement (positive)

late new data and reorganize related memories. Once the network is stimu-
lated in the first REM period, it is held in a state of readiness, and other parts
of it are more likely to be stimulated the next time REM occurs when older
elements of the same network are stimulated (see Figure 3).
Which network comes into play is governed by the type and level of
affect carried over from the day. When all is well with our world, the level is
low, and different networks may be stimulated in the same night. When we
are in an unstable state and affect runs high, the system may jam or fail to
work at its best. This appears to happen with higher then normal eye move-
ment density. Then dreams do not seem to progress across the night.
Dreaming appears to be a part of our normal information processing,
which relates waking experiences to past emotional experiences to past emo-

Table 4
DREAM CONTENT SCORES FOR DEPRESSED AND NONDEPRESSED
SUBJECTS
Dreamlike Fantasy' Affe c t b
Short Short
REM High Normal REM High Normal
Group Latency EMD REM Latency EMD REM

Depressed
M 3.40 2.38 3.00 1.90 1.oo 1.60
SD 1.07 0.92 0.71 0.88 0.76 0.55
Nondepressed
M 2.67 2.20 2.70 1.oo 1.40 1.60
SD 0.82 1.10 0.82 0.63 0.89 0.84

Note: EMD = eye movement density. For short REM latency, n = 18; for high EMD, n = 13;
for normal REM, n = 19.
Measured on a 1-5 scale, where 1 = no recall and 5 = well-developed dream. Measured
on a 1-3 scale, where 1 = neutraland 3 = strong.
188 ROSALIND CARTWRIGHT

Flgure 3 A network model of dream construction.

tional paradigms expressed as story images. Dreaming is involved in review-

ing, reorganizing, and rehearsing conceptions of who we are and how we are
doing in our own eyes. The system shows up best when we are under stress
and threatened with a major :ife change that requires new responses. The
system can be overloaded, which seems to happen in those with a biological
vulnerability for REM sleep disruption. This may be adaptive, with reduced
REM latency, or it may be maladaptive, with eye movement storms that
create chaos.

References
Beck, A. (1967). Depression. New York: Harper & Row.
Berger, R. (1963). Experimental modification of dream content by meaningful verbal
stimuli. British Journal of Psychiatry, 109, 722-740.
Breger, L. (1969). Dream function: An information processing model. In L. Breger
(Ed.), Clinical-cognitive psychology (pp. 182-227). New Jersey: Prentice-Hall.
Breger, L., Hunter, I., & Lane, R. ( 1971). The effect of stress on dreams. Psychological
Issues Monograph, 27(7, Suppl. 3).
Brown, J., & Cartwright, R. (1978). Locating NREM dreaming through instrumental
responses. Psychophysiology, 15, 35-39.
Cartwright, R. (1983). Rapid eye movement sleep characteristics during and after
mood-disturbing events. Archives of General Psychiatry, 40, 197-20 1.
Cartwright, R. ( 1986). Affect and dream work from an information processing point
of view. Journal of Mind and Behavior, 7,4 1 1-427.
Cartwright, R., Stephenson, K., Kravitz, H., & Eastman, C. (1987). Life events effects
on REM sleep. Sleep Research, 16, 267.
Crick, F., & Mitchison, G. (1983). The function of dreaming sleep. Nature, 304, 11 1-
114.
Network Model 189

Endicott, J., & Spitzer, R. (1978). A diagnostic interview: The schedule for affective
disorders and schizophrenia. Archives of General Psychiatry, 35, 837-844.
Foulkes, D. ( 1 966). Thepsychology of sleep. New York: Scribners.
Freud, S. (1955). The Interpretation of Dreams. New York: Basic Books. (Original
work published 1900.)
Gillin, J., & Borbely, A. (1985) Sleep: A neurobiologicalwindow in affective disorders.
Trends in Neuroscience, 8, 537-542.
Greenberg, R., Pillard, R., & Pearlman, C. (1972). The effect of dream (stage REM)
deprivation on adaption of stress. Psychosomatic Medicine, 34, 257-262.
Hobson, A., & McCarley, R. (1977). The brain as a dream state generator: An activa-
tion-synthesis hypothesis of the dream process. American Journal of Psychiatry,
134, 1335-1348.
Jones, R. (1962). Ego synthesis in dreams. Cambridge, MA: Schenkman.
Koulack, D., & Goodenough, D. (1976). Dream recall and dream recall failure: An
arousal-retrieval model. Psychological Bulletin, 83,975-984.
Kupfer, D. (1976). REM latency: A psychobiologic marker for primary depressive ill-
ness. Biological Psychiatry, 11, 159- 174.
Monroe, L., Rechtschaffen, A., Foulkes, D., & Jensen, J. (1965). Discriminability of
REM and NREM reports. Journal of Personality and Social Psychology, 2, 456-
460.
Reynolds, C., Shaw, D., Newton, T., Cable, P., & Kupfer, D. (1983). EEG sleep in
outpatients with generalized anxiety: A preliminary comparison with depressed
outpatients. Psychiatric Research, 8, 8 1-89.
Rush, A., Giles, D., Roliivarg, H., &Parker, R. (1982).Sleep EEG and dexamethasone
suppression test findings in outpatients with unipolar and major depressive disor-
ders. Biological Psychiatry, 17, 327-34 1 .
Spitzer, R., Endicott, J., & Robbins, E. (1978). Research diagnostic criteria: Rationale
and reliability. Archives General Psychiatry, 35, 773-782.
Trenholme, I., Cartwright, R., & Greenberg, G. (1984). Dream dimension differences
during a life change. Psychiatric Research, 12, 35-45.
Webb, W., & Cartwright, R. (1978). Sleep and dreams. Annual Review of Psychology,
29,223-252.
Zimmerman, W. ( 1970). Sleep mentation and auditory awakening thresholds. Psycho-
physiology, 6 , 540-549.
 CHAPTER 1 4

