Sleep Medicine Reviews, Vol. 5, No. 5, pp 365–376, 2001
doi:10.1053/smrv.2001.0151, available online at http://www.idealibrary.com on
Beta EEG activity and insomnia
Michael L. Perlis1,2, Helli Merica3, Michael T. Smith1and
Donna E. Giles1,4
1Sleep Research Laboratory, Department of Psychiatry, University of Rochester,2Department of Cognitive and
Social Psychology, University of Rochester,3Neurophysiology Unit, Department of Psychiatry, Geneva University
Hospital, Switzerland and4Department of Neurology, University of Rochester, USA
analysis, beta, EEG
elevated at around sleep onset and during polysomnographic sleep in patients with
insomnia. These findings suggest that insomnia may be characterized by central
nervous system (CNS) hyperarousal. In this article, the seven studies are critically
reviewed, two theoretical perspectives on beta EEG are presented, and the concept
of hyperarousal as a three component process is discussed.
2001 Harcourt Publishers Ltd
To date there have been seven studies which find that beta EEG is
the insomnia (in the absence of the original pre-
cipitating factors) and/or contribute to the con-
ditioned arousal that appears to be characteristic
of chronic insomnia [e.g. 4–6]. One of the most
potent aspects of the behavioral perspective on
insomnia is that it clearly implicates that the “per-
petuating factors” are the ideal targets for the
behavioral treatment. Such factors include, for ex-
ample, the tendency to unduly extend sleep op-
portunity in the effort to increase total sleep time.
The behavioral theory does not, however, address
what the “conditioned arousal” of chronic insomnia
is or how this factor may be related to sleep
initiation and/or maintenance problems and/or to
phenomenology of insomnia (i.e. the tendency of
patients to overestimate how long it takes them to
to polysomnographic measures [e.g. 7–10]).
Several investigators have refined the concept of
conditioned arousal [e.g. 6, 11–16]. Two constructs
in particular have received attention: somatic
to physiologic hyperarousal as assessed by such
measures as heart rate, respiration rate, EMG tone,
Since the mid 1980s, the etiology of insomnia has
been largely understood from within a behavioral
framework. This perspective provides a compelling
conceptualization and the treatments derived from
the theory have demonstrated clinical efficacy
[1, 2]. The behavioral model, as originally put forth
by Spielman and colleagues , posits that insomnia
occurs acutely in relation to both trait and pre-
cipitating factors and occurs chronically in relation
to perpetuating factors. Thus, an individual may be
prone to insomnia due to trait characteristics, but
experiences acute episodes because of precipitating
factors. Acute insomnia, itis hypothesized, becomes
subchronic when it is reinforced by maladaptive
coping strategies. These strategies, in turn, maintain
Correspondence should be addressed to: Michael L. Perlis,
PhD, Sleep Research Laboratory, Department of Psychiatry,
University of Rochester, 300 Crittenden Blvd., Rochester, NY
14642, USA. E-mail: Michael
Web Site: www.urmc.rochester.edu/smd/psych/SRL/
2001 Harcourt Publishers Ltd
M. L. PERLIS ET AL.
and/or galvanic skin response [e.g. 6, 11–16]. Cog-
nitive arousal refers the tendency of patients with
insomnia to be excessively ruminative, particularly
when trying to fall asleep. [e.g. 6, 11–16]. The extent
to which these different forms of arousal contribute
to chronic insomnia continues to be a subject of
investigation. Classic work by Monroe et al.  and
Freedman et al.  provide evidence that patients
with insomnia are somatically hyperaroused prior
to sleep onset and/or during PSG sleep. Recent
work by Bonnet and colleagues [4,5] has confirmed
such findings using a VO2 measure of metabolic
rate. Classic work by Lichstein and colleagues [6,
13] and Mitchell and colleagues  provide evi-
dence that patients with insomnia are prone to
intrusive cognitions at/around sleep onset. Recent
work by Hall and colleagues [15,17] provides evi-
dence that experimentally induced cognitive arousal
is associated with sleep initiation and maintenance
Apart from the classic dichotomy, recent work
on the spectral correlates of insomnia suggests that
patients with insomnia may exhibit a third form of
arousal: CNS arousal as measured by increased high
frequency EEG activity1at/or around sleep onset
and during both NREM and REM sleep [18–21,24].
In the following sections we: 1) review the seven
studies that provide spectral data on the sleep EEG
perspectives on how high frequency EEG activity
may be related to insomnia ; 3) discuss the
precedents and advantages of conceiving of hyper-
arousal as a three component phenomenon; and 4)
suggest possible avenues for future research on
beta/gamma activity in insomnia.
last two decades there has been substantial work
indicating that high frequency EEG activity during
wake in the beta [14–35Hz] and gamma [35–45Hz]
ranges are associated with attention in animals [e.g.
26,27] and with attention, perception, and, more
generally, cognitive function in humans [28–43].
Recently, seven investigations have demonstrated
that patients with insomnia exhibit an unusual
somnographically monitored sleep [18–21,24]. The
in beta activity was apparent in patients with in-
somnia compared with good sleeper controls for
time periods at/around sleep onset and/or during
NREM sleep. These findings were based on: 1) EEG
measurements from central sites using monopolar
orbipolar measuresthatincluded C3,Czor C4,and
O2; 2) relative, as opposed to absolute, measures of
power density; and 3) definitions of beta activity
ranging from 14–32Hz. Three of these studies
provide evidence that increased beta activity during
sleep occurs specifically in association with primary,
as opposed to secondary, insomnia [21,22,24]. Two
studies provide data to suggest that patients with
insomnia also have more beta activity during REM
sleep [18,19]. One study indicates that beta activity
varies with clinical course given non-pharmacologic
interventions . All seven investigations are sum-
In 1986, Freedman  published the first work
on the spectral concomitants of insomnia. In this
landmark study, the EEG power spectral profiles
of 12 patients with sleep onset insomnia were
compared with 12 controls. Patients with insomnia
were found to have increased beta power (high
frequency cutoff=30Hz) peri-sleep onset, during
stage 1 and during REM sleep. No differences in
beta power were observed in stages 2 through
4 of NREM sleep. Freedman concluded that the
observation that patients with insomnia have in-
creased physiological arousal should be extended
to include heightened “cortical arousal”. Three
methodological limitations were apparent: 1) only
1 min of EEG was analyzed for each sleep stage; 2)
the potential confounding influences of EMG activity
were not addressed; and 3) patients were not
screened for psychopathology.
In 1992, Merica and Gaillard  examined the
relative changes in beta/delta power during the
sleep onset period in a sample of 12 patients with
sleep maintenance insomnia and 23 good sleeper
BETA EEG AND INSOMNIA:
High frequency EEG activity in the 20–90Hz range
was first observed in the waking state by Berger in
1937. He hypothesized that such activity, observed
during mental arithmetic, must be a “material con-
comitant of mental processes” [cited in 25]. In the
1We will refer to the upper portion of the EEG spectrum
(14–45Hz) as “high frequency” activity. Although some
investigators prefer to use the term “fast frequency”, our
choice has the advantage of being consistent with the language
that is used to describe analog and digital filters (e.g. high and
low frequency filters).
BETA EEG ACTIVITY AND INSOMNIA
controls. In this study, beta activity was defined as
14.7–30Hz. They found that the rate of change
in beta/delta activity differentiated patients with
insomnia from normal sleepers. Specifically, patients
with insomnia had increased beta EEG and de-
creased delta activity during the sleep onset period
and this difference persisted into the first few
minutes of the first NREM cycle. The authors sug-
gest that abnormal high frequency activity after
sleep onset may explain the frequently observed
tendency of patients with insomnia to overestimate
sleep latency relative to EEG defined measures.
Strengths of this study include: moment-to-moment
spectral data for the 25-min sleep onset window
and an assessment of the reciprocal relationship
between delta and beta activity. Limitations of this
study were: 1) the sample was composed primarily
of patients with sleep maintenance complaints; and
2) that, although the time interval under study was
wake-free, the effect of short duration arousals (less
than 30 s) on the beta activity were not taken into
In 1997, La Marche and Ogilvie  revisited
the observation that patients with insomnia exhibit
increased high frequency EEG activity at/around
sleep onset and extended previous work by eval-
uating whether such activity is specific to primary
insomnia. They compared the EEG power spectral
data from six patients with primary insomnia to
patients with “psychiatric insomnia” (evaluated by
the MMPI) and normal controls. In this study beta
activity was defined as 15–25Hz. They found that
only the patients with primary insomnia dem-
onstrated higher relative beta power during peri-
sleep onset intervals and that patients with insomnia
secondary to psychiatric disorders showed lower
relative beta power overall. This study improves
upon prior work in three ways: the investigators
controlled for periods of wakefulness in their ana-
lyses; all subjects with insomnia had a sleep initiation
complaint and the addition of subjects with sec-
ondary insomnia (i.e. “psychiatric insomnia”). This
allowed for the first assessment of whether beta
activity is specifically associated with primary in-
somnia. Limitations of this study were: 1) the effect
of EMG activity and/or short duration arousals (less
than 30s) on the beta activity was not accounted
for; and 2) “psychiatric insomnia” group was not
In 1998, Merica and Gaillard  undertook a
second investigation of high frequency EEG activity
and insomnia. These investigators further explored
the possibility that spectral characteristics of EEG
sleep may differentiate patients with insomnia from
normal sleepers. Unlike previous work, these in-
at/around sleep onset but also examined the first
four NREM/REM sleep cycles. Beta activity was
defined as 14.7–30Hz. Consistent with previous
work, they found increased amounts of beta EEG
activity in the sleep onset period. Their analysis also
showed, however, that beta activity is elevated over
the course of the whole night, for both NREM and
REM sleep. The primary strengths of this study are
that it is the first to assess the temporal distribution
of beta activity both across the sleep period and
within the NREM and REM episodes, and that it is
the first empirical study on beta activity in insomnia
to place its findings within a theoretical context
(see “Theoretical perspectives” below).
The primary limitation of this study was that
not all episodes of wakefulness (inclusive of short
duration arousals) within each NREM/REM cycle
were removed prior to power spectral analysis.
or REM episodes that contained periods of wake-
fulness lasting 5 min or more. The authors spe-
cifically address this point in their methods section
(p.1830) and argue that wake and arousal data of
less than <5min in duration has negligible effects
on the average standardized NREM and REM power
spectrums. Their argument notwithoutstanding, it
to high frequency EEG activity may actually cor-
or short periods of wakefulness.