NIGHTMARES (DREAM
DISTURBANCES) IN POSTTRAUMATIC
STRESS DISORDER:
IMPLICATIONS FOR A THEORY OF DREAMING

MILTON KRAMER

Sleep that comes on quickly at an appropriate time and place, proceeds with-
out incident to terminate in a smooth awakening at a desirable time, and
leaves us feeling refreshed is what most of us seek from Somnos, the god of
sleep. Dreaming, Freud ( 1900) suggested, maintains the sleeping process by
binding the affective stimuli that might otherwise disrupt the continuity of the
process. How then are we to understand the disturbing dream that seems to
disrupt the process it was designed to protect? Is the disturbing dream, as
Freud ( 1920) suggested, a failure in an otherwise effective mechanism? Does
the dream lead to disruption because it cannot bind the emotion that surges
up periodically during the night?
Certainly this view of the sleepprotective function of dreaming is conso-
nant with two observations about REM sleep that are generally well-estab-
lished (Anch, Bowman, Mitler, & Walsh, 1988). First, if dreams maintain
sleep, and if the longer one sleeps the more likely one is to wake up after the
initial settling down process, then the distribution of REM sleep across the
usual sleep period is appropriate: It is positively accelerated, with more REM
occurring later in the night. This progression contrasts with that of deep sleep
(Stages 3-4), which decelerates (i.e., less occurs) as the night goes on. Second,
a period of dreaming sleep (REM sleep) often ends in a brief arousal. However,
sleepers generally do not remember this brief arousal. The importance of the
arousal for understanding the disturbing dream, which leads the sleeper to
awaken and recall a troublesome event, is that the nucleus for an awakening
190
Posttraumatic Stress Disorder 191

exists in the nature of the dreaming process itself: A brief, unrecalled arousal
is available to be converted into a troublesome awakening.
An example of a disturbing dream will concretize some of the problems
the experience presents and will call attention to some of the factors that pre-
dispose one to disturbed dreaming and that stimulate its appearance. The fol-
lowing dream was reported by a 26-year-old man who was a patient in a Veter-
ans Administration Hospital.
I dreamed I was back in Vietnam and I had thrown a grenade into one of
the Vietnamese huts. I went inside and there was one of the babies blown
up all over the inside of the hut. I woke up and was temfied, nauseated
and crying.

This dream meets our criteria for a disturbing dream: Troubling contents lead
to an awakening associated with negative affect and the recall of a prior trou-
bling dream.
Most dreams have no accompanying feeling. However, when a feeling
tone is recalled, a negative feeling is more common than a positive one (Hall
& Van de Castle, 1966). The romantic notion of the “land of dreamy dreams”
is not borne out by the experience of dreaming.
In fact, in a study (Taub, Kramer, Arand, &Jacobs, 1978)that compared
recalled dreams, recalled nightmares, and the concept of nightmares, the con-
cept of nightmares was reported to be more intense than the experience of the
nightmare. In addition, the nightmare and the recalled dream were reported
to have different contents. The disturbing dream is troublesome, does not
live up to its advance billing, and is clearly distinguishable from the ordi-
nary dream.
Although dream meaning is not the topic of this chapter, I will offer
an interpretation of this dream of infanticide that I arrived at without any
knowledge of the dreamer. The interpretation conforms with an analysis sub-
sequently provided by the dreamer’s therapist. (A more detailed examination
of this dream appears in Kramer & Roth, 1977.)
It was my impression that the dreamer felt he had committed a sanc-
tioned, aggressive (destructive) act, at a distance, against a fragile, possibly
pregnant woman and had caused her (or perhaps her unborn child) to die. The
dreamer did not want to be held responsible for the act (which he felt was
disgusting).
What do we know about the dreamer from the report given by his thera-
pist? He had been in Vietnam. He felt very guilty about some of his behavior
there. In his personal life, he had assaulted his second wife and had been ac-
cused of causing his first wife to abort by having intercourse late in the preg-
nancy and rupturing her membranes. He was the oldest of nine children in a
rural family. His father was a vagrant, violent alcoholic with whom he often
fought in defense of his mother. He had to assist on one occasion in delivering
one of his siblings because the doctor could not come and they could not get
his mother to the hospital. During his treatment, he reported a dream similar
192 MILTON KRAMER

to the one already described in which the baby was his daughter. The first
dream was described in the second session, and the dream about his daughter
was described in the sixth session.
Although one could discuss at length the meaning of this disturbing
dream and could raise a number of questions about predisposing factors, prox-
imal stimuli, the selection of content, internal organization,and so forth, these
speculations are not our current focus. We must focus inbtead on the awaken-
ing with negative affect that the dreamer reports.
The REM dream is often terminated by a brief arousal. The intensity of
the dream experience shows a systematic development across the REM period
(Kramer, Roth, & Czaya, 1975). The periodicity of the development of dream
content intensity is similar to the periodicity of the density of eye movements
seen during REM sleep. The intensity of the dream increases rapidly, peaking
at 10-20 min, then declines and starts up again. This rise and fall of emotion
across the dream period is consonant with the notion that there is a surge of
emotion during REM sleep, which the dream contains or attempts to contain.
The rise and fall of emotion, anxiety, fright, and so forth parallels physiological
activity, as reflected in eye movement density, which suggests a possible psy-
chophysiological parallel. Does the terminal arousal become an awakening
because the intensity surge exceeds the integrative capacity of the dream expe-
rience? This question must be explored further.
If an emotional surge exceeds the integrative capacity of the dream expe-
rience and results in a dream disturbance, is the surge related to the content
of the dream, to the preoccupation with and recall of dreams by the dreamer,
to the emotional condition (psychopathology)of the dreamer (i.e., predisposi-
tional issues), or to the responsivity of the dreamer to the experience (i.e.,
failure to bind or integrate an experience). Kramer, Schoen, and Kinney
(1984a) examined these issues and found that subjects who reported two or
more disturbed dream experiences per week had the same content categories
and dream frequencies as vivid dreamers. The vivid dreamers also reported
much higher dream recall in general than did the disturbed dreamers (89%
vs. 54%).
The disturbed dreamers, however, were more troubled people both from
a trait and a state point of view. Their scores were higher on all Minnesota
Multiphasic Personality Inventory (MMPI) scales and on the emotionally
based scales of the Cornell Medical Index, and they reported a greater number
of psychiatric hospitalizations. Their daily frequency of disturbed dreams ap-
peared to be related to current feelings (e.g., anger, sensitivity, and general
emotion). Of greatest interest was the fact that the disturbed dreamers showed
a greater degree of emotional responsivity to the dream experience than did
the vivid dreamers, and more of them had been described by others as acting
frightened in their sleep.
Dream disturbance sufferers, selected for having bad dreams, are more
psychologically troubled, recall fewer normal dreams, and are more respon-
sive to their dreams than are vivid dreamers. The increased responsivity sug-
PosttraumaticStress Disorder 193