Most recently, Nofzinger and colleagues  and
Perlis et al.  presented data at the Association
of Professional Sleep Societies annual national meet-
ings (1999 and 2000) that confirm and extend the
findings of the four aforementioned studies. While
these studies are very recent and must be con-
sidered as preliminary investigations, their data are
included here because they lend confirmatory sup-
port to the published investigations and/or because
they extend what is known about CNS arousal in
Nofzinger et al.  evaluated whole night power
spectral measures of 15 patients with primary in-
somnia, 15 patients with insomnia secondary to
depression and 15 matched healthy controls. Beta
activity was defined as 20–32Hz. Their analyses
M. L. PERLIS ET AL.
showed a trend towards increased NREM beta
activity in patients with primary insomnia compared
to controls and patients with secondary insomnia
(secondary to Major Depressive Disorder). This
Marche and colleagues, it contains two comparison
groups; this provides additional evidence that in-
creased beta activity occurs specifically in as-
sociation with primary insomnia. Second, the entire
night of NREM sleep is profiled in terms of average
NREM sleep; this provides additional evidence that
increased beta activity occurs throughout NREM
sleep. Third, their analysis and data were presented
in 1Hz bins; this allows us to appreciate that the
increase in beta activity appears to occur in a
increasing linear fashion across the upper portion
of the EEG spectrum (14–32Hz). Such a trend
underscores the importance of evaluating whether
this process extends across the entire EEG spec-
trum and into the domain thought to correspond
to EMG activity. Fourth, and most important, the
technique used for averaging the spectral data in-
cluded removing all of the waking epochs from each
NREM/REM interval and the use of a sophisticated
artifact rejection routine (Eric A. Nofzinger, and
Daniel J. Buysse, personal communication); this en-
sures that the increased beta activity that occurs
during NREM sleep in insomnia is not due to
increased wake after sleep onset time or to an
increased occurrence of short duration arousals.
Details regarding the artifact rejection procedure
can be found in work by Brunner and colleagues
In the most recent study by Perlis and colleagues
, patients with primary insomnia (n=9) were
compared to patients with insomnia secondary to
Major Depression (n=9) and good sleeper controls
(n=9) on average NREM measures of beta-
1(14–20Hz), beta-2 (20–35Hz), gamma (35–45Hz)
and omega (45–125Hz) activity. In addition, high
frequency activity was assessed for its association
with subjective–objective discrepancies for sleep
latency and total sleep time (difference scores be-
tween subjective report and PSG measures). It was
found that: 1) the groups differed for average NREM
activity for all four fast frequency bandwidths; 2)
subjects with primary insomnia exhibited sig-
nificantly more beta-1, beta-2 and gamma activity
than subjects with Major Depression (MDD) or
good sleeper controls; and 3) the subjects with
MDD showed the least amount of beta and gamma
and the most amount of omega activity. Cor-
activity was significantly correlated with subjective–
objective discrepancies for total sleep time (beta-1
and beta-2 r=−0.46, P<0.02) and tended to be
significantly associated with subjective–objective
discrepancies for sleep latency (beta-1=0.33,
P<0.09). These data, like those of La Marche et al.
 and Nofzinger et al. , suggest that increased
beta and gamma activity occur uniquely in as-
sociation with primary insomnia. These results also
confirm that increased beta activity occurs through-
out NREM sleep. Finally, the results from this study
are the first to demonstrate that increased beta
activity is associated with sleep state misperception
and that beta/gamma activity is not likely to be the
result of increased wake after sleep onset time or
related to increased motor activity during sleep.
The primary strengths of this study were that
multiple groups were evaluated, a larger portion of
the high frequency spectrum was evaluated and
substantial efforts were made to eliminate wake-
fulness, movement times, mini-arousals (5–15s EEG
One limitation of this study was that first night data
were used for the spectral analyses. Such data
are likely to be confounded by first night effects.
Another limitation is that no attempt was made to
correlated with the high frequency components of
the EEG signal (omega). A low correlation would
have provided additional evidence that the two
domains (beta/gamma versus omega) are relatively
independent of one another.
Apart from the above studies, one investigation
evaluated both between and within group dif-
ferences. Specifically, Jacobs and Benson  con-
ducted a controlled treatment outcome study of
insomnia and utilized pre-sleep beta power (high
frequency cutoff=30Hz) as a major dependent
variable. As with the prior studies, they found at
pre-treatment, that their sample of 14 patients with
primary sleep onset insomnia had higher levels of
beta activity compared to matched controls at/
around sleep onset. Following treatment, the in-
somnia group exhibited significant reductions in
with insomnia maintained higher levels compared
to healthy controls. This study is important because
it is the first demonstration that beta activity in
insomnia varies with clinical state. What remains to
BETA EEG ACTIVITY AND INSOMNIA
be determined is the extent to which the high
frequency EEG activity in patients with insomnia
exhibits both state and trait properties. For ex-
ample, persistent beta activity following effective
treatment may reflect that the insomnia is either
not completely resolved (state) and/or that there
is an underlying persistent vulnerability (trait) that
may make patients with insomnia more likely to
have recurrent episodes.
chronic insomnia, high frequency EEG activity oc-
curs as a result of classical conditioning. That is,
high frequency EEG is elicited in response to the
visual and/or temporal cues usually associated with
sleepiness and sleep (e.g. bedroom, bed, bedtime)
and this occurs in the absence of situational stress-
ors. High frequency EEG activity, in turn, as a
correlate of attention, perception, and cognitive
function in humans, is hypothesized to be related
to increased sensory and information processing
and the attenuation of normal mesograde amnesia
of sleep [45,46].
ON BETA ACTIVITY IN
Enhanced sensory processing and
Enhanced sensory processing at/around sleep onset
and during NREM sleep may be directly associated
with sleep initiation and/or sleep maintenance dif-
ficulties. If the individual is particularly vulnerable
to perturbation by environmental stimuli, he/she
will be more likely to have trouble falling asleep
at natural sleep onset and from awakenings that
naturally occur across the night.
The findings from the above studies are intuitively
appealing, in part because they are consistent with
psychological conceptualizations of insomnia, and in
part because they offer a clearer perspective on
the possible links between presenting symptom,
cognitive process, and underlying neurophysiology.
The beta findings are consistent with the psy-
chological data suggesting that patients with in-
somnia may be hypervigilant and/or excessively
ruminative at sleep onset and/or during sleep [e.g.
13,14]. The beta findings lend themselves to a
clearer perspective regarding the pathophysiology
of insomnia because they point to processes (sens-
memory) and implicate brain structures (e.g. thal-
amus, sensory cortex, prefrontal cortex and the
hippocampus) that may be related to sleep initiation
etical perspectives have been put forward regarding
how high frequency activity at/around sleep onset
and during NREM sleep may be related to insomnia;
one by Perlis and colleagues  and one by Merica
and colleagues .