gests an integrative failure that may be affected by the altered emotional state
(increased psychopathology) of disturbed dreamers.
The work of Kramer, Schoen, and Kinney (1984a) would lead to the
prediction that disturbed dreamers would be more responsive to stimuli (e.g.,
the dream experience)than normal subjects. Indeed, that was the case. Above-
threshold tones were presented during sleep to a group of disturbed dreamers
and a group of control subjects (Kinney & Kramer, 1985). Control subjects
responded in 52% of the trials, but disturbed dreamers responded in 93% of
the trials. Related to this was a stimulus identification in 40% of the control
subjects but in only 8% of the disturbed dreamers.
The dreaming process is linked to a process of repeated arousal. Arousal
followed by awakening in disturbed dreamers appears to result from increased
responsivity to the dream experience,which is related to the psychophysiologi-
cal alterations associated with being emotionally troubled. In chronic sufferers
from dream disturbance (e.g., patients with chronic delayed posttraumatic
stress disorder), the arousal threshold is increased (Schoen, Kramer, & Kin-
ney, 1984), which suggests that their emotional focus may be internal (i.e., on
the dream experience) and supports the theory that patients are more respon-
sive to the negative dream experience.
There is reason to believe that the sleep process is related to alterations
in subjective state, particularly affect, mood, or feelings. If the sleep process
has been successful, it is generally agreed that one feels down, tired, and sleepy
before sleep and feels refreshed after sleep. This view of sleep relates, in a gen-
eral way, to the theory that dreaming is related to emotional alterations during
sleep and that bad dreams are related to (and often the consequence of) altered
feeling states during the day (Kramer, Schoen, & Kinney, 1984a).
To examine the relation between sleep and subjective states, we assessed
mood, as indexed by an adjective check list, before and after sleep (Kramer &
Roth, 1972; Roth, Kramer, & Roehrs, 1976). Mood at night, before sleep, is
generally experienced more intensely;that is, one is more unhappy, aggressive,
friendly, clear thinking, or anxious prior to going to sleep at night than in the
morning. Further, moods experienced at night are more variable than those
experienced in the morning. Mood, therefore, generally decreases from night
to morning and becomes less variable.
As mood changed systematically from night to morning, the relation of
sleep physiology (Kramer, Roehrs, & Roth, 1976)and dream content (Kramer
& Roth, 1980)to mood change across the night was studied. The decrease in
unhappiness from night to morning was found to be related to whom and (to
a lesser extent) to what one dreams about, and changes in sleepiness and clear
thinking across the night were related to the amount of non-REM (NREM)
sleep the dreamer had experienced. These observations were utilized in formu-
lating the mood regulatory theory of sleep (Kramer, in press). From the per-
spective of the dream, the theory suggests that the dream subserves a selective,
affective regulatory function (Kramer & Roth, 1978). The dream may be
194 MILTON KRAMER

thought of as an “emotional thermostat” that corrects the level and range of

the person’s mood.
If dreaming functions to protect sleep by absorbing the emotional surge
that appears to occur during REM sleep, then dreaming ought to be related to
the emotional state of the dreamer. That is indeed the case: Changes in emo-
tional state across the night are related to the content of the dream.
Multiple dream periods across the night are part of the periodicity of
REM sleep. These multiple dreams are also related to the emotional preoccu-
pations or current concerns of the dreamer. Just as pre- and postsleep mood
is related to the intervening dreams and sleep physiology of the night, so is the
dream content of the multiple REM periods of the night related to what is on
the dreamer’s mind before and after sleeping (Kramer, Roth, Arand, & Bon-
net, 198 1; Kramer, Moshiri, & Scharf, 1982; Kramer, Roth, & Palmer, 1976).
Two principal patterns of thematic dream development across the night
are discernable (Kramer, Whitman, Baldridge, & Lansky, 1964): A progres-
sive-sequential pattern in which problems are stated, worked on, and resolved
and a repetitive-traumatic pattern in which the problem is simply restated and
no progress occurs.
The effectiveness of the night’s dreaming varies from night to night
within the same subject. This may result from the differential pattern of
dreaming across the night. If one has experienced a progressive-sequential
dream pattern, one may also experience a positive alteration in emotional
state. The problem solving that takes place is emotional in nature. If the prob-
lem that one goes to sleep with is simply restated and is not solved, then a less
successful night’s dreaming has occurred. It could be through this mechanism
of success or failure in problem solving (French & Fromm, 1964) that the
affective alteration takes place. The problem-solving function may also paral-
lel the change in the degree of unhappiness and may be related to the appear-
ance of appropriate character types.
It may be useful to illustrate the patterns of dream development by re-
viewing the experiences of a subject who slept in the experimental laboratory
who showed both patterns of dreaming. Subjects do show both patterns, which
suggests that there is not universal success in altering the emotional preoccu-
pations of the previous day. This may account for some of the variability in
how one feels on awakening from a night’s sleep.
SEQUENTIAL DREAM PATTERN
The sequential pattern shows an alternating ascendancy of disturbing and re-
active motives and a concomitant tension accumulation, discharge, and re-
gression pattern. Each dream report obtained was from an REM-period awak-
ening (Kramer et al., 1964). This illustration is from a female subject on the
sixth dream night.
Sequential Dreams
1. This little girl was asleep. She was being real cute, prolonging things
for money or to stay in the hospital longer.
PosttraumaticStress Disorder 195

2. I passed Frank‘s wife in a car. She saw me come . . . she pulled

away. I got kind of mad. I decided that it didn’t make any difference.
3. I was playing tennis. I hit it back real hard. We won the game.
4. A patient didn’t need the doctor after all. She started out thinking
she needed a doctor but she didn’t. She had a big bandage on her
stomach.
5. Doctor was not able to treat patient because he was not properly
licensed. The patient is planning to use surgery against the doctor.

Sequential Interpretations
1. The thematic pattern focuses on the subject’s wish to depend on or
cling to the doctor and hospital even if she has to claim illness, but
the cuteness reveals a seductive motive as well.
2. The subject expresses, vis-A-vis the wife of a friend, the feared and
expected rejection, which she tries to minimize.
3. This dream can be seen as a turning point because the victory in the
tennis game appears to discharge effectively the tension generated by
the conflict between the wish to be close to the doctor/experimenter
for care and love (the disturbing motive) and the feared rebuff of
abandonment (reactive motive). The subject switches to a successful
competition that she wins with a partner and thus finds a successful
solution. The inference is that whatever she was struggling with has
been conquered.
4. The successful victory with a partner in the previous dream permits
the subject to deny her need for the doctor, which she admits she
once had. The bandage provides evidence that the need continues.
5. Recognition of her continuing repressed need for the doctor for care
and love causes the patient to intensify her rejection of him by being
critical of the doctor’s qualifications and by expressing her wish to
get even or to hurt him. It is the familiar double assertion: “Not only
do I not need you, but you’re no good anyway.’’
The sequential pattern in this series expresses a dependent/sexual longing to-
ward the experimenter/doctor, which leads to a feared (expected) rejection by
the wife/mother in the second dream. The subject masters the conflict in the
third dream by an aggressive victory with her own partner. The fourth dream
reveals a rejection of the previous need, although a recognition is present that
the need still exists. In the last dream of the night, a more intense rejection in
the form of an attack on a doctor serves to deny the need further.

REPETITIVE DREAM PATTERN

The repetitive pattern provides a restatement of the conflict, often in different
settings and at different degrees of concreteness or abstraction. Although some
interrelation of events is suggested, the dreams predominantly restate the situ-
ation within a narrow range without sequential progression. Each dream was
196 MILTON KRAMER

obtained from a female subject (on the third dream night) after an REM-pe-
riod awakening (Kramer et al., 1964).

Repetitive Dreams
I. Somebody was lost. It was a dog and they were trying to find out
where it lived. It wasn’t my dog, though. I wasn’t lost. This person
who was lost was always fumbling around leading everybody else
around because he didn’t know what he was doing. Some boy, I
think. Somehow we had telephone numbers trying to find the right
one. It was supposed to be that little boy that was lost.
2. They filled up the car. There wasn’t enough room, unless I went
back with the people we went back with before. I could go back with
someone else. The place we were going was an orphanage
someplace, some house, a place like that.
3. I was dreaming about visiting, I think it was some EEG laboratory,
or something like that where the mothers could leave their children,
and they could go shopping. I doubt whether they could, there
wouldn’t be enough room for all of these people.

Repetitive Interpretations
1. The boy (association = experimenter) is seen as lost and misleading
others. Not the subject but the boy is lost. There is concern in the
dream that she will be misled by the experimenter because of his
inexperience, so she tries to call home.
2. In a setting of abandonment (orphanage),the subject hopes
anxiously that there will be room for her to return home. She has a
ride home, but the car is crowded.
3. The subject is in the laboratory temporarily and her mother is going
to return for her (implied). There isn’t enough room. In all the
dreams of the night, the subject deals with her fear of being
abandoned and her method of recontacting her family: calling on
the phone, riding in a car, or being picked up.
These patterns are elaborations of the general thesis that dreams may
subserve a problem-solving function (French & Fromm, 1964). Through this
effort at emotional problem solving, the dream may succeed or fail to contain
the emotional surge that leads to the arousal/awakening that is the hallmark
of disturbed dreaming. Difficulty in emotional problem solving in waking life
is certainly more characteristic of the psychopathologically disturbed subjects
who have been found to be more responsive to arousal/awakening after a
REM period and who report more bad dreams.
Sleep is generally a successful process, although there is much concern
expressed about a disturbed night’s sleep. Given that the process is generally
successful, one would expect to see the progressive-sequential pattern occur
Posttraumatic Stress Disorder 197