Perlis and colleagues have proposed , as part
of the Neurocognitive Model of Primary Insomnia,
that beta/gamma activity at/around sleep onset and
during NREM sleep is a primary feature of chronic
insomnia; beta activity is a conditioned form of
CNS arousal that permits a variety of sensory and
cognitive phenomena that do not occur (or are at
least diminished and/or suppressed) in good sleep-
ers. In specific, it is suggested that as insomnia
becomes chronic via perpetuating factors , there
is an increase in high frequency EEG activity at/
around sleep onset and/or during NREM sleep. In
Enhanced information processing and
The perception of environmental stimuli at/around
sleep onset and during NREM sleep may be directly
related to the perception of sleep and wakefulness.
The ready perception of environmental stimuli, par-
ticularly during NREM sleep, may blur the phen-
wakefulness. That is, one cue for “knowing” that
one is asleep may be a lack of awareness for events
occurring during sleep. If one is asleep by traditional
PSG criteria, and yet still capable of processing
information and retaining it in short-term memory,
then when awakened following such an interval, the
subject would be expected to declare wakefulness.
Enhanced information processing may therefore
account for the tendency in insomnia to judge PSG
sleep as wakefulness.
Enhanced long-term memory function
The ability to remember events at/around sleep
onset(s) and during NREM sleep may interfere
with the attribution process that affects morning
judgments about ease of initiation and duration of
M. L. PERLIS ET AL.
Vulnerability or extant physical and/or psychiatric illness
Acute biopsychosocial event
Maladaptive coping strategies
Increased time in bed
Staying awake in bed
Long term memory formation
Complaint of insomnia
Can't fall asleep
Wake up frequently
Perceived wakefulness vs PSG sleep
Overestimation of wakefulness
between the various factors. The schematic of the relationship between somatic and cognitive arousal is
intended to denote a feedback/feedforward loop. Cortical arousal (CNS arousal) is centered between the
somatic and cognitive factors to feature this component as a moderating, if not mediating, variable.
The dual sided arrows in diagram represent the likely possibility that their are reciprocal interactions
sleep. Normally, subjects cannot recall information
from periods immediately prior to sleep [45–48],
during sleep [49–51], or from brief arousals during
information is, to a larger extent, intact in insomnia,
this would be expected to influence judgments
substantially about latency to sleep onset and sleep
A diagram of the Neurocognitive model is pro-
vided below. This diagram was originally published
in 1997  and is included here with permission
from the Senior editor of the Journal of Sleep Re-
search. Please note that the Behavioral model  is
represented in this diagram and that the Neuro-
cognitive Model is an extension of that perspective.
The behavioral model serves to explain how in-
somnia is developed and maintained. Our con-
tribution to the perspective (represented in the
center box) is that behavioral contingencies and
classical conditioning result not only in somatic
but also CNS arousal. The latter phenomena, as
represented by high frequency EEG activity, is hy-
pothesized to result in the occurrence of sensory
and cognitive processes that interfere with sleep
initiation, sleep maintenance, and alter the per-
ception of sleep quantity and quality.
Merica and colleagues  have suggested that
increased beta activity during PSG defined sleep
(NREMand REMsleep) maybe understoodin terms
of neuronal group theory or neuronal population
statistics theory, both of which are embraced by
their neuronal transition probability (NTP) model
. That is, brain activity may be viewed as the
coordination of a number of neuronal groups each
in a specific mode of activity. If a sufficient number
of neurons are in a mode that is consistent with
sleep, the end state is sleep. However, as multiple
groups act in concert to produce the state of sleep,
it is possible that a failure in the co-ordination
process could yield an intermediate state, one in
which the dominant mode is consistent with sleep,
but with neuronal groups still active that are related
to the wake state. Thus, in chronic insomnia “a
larger part of the brain is awake”. The persistence
of beta activity, especially as observed by Merica
and colleagues, fits nicely with this model. Their
data show that the rate of decline in beta activity
over the first NREM sleep cycle is slower and thus
may represent fewer neuronal groups “going off-
line”. Also consistent with their model is that the
NREM spindle peak is wider and that slow wave
activity peak is lower, indicating a slower rate of
transition of the neuronal population to deep sleep.
Their NTP model explained for the first time the
BETA EEG ACTIVITY AND INSOMNIA
model addresses “what beta activity in insomnia is
related to”. “What beta activity in insomnia is”,
according to Merica and colleagues, is an abnormally
populations do not progress normally and/or uni-
formly from high to low frequency modes. “What
beta activity in insomnia is related to” is the con-
tinued activation of systems related to attention,
perception, and memory. The primary difference
between the two models is that Merica and col-
leagues view beta activity at/around sleep onset and
during NREM sleep as, in essence, “a shut down
problem”. Perlis et al., on the other hand, view beta
activity as an elicited response, i.e. a product of
classical conditioning. In the final analysis, it is likely
that both models will possess some explanatory
HYPERAROUSAL AS A THREE
NREM duration (%)
NREM duration (%)
model applied to a comparison of the EEG of normal
and insomniac subjects. Simultaneous time courses
and delta (slow-wave activity) as calculated by NTP
theory. Time is expressed as a percentage of the
NREM episode duration. The ordinate is the ratio of
the number of neurons in a given frequency mode
to the initial number in the beta mode.
The neuronal transitias probability (NTP)
is not a singular construct but rather may be sep-
arated into three relatively distinct components:
EKG or V02 activity) cognitive arousal (as measured
by self report devices such as the Pre-Sleep Arousal
Scale  and/or CNS arousal (e.g. as measured by
beta/gamma EEG activity). Also implied throughout
is that the three systems may be measured in-
dependently and activation within one domain may
occur without a parallel and/or an equal level of
activation within the other domains. Some may view
these premises as controversial ; others might
argue that the pluralistic perspective on arousal is
simply “unnecessarily complicated”. The multiple
factor point of view, however, has several pre-
cedents and advantages.
As for precedents, there are several from both
within and outside of sleep research. Perhaps the
best example outside of our field comes from the
emotion research literature. Emotion researchers
between cognitive and physiologic aspects of emo-
tional arousal. Within these circles, the central
question has been “which form of arousal is prim-
ary” [e.g. 58–60]. From within our own field there
are also several precedents. Perhaps most com-
pelling is that the physiologic and neurophysiologic
qualities of REM sleep itself suggest that there may
be a substantial separation between somatic and
relative evolution during NREM of the number of
neurons (i.e. the power) in each frequency mode.
The application of the NTP theory to the problem
of insomnia is shown in Figure 2. The increase of
Beta, the persistence of spindle activity, and the
decrease of slow wave activity are shown in this
theory to result simply from the lowering of the
probabilities of frequency transition. What remains
to be determined is what neurophysiological mech-
anism could lead to the lowering of these prob-
abilities. One should also note that the NTP theory
is also consistent with the more general ideas
presented by Krueger and Obal  in their neur-
onal group theory of sleep, the difference being
that the “localization” in NTP is functional rather
are related. The NTP model addresses “what beta
activity in insomnia is” while the Neurocognitive
M. L. PERLIS ET AL.
cognitive/CNS arousal: during REM sleep the mind
and brain are active while the peripheral con-
comitants of arousal are largely quiescent. This
alone attests to the possibility that these systems
need not be activated in parallel. More to the poiint,
however, is that several investigators within the
field of insomnia research have suggested, based
on theoretical grounds and/or empirical data, that
hyperarousal in insomnia may be separated into
somatic, cognitive and/or CNS components [11, 13,
16, 56, 61].
As for advantages, there are at least four. First,
allowing for the three domains does not require
that one assume that the association between hy-
perarousal (as a global construct) and insomnia
simply resides in the fact that the two states are
incompatible. Instead, specific processes are im-
plicated. For example, insomnia may not merely be
a problem resulting from “hyperarousal” but the
result of a conditioned form of CNS activity which:
1) in and of itself is incompatible with the de-
activation of cortical processes; and 2) allows for
levels of sensory and/or information processing that
interfere with sleep initiation and/or maintenance.
Second, allowing a differentiation between CNS and
cognitive arousal helps us to build upon established
perspectives regarding the etiology of insomnia. For
example, cognitive arousal in the form of worry
and rumination may give rise to acute insomnia
which has as a feature increased sleep onset and
NREM beta/gamma activity (CNS arousal). In the
chronic course, however, CNS arousal may occur
simply as a conditioned phenomena. Third, allowing
for the concepts of cognitive and CNS arousal may
provide a level of insight into the phenomena of
“sleep state misperception” that is not afforded by
either the global concept of hyperarousal or the
concept of somatic arousal. For example, sleep
state misperception may occur when the normal
mesograde amnesia of sleep is attenuated owing to
CNS arousal. Fourth, and finally, the distinction
between cognitive and somatic arousal has the
heuristic value of being compatible with what
patients report: most patients with insomnia prim-
arily complain of cognitive arousal (i.e. “the inability
to turn their minds off”) . In sum, all of the
above advantages suggest, and the data summarized
in this article support, the utility of construing
“hyperarousal” as a multidimensional construct.
If one is willing to allow for the possibility that
that arousal is not necessarily a singular construct,
there is still the possibility that at least somatic (as
measured by EMG) and CNS arousal (as measured
by high frequency EEG) simply represent two sides
of the same coin. Although there are some data to
support this position,  we believe that the
preponderance of evidence does not support this
position and for essentially two reasons.
Beta/gamma activity is not derived
from, or the same as, EMG activity
• While more extensive artifact rejection pro-
cedures are required than have been undertaken
bymost ofthe studiesof beta/gammain insomnia,
at least two of the seven investigations to date
have sampled the sleep EEG in such a way as to
minimize the influence of overt EMG activity and
yet continued to find significant differences in
beta activity between subjects with and without
• Inter-cranial measures of EEG are able to detect
high frequency EEG. Thus, it is unlikely that
surface measures of beta/gamma activity are ac-
tually EMG in origin .
• High frequency EEG activity, particularly within
the Gamma band, tends to be rhythmic. In con-
trast, signals that are EMG in origin are highly
variable with respect to their frequency and
voltage profiles . Thus, while both kinds are
‘‘fast frequency’’, they are unlikely to be one
and the same given their distinct wave-form
• If somatic and CNS arousal represent the same
phenomena and/or are equally contributory to
the occurrence of insomnia, then one would
expect the two measurement strategies would
equally discriminate between patients with in-
somnia and good sleepers. This is not the case.
Using data from the literature, effect sizes for the
V02 and beta/gamma measures of the differences
between patients with insomnia and good sleeper
controls indicate that the effect size for beta/
gamma is nearly twice that of the V02 measure
(?=1.54 vs ?=0.89).