more often than the repetitive-traumatic pattern: In the two laboratory sub-
jects, that was the case (50%vs. 32%;Kramer et al., 1964).
If the progressive-sequential pattern is more common, one ought to see
dream content change across the REM periods of the night. This change in-
deed was found when the dreams were analyzed of 22 subjects who slept for
20 consecutive nights in the laboratory and described dream content from
each REM period of the night (Kramer, McQuarrie, & Bonnet, 198 1). Word
count and dream content showed systematic and statistically significant
change across the night. Three character variables in particular showed an
increase and decrease across the night, in keeping with the inverted U-shaped
curve that the progressive-sequential pattern describes. And the character vari-
able was best correlated with change on the dimension of unhappy mood
across the night (Kramer & Roth, 1980).
A number of studies have shown evidence that dreams are a regular and
orderly process (Kramer, 1982). The dreams of the same person on one night
are distinguishable from those on another night. This supports the notion that
the dreams of the night are indeed focused on a single topic, whether they are
handled in a progressive-sequential or a repetitive-traumatic manner. Dreams
in individuals are differentiable,but the dreams of one night are highly corre-
lated with the dreams of the next night. Dreams, therefore, are unique to the
individual dreamer.
The emotional preoccupations or current concerns of the dreamer, as-
sessed from Thematic Apperception Test (TAT) stories (Kramer, Roth, &
Palmer, 1976) and presleep and postsleep verbal samples (Kramer, Roth, Ar-
and, & Bonnet, 198 I), have been shown to be significantly correlated with the
content of a night's dreams. The dream content is linked thematically
(Kramer, Moshiri, & Scharf, 1982) both to the current concerns of the subject
before sleep and to the emotional preoccupation of the dreamer the subse-
quent morning. The notion of emotionally relevant material being processed
by the dreamer during his or her dreams is supported by these observations.
Thus, dream content is related across the waking-sleeping-wakingcontinuum.
What occurs at times of emotional upheaval, when disturbing dreams
occur with increased frequency, such as those that occur in chronic delayed
posttraumatic stress disorder (CDPTSD)? There may be a process of reactiva-
tion of previously troubling life experiences. This reactivation may be a re-
sponse to a breakdown in a current emotionally significant relationship (e.g.,
marriage), and the previously troubling experiences may become a metaphor
to express feelings such as hostility and rage stirred by the breakdown. This
previous experience is then linked to earlier childhood experiences.
This rather traditional psychodynamic view is supported by a study of
the sleep (Kramer & Kinney, 1988), dreams (Kramer, Schoen, & Kinney,
1984b), and psychological states (Kinney & Kramer, 1985) of CDPTSD pa-
tients. The sleep and dreams of Vietnam combat veterans with disturbed
dreaming were explored. They were then compared with combat-matched
control subjects without CDPTSD and with normal nonveterans.
198 MILTON KRAMER

Dream-disturbed Vietnam combat veterans with CDPTSD have clear dis-

turbances in their sleep, both spontaneously and in response to nonspecific
stimuli (Schoen et al., 1984). These sleep difficulties may help sustain their
disturbance. These subjects sleep more poorly and awaken and arouse sponta-
neously in NREM sleep (across the night) more often than control subjects.
In response to auditory stimuli above waking thresholds (Kinney & Kramer,
1984), dream disturbed subjects have more physiological systems activated
in NREM sleep during the first half of the night (especially respiratory and
motor) than control subjects but report less awareness of the source of the
arousals.
This pattern of disturbance in dream-disturbed Vietnam combat veter-
ans suggeststhat their sleep is potentially more disruptable by random (mean-
ingless) noise that can lead to an arousal/awakeningwithout subjects knowing
what the source of the arousal/awakening is. Subjects are then in the position
of being awakened without cause, which is often experienced as frightening.
Were these dream-disturbed subjectshypervigilant? In Vietnam, combat
veterans had felt it wasn’t safe to sleep in the field. Therefore, one might predict
that they would have lower arousal thresholds to auditory stimuli. On the con-
trary, they had higher arousal thresholds (Schoen, Kramer, & Kinney, 1984)
and appeared to be more inwardly vigilant than externallyvigilant. The source
of their disturbance seemed to exist on the inside rather than on the outside.
An investigation of CDPTSD subjects’ internal experience (i.e., dream
content) showed that they recalled dreams more often than did control sub-
jects (55% vs. 45%) but were alone in reporting military references in their
dreams (Kramer, Schoen,& Kinney, 1984b).The threat, if the military experi-
ence indexes a threat, is now within the individual, and he is focused on it.
Interestingly, only half of the dreams of the CDPTSD subjectswere mani-
festly about the Vietnam experience. The other half were about other aspects
of their lives. Therefore, these subjects were not exclusively focused on Viet-
nam, although in the waking state they certainly felt this to be the case.
It might be helpful to illustrate the dream findings from these CDPTSD
patients with a case example. A Vietnam veteran volunteered to sleep in the
laboratory at the Veterans Administration Hospital. He was stymied in his
therapeutic work on his problems. Two years earlier, his wife had left him.
Following that, he became more and more troubled and preoccupied with his
Vietnam experiences. Approximately 4 months earlier, he had quit his job
and had given his gun to his boss for fear of getting into difficulty with it. A
brief synopsis of his dreams from five consecutivedream periods of one night
follows (Kramer, Schoen, & Kinney, 1987).
1. He tries to cross a river, and a big guy is trying to help him. It seems
that they are in the Mekong Delta. Every time they try to get to the
other side, they are back at the beginning.
2. He is a young boy at home and is fightingwith one of the
neighborhood kids. His brother is around. The kid they are fighting
with is later killed in Vietnam.
PosttraumaticStress Disorder 199

3. He goes with others to a construction site. He is a sapper, like in

Vietnam. He blows the site up. The consequence of blowing up the
construction site is that his brother’s house is flooded.
4. He remembers being a young boy at his parents’ house. There is
some argument and he leaves. He writes them post cards without a
forwarding address. When they try to contact him, he has moved
elsewhere.
5. He is driving very fast, and his wife is worried that he will have an
accident.
The patient has clearly interdigitated his current life stresses, his Vietnam ex-
periences, and his childhood. A brief review of the dynamic implication of the
material may be useful.
In the first dream, the subject might well be referring to his therapeutic
efforts with the doctor (the big guy). He is making efforts to go across a danger-
ous area, the Mekong Delta, to get to the other side. This could be a metaphori-
cal statement about going from one state to another, from one position to
another, or from one place in life to another. The more he tries, the more he
ends up in the same place. His dream may reflect the stalemated condition he
alluded to in coming to seek help.
In response to the subject’s frustration with change, he recalls an earlier
difficulty in succeeding. Themes related to moving ahead, and problems asso-
ciated with moving ahead, are clearly repeated in his dreams. He is struggling
with a neighborhood youngster in his dreams, and his brother is in the back-
ground. The young man with whom he is struggling is one who later dies in
Vietnam. Peer competition (fighting, hostility) in his dreams is linked to death
in Vietnam. Sibling competitive issues are further suggested by the brother’s
presence.
In the third dream, the competitive situation becomes a bit clearer, as
does the link to Vietnam. In performing his job in Vietnam, which is destruc-
tive but legitimate, he damages things, which leads to the damage of his broth-
er’s home. One recalls that in the second dream he was fighting with a peer,
with his brother in the background. In the third dream, the direct consequence
of his action is to bring damage to his brother’s possession. The potential for
a sibling problem is clearly underlined.
A tension with his parents is reflected in the fourth dream. The point of
sibling rivalry is to gain the approbation of his parents. He flees from tension
with the parents (i.e., the argument) but longs to remain in contact with them.
He writes them and invites a response, but directs them to where he has been,
not to where he is going. He cannot form an effective relationship with his
parents, although he would like to. The dream has a Tom Sawyer quality:
“They’ll miss me when I’m gone.”
The last dream reflects a current peer (but possibly a parental) relation-
ship as well. His wife (i.e., a peer/parent) is concerned about him. She’s afraid
that his being out of control will lead to an accident. She cares about him as
he wants his parents to care.
200 MILTON KRAMER

The subject was surprised that a significant portion of his dreams focused
on his family. We learned later that, followingthe experience in the laboratory,
he went back and sought out his wife to see if there could be a reconcilia-
tion. His therapeutic work was rejuvenated by work on both past and current
problems.
The dreams of this subject reflect his use of the Vietnam situation in a
metaphorical sense to reflect his difficulties (Le., his difficulties in changing).
His hostility, which goes back to childhood and is directed toward his brother,
is manifested again in Vietnam, but it affects both a brother in arms and his
own brother from the past. His act of sibling rivalry results in a disruption of
his relationship with his parents. He is ambivalent about this disruption and
flees from them but also wishes to remain in contact with them. He cannot
resolve this tension. His plight leads him to be potentially out of control, per-
haps because of the intensity of his emotion. This potential loss of control is
reflected in his current situation with his wife. It is the breakdown with his
wife that stimulated his current difficulty, which is focused on Vietnam.
It is apparent from this case that the Vietnam experience can become a
metaphor for chronic interpersonal difficulty with peers and parental figures,
both current and past. The wish to dominate the peer leads to competition
with destructive consequences. It serves to alienate the subject from the par-
ents (and parental support figures) and potentially leads to a loss of control.
This example illustrates three levels of problems in the Vietnam veteran.
A breakdown in current life situations echoes and stirs up similar problems
from the past. Vietnam then serves as a metaphor to express the difficulties.
Those aspects of Vietnam that are in focus currently are those that have some
meaning as determined by events in the subject’s past, prior to Vietnam. The
Vietnam experience is a factor, but not the only factor, that dominates the life
of the Vietnam veteran.
The current precipitant is the new trauma in CDPTSD. In the illustrative
case, the current “universal stressor” is marital disruption. These stressful
events may occur in predisposed individuals. The Vietnam combat control
group may be a high-risk group for developing CDPTSD. The sleep of the con-
trol group (Kramer & Kinney, 1988)was disturbed when compared with that
of the normal group: 87% sleep efficiency, eight awakenings per night, more
light sleep, less REM sleep, and decreased deep sleep. Furthermore, their
dream recall was strangely low both overall (45%) and in REM sleep (50%)
(Kramer, Schoen, & Kinney, 1984b).
It may be that these combat veterans without dream disturbances are
vulnerable to developing CDPTSD in response to a current stress. This develop-
ment would facilitate the return of dreams about Vietnam, and those dreams
would also be expressive of the subject’s current problems. Disturbing dreams
and troubled awakenings are reinforced (perpetuated) by nonspecific, inter-
mittent stimuli (apparently meaningless noise) at night.
The experience of the Vietnam veteran with CDPTSD may be placed in a
more traditional psychodynamic perspective. The Vietnam experience re-
PosttraumaticStress Disorder 20 1

flected but did not completely determine their later lives. The patients with
CDPTSD had a reactivation of memories of Vietnam brought about by some
breakdown in their current life equilibrium. The case illustration involved a
disruption in a marital relationship. The current preoccupations of the patient
echoed both his Vietnam experience and his childhood experience. This form
of illness is not particularly unusual. Once the dynamic is started, psychophys-
iological factors (i.e., alterations in sleeping and dreaming) may serve to keep
it alive.
Theories of dreaming need to address dreams that go on automatically,
outside of awareness, and those that enter awareness and have the potential
for a more direct effect on the dreamer’s consciousness (Kramer, 198 1). One
needs both an assimilative and an accommodative theory of dreaming, with
the former being reductive and the later being potentially transforming. The
selective-affective mood regulatory theory of dreaming (Kramer & Roth,
1978; Kramer, in press) is an example of a n assimilative theory. The experi-
ence of the disturbing dream opens the possibility for a n extension of this
assimilative view to encompass some degree of transformation and accommo-
dation,

References
Anch, A. M., Bowman, C. P., Mitler, M. M., & Walsh, J. K. (1988).Sleep: A scientific
perspective. Englewood Cliffs, NJ: Prentice-Hall.
French, T., & Fromm, E. (1964).Dream interpretation. New York: Basic Books.
Freud, S. (1900). The interpretation of dreams. In J. Strachey (Ed. and Trans.), The
standard edition of the complete psychological works of Sigmund Freud (Vols. 4-
5 ) . London: Hogarth Press.
Freud, S. (1920). Beyond the pleasure principle. In J. Strachey (Ed. and Trans.), The
standard edition of the complete psychological works of Sigmund Freud (Vol. 18).
London: Hogarth Press.
Hall, C., & Van de Castle, R. ( 1966). The content analysis of dreams. New York Apple-
ton.
Kinney, L., & Kramer, M. (1984).Sleep and sleep responsivity in disturbed dreamers.
Sleep Research, 13, 102.
Kinney, L., & Kramer, M. (1985). Personality differences among disturbed dreamers,
“sick” controls and normal veterans. Sleep Research, 14, 130.
Kramer, M. (1981). The function of psychological dreaming: A preliminary analysis.
In W. P. Koella (Ed.), Sleep 1980 (pp. 182- 185). Basel, Switzerland:Karger.
Kramer, M. (1982). The Psychology of the dream: Art or science. Psychiatric Journal
of the University of Ottawa, 7, 87- 100.
Kramer, M. (in press). The mood regulatory function of sleep. In A. Moffit &
M. Kramer (Eds.), Thefunction of dreaming. Albany, NY: SUNY Press.
Kramer, M., & Kinney, L. (1988). Sleep patterns in trauma victims with disturbed
dreaming. Psychiatric Journal of the University of Ottawa. 13, 12- 16.
Kramer, M., McQuame, E., & Bonnet, M. (1981). Problem solving in dreaming: An
empiricaltest. In W. P. Koella (Ed.), Sleep 1980 (pp. 174- 178).Basel, Switzerland
Karger.
202 MILTON KRAMER