• Data from our group  using power spectral
analysis of the NREM EEG confirm that patients
(45–125Hz) activity. In fact, patients with in-
somnia show less omega activity than good sleep-
ers. If somatic arousal, as measured by EMG,
BETA EEG ACTIVITY AND INSOMNIA
were a prominent feature of primary insomnia,
one would expect elevations in both the beta/
gamma and omega spectral bands. (The dis-
association between beta/gamma and omega is
also nicely consistent with the tendency of in-
somnia patients to report more cognitive than
• Data from our group  using power spectral
analysis of the NREM EEG confirm that patients
with insomnia secondary to Major Depression
do not show elevated beta/gamma activity. In
fact, patients with MDD appear to exhibit less
beta/gamma activity than either patients with
primary insomnia or good sleepers. In contrast,
patients with MDD show dramatically more
omega activity than either patients with primary
insomnia or good sleepers. This “double dis-
association” supports the point of view that
somatic and CNS arousal may be “separable”
and differently contribute to the primary and
Even in this instance, there is evidence that these
components may be distinct. That is, high frequency
EEG activity is not necessarily isomorphic with
cognitive and/or behavioral activation. For example,
benzodiazepine receptor agonist hypnotics [e.g. 65,
66] and many forms of anesthesia (at least at lower
doses) produce high frequency EEG activity and yet
do not promote, but rather inhibit, cognitive activity
. This then suggests that these more closely
aligned forms of arousal may also occur in-
In the final analysis, the central questions are to
what extent can the three forms of arousal be
independently measured and which of the compo-
nenets of arousal are most associated with primary
insomnia. These remain open empirical questions.
At this time, the clinical implications of beta/
gamma EEG activity in insomnia have not yet
been realized. If it is established that this form
of CNS activity is indeed fundamentally related to
the pathophysiology of insomnia, this “discovery”
may profoundly impact practice. Polysomno-
graphy may become a necessary test for dif-
ferential diagnosis and CNS hyperarousal may
become a specific target for treatment inter-
Beta/gamma activity is reliably
associated with cognitive processes
• There is an extensive neuroscience literature
that has documented the association between
attention, perception and cognition and the oc-
currence of high frequency EEG activity. These
data suggest that beta/gamma activity is strongly
correlated with cognitive processes in a way that
EMG activity is not [28–43].
• Two investigators have found, in both patients
with insomnia and in good sleeper subjects, that
beta/gamma is elevated during REM sleep [18,
19]. This is consistent with the cognitive ac-
tivation that occurs with dreaming. (It is also
unlikely that tonic EMG activity accounts for the
increase in high frequency EEG during REM sleep
given the motor inhibition related atonia that
occurs during this state.) REM sleep Gamma
frequency activity (40Hz) has also been docu-
mented in at least one animal study .
• Data from our group indicate that the perception
of wakefulness during sleep is substantially cor-
related with beta/gamma activity and is not as-
sociated with omega activity .
between somatic and CNS arousal. Less clear, how-
ever, is whether cognitive arousal is separable from
its putative CNS counterpart (high frequency EEG).
The combination of the various studies sum-
marized in this review provide good evidence
that beta/gamma activity at/around sleep onset
and during NREM sleep is increased in patients
with insomnia in comparison to both good
sleeper controls and MDD patients with in-
somnia. There are also some data to support the
possibilities that beta EEG activity:1) is negatively
correlated with perception of sleep  and 2)
varies in association with clinical course .
Cumulatively, these studies suggest that further
research in this arena will be fruitful. Possible
directions might include additional power spec-
tral studies of the sleep EEG in insomnia with
high density EEG brain mapping, brain imaging
vestigations which concomitantly assess somatic,
cognitive and CNS arousal as they occur at sleep
onset and during PSG sleep.
M. L. PERLIS ET AL.
6. Monroe LJ. Psychological and physiological dif-
ferences between good and poor sleepers. J Abnorm
Psychol 1967; 72: 255–264.
7. Bixler E, Kales A, Leo I, Slye E. A comparison of
subjective estimates and objective sleep laboratory
findings in insomnia patients. Sleep Res 1973; 2: 143.
8. Carskadon MA, Dement W, Mitler M, Guilleminault
C, Zarcone V, Spiegel R. Self-report versus sleep
laboratory findings in 122 drug-free subjects with
complaints of chronic insomnia. Am J Psychiatry
1976; 12: 1382–1388.
9. Armitage R, Trivedi M, Hoffmann R, Rush AJ. Re-
lationship between objective and subjective sleep
measures in depressed patients and healthy con-
trols. Depression Anxiety 1997; 5: 97–102.
∗10. Edinger JD, Fins A. The distribution and clinical
significance of sleep time misperceptions among
insomniacs. Sleep 1995; 4: 232–239.
∗11. Freedman RR, Sattler HL, Physiological and psy-
chological factors in sleep-onset insomnia. J Abnorm
Psychol 1982; 91: 380–389.
∗12. Bonnet MH, Arand DL. Hyperarousal and insomnia.
Sleep Med Rev 1997; 1: 97–108.
of cognitive versus somatic determinants of sleep
disturbance. J Abnorm Psychol. 1980; 89: 105–107.
14. Mitchell KR. Behavioral treatment of presleep ten-
sion and intrusive cognitions in patients with severe
predormital insomnia. J Behav Med 1977; 2: 57–69.
15. Hall M, Buysse DJ, Reynolds CF, Kupfer DJ, Baum
A. Stress-related intrusive thoughts disrupt sleep
onset and contiguity. Sleep Res 1996; 25: 163.
∗16. Perlis ML, Giles DE, Mendelson WB, Bootzin RR,
Wyatt JK. Subjective – objective discrepancies in
psychophysiologic insomnia: A neurocognitive per-
spective. J Sleep Res 1997; 6: 179–188.
17. Hall M, Buysse DJ, Dew MA, Prigerson HG, Kupfer
behaviors are associated with sleep disturbances in
bereavement-related depression. Depression Anxiety
1997; 6: 106–112.
18. Freedman R. EEG power in sleep onset insomnia.
Electroencephalograph Clin Neurophysiol 1986; 63:
∗19. Merica H, Blois R, Gaillard JM. Spectral char-
acteristics of sleep EEG in chronic insomnia. Eur J
Neurosci 1998; 10: 1826–1834.