Kramer, M., Moshiri, A., & Scharf, M. (1982). The organization of mental content in
and between the waking and dream state. Sleep Research, 11. 106.
Kramer, M., Roehrs, T., & Roth, T. (1976). Mood change and the physiology ofsleep.
Comprehensive Psychiatry, 17, 161-165.
Kramer, M., & Roth, T. (1972). The mood-regulatory function of sleep. In W. P.
Koella & P. Levin (Eds.), Sleep 1972 (pp. 563-57). Basel, Switzerland: Karger.
Kramer, M., & Roth, T. (1977). Dream translation, Israel Annual of Psychiatry, 15,
336-3561.
Kramer, M., & Roth, T. (1978). The dream as selective affective modulator. Acad.
Form, 22, 13.
Kramer, M., & Roth, T. (1980). The relationship of dream content to night-morning
mood change. In L. Popoviciv, B. Asgian, & G. Badiv (Eds.), Sleep 1978 (pp. 62 1-
624). Basel, Switzerland: Karger.
Kramer, M., Roth, T., Arand, D., & Bonnet, M. ( 198 1). Waking and dreaming menta-
tion: A test of their interrelationship, Neuroscience Letters, 22,83-86.
Kramer, M., Roth, T., & Czaya, J. (1975). Dream development within a REM period.
In Sleep I974 (pp. 406-408). Basel, Switzerland: Karger.
Kramer, M., Roth, T., & Palmer, T. (1976). The psychological nature of the REM
dream report and T.A.T. stories. Psychiatric Journal of the University of Ottawa,
I , 128-135.
Kramer, M., Schoen, L. S., & Kinney, L. (1984a). Psychological and behavioral fea-
tures of disturbed dreamers. Psychiatric Journal of the University of Ottawa, 9,
102- 106.
Kramer, M., Schoen, L. S., & Kinney, L. (1984b). The dream experience in dream-
disturbed Vietnam veterans. In B. A. Van der Kolk (Ed.). Washington, DC:Amer-
ican Psychiatric Press.
Kramer, M., Schoen, L. S., & Kinney, L. (1987). Nightmares in Vietnam veterans.
Journal of the American Academy of Psychoanalysis, IS,67-8 1.
Kramer, M., Whitman, R., Baldridge, B., & Lansky, L. (1964). Patterns of dreaming:
The interrelationship of the dreams of a night. Journal of Nervous and Mental
Disorders, 139,426-439.
Roth, R., Kramer, M., & Roehrs, T. (1976). Mood before and after sleep. Psychiatric
Journal of the University of Ottawa, 1, 123-127.
Shoen, L. S., Kramer, M., & Kinney, L. (1984). Auditory thresholds in the dream
disturbed. Sleep Research, 13, 102.
Taub, J., Kramer, M., Arand, D., & Jacobs, G. (1978) Nightmare dreams and night-
mare confabulations, Comprehensive Psychiatry, 19,285-29 1.
 INDEX

Across States Connectionist Model of Imagery reaction time. See Reaction time
and Thought (ASCIT), 5,7,9 time estimation. See Time perception
Activation, Input Source, Modulation (AIM), vigilance, 163
25-40 Cornell Medical Index, 192
Alzheimer’s disease, 100-101
Amnesia, 80,83,85,98, 134, 160 Demand characteristics, 1 1 1, 152
anterograde, 69,72-74, 144 Depression, 184-1 88
waking, 85,92,97 Dichotic listening, 134
Anesthesia, 127-135 Disorders of Initiating and Maintaining Sleep
awareness under, 128 (DIMS). See Insomnia
cognitive processing under, 128-1 33 Dream deprivation, 181- 182
Animal models, 27 Dreams, 179- 189
Anxiety, 152-1 54 Activation-Synthesis model, 7- 10, 17
Artificial intelligence, 73 and affect, I9 1- I97
Auditory item presentation bizarreness, 8- 10, 17-20
air conduction, 94 Crickand Mitchison theory, 17-18, 181
bone conduction, 94 and depression, 183-1 88
Atonia, 33 Freud’s theory of, 181, 190
Awakenings, 62,155, 168-169 function of, 181-183, 186-188
lucid dreams, 36-38,109-126
Beck Depression Inventory, 183-1 86 network model, 179- I89
Behavioral response NREM dreams, 180
after anesthesia, 130 recall of, 90,96, I66
in sleep, 77-87 repetitive dreams, 194- 197
response latency, 64-65,80 reports, 160, 179-180, 198-200
Blind sight, 134 and stress, 182-202
Brain metabolism, 170-1 7 1 combat, 198-201
divorce, 182-188
Cerebellum, 14, 16-17,20,73
Circadian rhythm, 151-152, 161,170, 172 Electroencephalogram(EEG)
Classical conditioning, 70-75, 128 alpha, 60,62,69-72,78-79,83-85,88-
conditioned stimulus, 7 I 92, 120,149
in sleep, 70-72,78,89 K-complexes,49,7 1,95
of tolerance to hypnotics, 145 NREM sleep. See NREM sleep
unconditioned stimulus, 7 1 REM latency, 183- 188
Cognitive performance tasks REM sleep. See REM sleep
clock reversal, 163 sleep efficiency, 140
letter cancellation, 163 sleep onset latency. See Sleep onset
logical reasoning, 162 latency
memory tasks, 162 sleep spindles, 63, 149, 151
mental arithmetic, 161-168 slow wave sleep, 140, 142, 144, 165- I 72