20. Merica H, Gaillard JM. The EEG of the sleep onset
period in insomnia: a discriminant analysis. Physiol
Behav 1992; 52: 199–204.
21. Lamarche CH, Ogilvie RD. Electrophysiological
changes during the sleep onset period of psy-
somniacs, and normal sleepers. Sleep 1997; 20:
Additional power spectral studies would be
useful to confirm that high frequency EEG activity
is limited to beta/gamma frequencies and/or to
assess how high frequency activity is associated
with EEG frequencies that are putatively related
to the homeostatic regulation of sleep (i.e. delta
activity and Process S regulation of sleep [68–70].
High density EEG brain mapping might be used
to more precisely define the cortical distribution
of beta/gamma activity. Brain imaging studies
is CNS in origin and to determine the source
generators of high frequency activity. Chron-
obiologic studies might serve to determine how
beta activity distributes across the 24-h day and/
or varies with non-preferred phase sleep [e.g.
71]. The concomitant assessment of somatic,
cognitive and CNS arousal would allow us to
model the relative contributions of the three
domains to the pathophysiology and phe-
nomenology of insomnia.
UR Salzman Award, NARSAD Foundation Award,
NIH: MH56869-01 Swiss National Science Founda-
tion grant 3100-050765.97.
∗1. Morin CM, Culbert JP, Schwartz SM Non-
analysis of treatment efficacy. Am J Psychiatry 1994;
2. Murtagh DR, Greenwood KM. Identifying effective
psychological treatments for insomnia: a meta-ana-
lysis. J Consult Clin Psychol 1995; 63: 79–89.
∗3. Spielman A, Caruso L, Glovinsky P. A behavioral
perspective on insomnia treatment. Psychiatr Clin
North Am. 1987; 10: 541–553.
4. Bonnet MH, Arand DL. 24-Hour metabolic rate in
insomniacs and matched normal sleepers. Sleep,
1995; 18: 581–588.
5. Bonnet MH, Arand DL. Physiological activation in
patients with sleep state misperception. Psychosom
Med 1997; 59: 533–540.
∗The most important references are denoted by an asterisk.
BETA EEG ACTIVITY AND INSOMNIA
22. Nofzinger EA, Nowell PD, Buysee DJ, Vasco RC,
Thase ME, Frank E, Kupfer DJ, Reynolds CF. To-
wards a Neurobiology of Sleep Disturbance in
Primary Insomnia and Depression: a Comparison
of Subjective, Visually Scored, Period Amplitude,
and Power Spectral Density Sleep Measures. Sleep
1999; 22(Supp. 1): S99.
23. Jacobs GD, Benson H, Friedman R. Home-based
central nervous system assessment of a multifactor
behavioral intervention for chronic sleep-onset in-
somnia. Behavior Ther 1993; 24: 159–174.
∗24. Perlis ML, Smith MT, Orff H, Andrews P, Giles DE.
Beta/gamma activity in patients with insomnia and
in good sleeper controls. Sleep 2001; 24: 110–117.
S, Gallen C. Human auditory evoked gamma-band
magnetic fields. Proceedings of the National Academy
of Sciences of the United States of America 1991; 88:
26. Bouyer JJ, Montaron MF, Rougeul A. Fast fronto-
parietal rhythms during combined focused attentive
behavior and immobility in cat: cortical and thalamic
localizations. Electroencephalograph Clin Neurophysiol
1981; 51: 244–252.
27. Rougeul A, Bouyer JJ, Dedet JJ, Debray O. Fast
somato-parietal rhythms during combined focal at-
tention and immobility in baboon and squirrel mon-
key. Electroencephalogr Clin Neurophysiol 1979; 46:
28. Basar-Eroglu C, Struber D, Schurmann M, Stadler
M, Basar E. Gamma-band responses in the brain: a
short review of psychophysiological correlates and
functional significance. Int J Psychophysiol 1996; 24:
29. Galambos R, Makeig S, Talmachoff PJ. A 40-Hz
auditory potential recorded from the human scalp.
Proc Nat Acad Sci USA 1981; 78: 2643–2647.
30. Goertz R, Jokeit H, Kuchler E. Event related dy-
31. Joliot M, Ribary U, Llinas R. Human oscillatory brain
activity near 40Hz coexists with cognitive temporal
32. Leung LS. Generation of theta and gamma rhythms
in the hippocampus. Neurosci Biobehav Rev 1998;
33. Llinas R, Ribary U. Coherent 40-Hz oscillation
characterizes dream state in humans. Proc N Acad
USA 1993; 90: 2078–2081.
34. Loring DW, Ford M, Sheer D. Laterality of 40Hz
EEG and EMG during cognitive performance. Psy-
chophysiology 1984; 21: 34–38.
35. Lutzenberger W, Pulvermuller F, Birbaumer N.
Words and pseudowords elicit distinct paterns of
30-Hz EEG responses in humans. Neurosci Lett.
1994; 176: 115–118.
36. Lutzenberger W, Pulvermuller F, Elbert T, Bir-
baumer N. Visual stimulation alters local 40-Hz
responses in humans: an EEG-study. Neurosci Lett.
1995; 183: 39–42.
37. Makeig S, Inlow M. Lapses in alertness: Coherence
of fluctuations in performance and EEG spectrum.
Electroencephalogr Clin Neurophysiol 1993; 86: 23–35.
38. Pantev C. Evoked and induced gamma-band activity
of the human cortex. Brain Topograph 1995; 7:
39. Pfurtscheller G, Neuper C, Kalcher J. 40-Hz os-
cillations during motor behavior in man. Neurosci
Lett. 1993; 164: 179–182.