203
 204 INDEX

spectral analysis, 64, 141 classical conditioning. See Classical

speeding, 62 conditioning
stage shifts, 62 escape/avoidance, 58-59
EMG, 62,141 extinction, 67
submental, 4 habituation. See Habituation
Event-related potentials. See evoked reinforcement, 58-60
potentials in sleep. See Sleep learning
Evoked potentials, 12-1 3,43-57
auditory, 45-46,128 Memory
brainstem evoked response, 45-46 declarative, 72-75
contingent negative variation, 44,47-50, and dreams, 186- 187
52 echoic, 50
middle latency responses, 46 episodic, 144
mismatch negativity, 44, 50-5 1 explicit, 97-99
readiness potential, 44,52 implicit, 52,97-99
selective attention, 44 long-term, 68-75,89-90
semantic processing, 44,52-53 procedural, 52,72-75
in sleep, 44-53 recall, 96-97, 128- 129, 131- 133
in NREM sleep, 45-46,49-5 I recognition, 96-99, 128-1 3 I
in REM sleep, 48-52 retrieval, 9 1,93-94, 128- 129
Eye movements, 3-21,29-3 I , 33,37,95,110, short-term, 68-75,89-90
112-117, 119,180, 183-188 and sleep, 67-76
and visual imagery. See Visual imagery state-specific, 70,9 1-94
alcohol, 93
Galvanic skin response, 89, 130 drugs, 93
mood, 93
H-reflex, I 12 Mental states
Habituation, 70-72,74,89 subjectivity, 26-27
Heart-rate, 7 1, 89, 1 I2 isomorphism, 27
Hippocampus, 69,72-75 Middle ear muscle activity (MEMA), 4,9
Homophones, 98-99 Minnesota Multiphasic Personality Inventory
Hypnogogic state, 160,17 I - 172 (MMPI), 155,192
Hypnopompic state, 159- 175 Multiple Sleep Latency Test (MSLT), 141
Hypnosis, 69,77-86, 129, 133-134
posthypnotic amnesia, 85-86 Naps, 152-153,161, 163-167
Hypnotics, 139- 147 Neural nets, 17- 18
barbiturates, 142 Neurotransmitters
benzodiazepines, 140- 145 aminergic, 32-35,38-39
Hypnotic susceptibility, 81-82, 85, 101-105 cholinergic, 32-35,38-39,48
Harvard Group Scale of Hypnotic Nightmares, 122,190-202
Susceptibility, 8 I-82,85 Night terrors, 38
Stanford Hypnotic Susceptibility Scale, Nocturnalmyoclonus, 139, 141,143, 155
81-82,85,104 NREM sleep
evoked potentials in. See Evoked
Information processing, 67-69, 128-129, 181, potentials
187 neurophysiology,4,9, I I, 14-15, 18-19,
Insomnia, 139- 147, 150- 157 26-30,34-38
psychophysiological, 140, 145 sleep learning. See Sleep learning
subjective, 139- 140, I45 visual imagery in. See Visual imagery

Korsakoff patients, 98, 133 Orienting response, 128- 130

Language processing, 52-53 Parallel processing, 7-8,

Learning Periodic leg movements. See Nocturnal
acquisition, 67 myoclonus
INDEX 205

Periorbital Integrated Potential (PIP), 4,9 dishabituation, 7 I

Placebo, 145 habituation. See Habituation
Pontine-geniculate-occipital(PGO) activity, and hypnosis, 83-85,101-105
3-21,28-31,33,38, 180 learning set, 10 I- 105
Posttraumatic stress disorder, 190-202 in NREM sleep, 59,6 I , 64,84-85,90-9 I ,
96, 103-104
Reaction time, 47-48, 162 in REM sleep, 59,64-65,69,78-85,96,
REM sleep 103-104
dreams. See dreams Soviet Union studies, 83-85,89,91,95,
evoked potentials. See Evoked potentials 101- 103
neurophysiology, 3-2 I, 26-36 stimulus control, 58-66
phasic events. See Eye movements, subject factors, 99-105
Middle ear muscle activity, or Sleep onset latency, 91, 143, 148
Periorbital integrated potential perception of, 148-158,
sleep learning. See Sleep learning Sleepiness, 159- I75
visual imagery. See Visual imagery daytime, 63,141
Research diagnostic criteria, 184 postdormital, I59
Respiration, 1 12 Somnambulism, 90
Reticular formation, 6, 11- 13, 15, 17,47-48 Subliminal perception, 134
Retina, 10,26 Suggestibility.See Hypnotic susceptibility
Reverie, 38-39,81,159-175
EEG during, 167- 168 Temperature
ambient, 144
Saccade. See Eye movements body,45,64, 143-144, 151-154, 169-
Schedule of Affective Disorders and 171
Schizophrenia, 184 Thematic Apperception Test (TAT), 197
Sleep Timepercpetion, 151-154, 163, 166
auditory thresholds, 35,45, 142, 145, 150
awareness during, 60,109-126 Vestibular activity. See Cerebellum
perception of, 142 Visual imagery
Sleepapnea, 139, 141, 143, 155 brightness, 19
Sleepdeprivation, 38, 141, 159-169 clarity, 19-20
Sleep drunkenness, I59 color saturation, I9
Sleep inertia, 159-163 dreams, 10- 14, 18-20
and sleep depth, 168- I69 and eye movements, 18-20, 1 19
Sleep learning, 58-76,78,88-108 measurement of, 19-20
classical conditioning. See Classical NREM sleep, 18-20
conditioning REM sleep, 18-20