40. Pulvermuller F, Lutzenberger W, Preissl H, Bir-
baumer N. Spectral responses in the gamma-band:
Physiological signs of higher cognitive processes?
Neuroreport. 1995; 6: 2059–2064.
∗41. Sheer D. Focused arousal and 40Hz EEG. In: Knight
R, Bakker D. (eds) The Neuropsychology of Learning
Disorders. Baltimore, MD: University Press, 1976:
42. Spydell J, Sheer D. Effect of problem solving on
right and left hemisphere 40Hz EEG activity. Psycho-
physiology 1982; 19: 420–425.
∗43. Tiitinen H, Sinkkonen J, Reinikainen K, Alho K.
Selective attention enhances the auditory 40-Hz
transient response in humans. Nature 1993; 364:
44. Brunner DP, Vasko RC, Detka CS, Monahan JP,
Reynolds CF3, Kupfer DJ. Muscle artifacts in the
sleep EEG: automated detection and effect on all-
night EEG power spectra. J Sleep Res 1996.
45. Wyatt J, Bootzin R, Anthony J, Bazant S. Sleep
amnesia. Sleep. 1994; 17: 502–511.
46. Wyatt JK, Allen JJBA, Bootzin RR, Anthony JL.
Mesograde amnesia during the sleep onset trans-
ition: Replication and electrophysiological cor-
relates. Sleep. 1997; 20: 512–522.
of excessive daytime sleepiness. In: Drucker-Colin
R, McGaugh J. (eds) Neurobiology of Sleep and Mem-
ory.New York,NY: AcademicPress; 1977:439–456.
48. Portnoff G, Baekeland F, Goodenough DR, Karacan
immediately prior to onset of non-REM sleep. Per-
cept Mot Skills. 1966; 22: 751–758.
49. Wood J, Bootzin R, Kihlstrom J, Schachter D.
Implicit and explicit memory for verbal information
50. Bootzin R, Fleming G, Perlis M, Wyatt J, Schachter
D. Short and long term memory for stimuli pre-
sented during sleep. Sleep Res 1991; 20: 258.
51. Koukkou M, Lehmann D. EEG and memory storage
experiments with humans. Electroencephalograph
Clin Neurophysiol 1968; 25: 455–462.
M. L. PERLIS ET AL. Download full-text
52. Goodenough D, Sapan J, Cohen H. Some ex-
periments concerning the effects of sleep on mem-
ory. Psychophysiology 1971; 8: 749–762.
53. Bonnet M. Memory for events occurring during
arousal from sleep. Psychophysiology 1983; 20: 81–
∗54. Merica H, Fortune RD. A neuronal transition prob-
ability model for the evolution of power in the
sigma and delta frequency bands of sleep EEG.
Physiol Behav. 1997; 62: 585–589.
55. Rao U, Hammen C, Daley SE. Continuity of de-
pression during the transition to adulthood: a 5-
year longitudinal study of young women. J Am Acad
Child Adolesc Psychiatry 1999; 38: 908–915.
56. Nicassio P, Mendlowicz M, Fussell J, Petras L. The
phenomenology of the pre-sleep state: the de-
velopment of the pre-sleep arousal scale. Behav Res
Ther 1985; 23: 263–271.
57. Bonnet MH, Arand DL. The impact of activity upon
spectral EEG parameters. Sleep 2000; 23(Suppl. 2),
58. Zajonc RB. Feeling and Thinking: Preferences need
no inferences. Am Psychologist 1980; 35: 151–175.
59. Zajonc RB. On the primacy of affect. Am Psychologist
1984; 39: 117–123.
60. Schachter S, Singer J. Cognitive, social, and physio-
logical determinants of emotional state. Psychol Rev
1962; 69: 379–399.
∗61. de la Pena A. Toward a psychophysiologic con-
ceptualization of insomnia. In: Williams RL, Karacan
York, NY: John Wiley and Sons, Inc., 1978: 101–143.
62. Gross DW, Gotman J. Correlation of high-fre-
quency oscillations with the sleep-wake cycle and
cognitive activity in humans. Neuroscience 1999; 94:
63. Cacioppo J, Tassinary L, Frilund A. The skeletal
motor system. In: Cacioppo J, Tassinary L. (eds)
Principles of Psychophysiology. Cambridge, MA: Cam-
bridge University Press, 1990.
of fast (30–40Hz) spontaneous cortical rhythms
during brain activation. J Neurosci. 1996; 16: 392–
65. Borbely A, Mattmann P, Loepfe M, Strauch I,
Lehmann D. Effect of benzodiazepine hypnotics on
all night sleep EEG spectra. Hum Neurobiol 1985; 4:
66. Johnson LC, Spinweber CL, Siedel WF, Dement
WC. Sleep spindling and delta changes during
chronic use of a short-acting and long-acting ben-
zodiazepine hypnotic. Electroencephalograph Clin
Neurophysiol 1983; 55: 662–667.
67. Rao U, Poland RE, Lutchmansingh P, Ott GE, Mc-
Cracken JT, Lin KM. Relationship between ethnicity
and sleep patterns in normal controls: implications
for psychopathology and treatment. J Psychiatr Res
1999; 33: 419–426.
68. Borbely AA. A two process model of sleep re-
gulation. Hum Neurobiol. 1992; 1: 195–204.
69. Borbely AA. The two-process model of sleep re-
gulation: Implications for sleep in depression. In:
DJK, THM, JDB. (eds) Biological rhythms and mental
disorders. New York, NY: Guilford Press, 1988:
∗70. Borbely AA. The S-deficiency hypothesis of de-
pression and the two-process model of sleep re-
gulation. Pharmacopsychiatry. 1987; 20: 23–29.
71. Rao U, Lutchmansingh P, Poland RE. Contribution
preliminary report. Neuropsychopharmacology. 2000;