Content uploaded by Richard R Bootzin
Author content
All content in this area was uploaded by Richard R Bootzin
Content may be subject to copyright.
INTRODUCTION
INSOMNIACS’ OVERESTIMATION OF TIME SPENT
AWAKE OVERNIGHT HAS BEEN MAINLY INVESTIGAT-
ED THROUGH RETROSPECTIVE SUBJECTIVE EVALUA-
TION OF SLEEP COMPARED WITH OBJECTIVE
POLYSOMNOGRAPHIC RECORDINGS. Insomniacs typically
underestimate time asleep and overestimate time awake (eg,
1,2
).
Frankel, et al
3
compared overnight polysomnography (PSG) for
five consecutive nights from 18 insomniacs reporting chronic
difficulty initiating and/or maintaining sleep with age- and gen-
der-matched controls. A questionnaire on the night’s sleep was
completed following final awakening each morning. Insomniacs
showed reduced PSG-defined total sleep time (TST) and
increased time awake compared to good sleepers. Subjectively,
insomniacs disproportionately overestimated wake time and
underestimated total sleep time. Good sleepers, however, under-
estimated wake time and overestimated total sleep time.
Morning evaluation of sleep may be influenced by subjects’
expectations and beliefs surrounding their sleep. Insomniacs may
have a bias toward a pessimistic evaluation of their sleep. In
addition, insomniacs may be less able to discriminate between
sleep and wake states. Specifically, insomniacs may have diffi-
culty recognising an awakening from sleep and assume, instead,
they have already been awake. The aim of the present study is to
evaluate sleep-wake discriminability across the night and com-
pare it with retrospective, whole-night reports of sleep.
Borkovec et al
4
investigated this issue by comparing 25
insomniacs and 10 good sleepers on their reporting of prior
wakefulness when questioned during stage 2 sleep. Subjects were
called following the first 5 minutes of continuous Stage 2 sleep
on two consecutive nights and during an intervening afternoon
nap. Insomniacs were more likely than good sleepers to report
having been awake immediately prior to being called on all three
occasions. Furthermore, insomniacs expressed a higher degree of
certainty regarding their decision when presented with a three-
point scale of confidence.
Coates et al
5
incorporated an awakening paradigm in a home
recording of 12 good sleepers and 12 insomniacs. In addition to
being asked about prior sleep-wake state overnight, subjects were
asked to estimate elapsed time and total sleep since they were last
called. As demonstrated by Borkovec, et al
4
, the insomniacs were
more likely than good sleepers to report being awake already
when questioned out of Stage 2 sleep at the beginning of the
night. Subsequent probes during sleep were presented alternately
in REM and Stage 2 sleep. These yielded no difference between
groups in terms of the likelihood of reporting being awake, or
accuracy of time and sleep estimates. Results from probes occur-
ring during spontaneous awakenings, however, indicated that
insomniacs underestimated time asleep and time elapsed since
previously called. Conversely, good sleepers overestimated both
elapsed time and sleep time following spontaneous arousal. In
summary, insomniacs overestimated their difficulty sleeping
early in the sleep period and upon awakening overnight, but their
accuracy at detecting sleep when woken by external stimuli
throughout the night was no different than for good sleepers.
Sewitch
6
used a Signal Detection Theory analysis of respons-
es to sleep-wake probes overnight in order to separate discrim-
inability of sleep-wake status from the bias toward reporting
being already awake when woken from sleep. Over five noncon-
secutive nights in the laboratory, subjects were required to iden-
tify if they were awake or asleep just prior to hearing a bell ring
Insomniacs’ Perception of Wake Instead of Sleep—Mercer et al
SLEEP, Vol. 25, No. 5, 2002
559
Insomniacs’ Perception of Wake Instead of Sleep
INSOMNIA
Jeremy D Mercer BA (Hons),
1,2
Richard R. Bootzin PhD,
3
and Leon C. Lack PhD
1,2
1
School of Psychology, Flinders University, South Australia;
2
Sleep Disorders Unit, Repatriation General Hospital, Daw Park, South Australia;
3
Department of Psychology, University of Arizona
Study objectives: To establish if insomniacs’ underestimation of sleep
time is due to reduced ability to discriminate between sleeping and wak-
ing states.
Design: Two night’s home polysomnography were compared to sleep
diaries. Five laboratory nights employed a series of recorded questions
regarding perception of prior sleep-wake state, which were presented dur-
ing sustained wake and interrupted Stage 2 and REM sleep.
Setting: Sleep laboratory and participants’ homes
Participants: Fourteen insomniacs were compared to 8 good sleepers.
Mean age for both groups was 58 years.
Interventions: N/A
Measurements and Results: A signal detection theory analysis was
applied to participants’ responses to questions presented overnight in the
laboratory concerning judgement of prior sleep-wake state and confidence
in their decision. Insomniacs had reduced sleep-wake discriminability in
addition to a greater bias toward reporting prior wakefulness in the labo-
ratory compared to good sleepers. These measures correlated signifi-
cantly with the degree of underestimation of total sleep and overestima-
tion of wake recorded at home.
Conclusions: Insomniacs’ underestimation of total sleep time is the prod-
uct of prior sleep being misperceived as wake time upon awakening
overnight. This misperception may play a role in the perpetuation of
insomnia.
Key words: Insomnia; sleep-wake perception; sleep-state misperception
Disclosure Statement
Supported by the Australian Research Council and the National Health and
Medical Research Council.
Submitted for publication October 2001
Accepted for publication April 2002
Address correspondence to: Jeremy Mercer, Sleep Disorders Unit, Repatriation
General Hospital, Daws Road, Daw Park, South Australia, 5041;
Tel: +61 8 8275 1149; Fax: +61 8 8277 6890; E-mail: jdm@senet.com.au
and whether they were positive, pretty sure, or not so sure about
their decision. These probes occurred during PSG-defined wake,
and after 60 seconds of Stage 2 or (REM) sleep. A Signal
Detection Theory analysis produced measures of sleep-wake dis-
criminability for each subject derived from responses to the first
question. A similar analysis incorporating both the first and sec-
ond question yielded a measure of bias toward reporting prior
wakefulness. Compared with 11 good sleepers, the one insomni-
ac subject in the study showed both decreased sleep-wake dis-
criminability and increased bias toward reporting being awake.
The present investigation incorporated a Signal Detection
Theory analysis of sleep-wake perception in the laboratory on a
larger sample of insomniacs. These laboratory measures were
then compared with subjective and objective home sleep mea-
sures for both groups, in order to investigate the relationship
between sleep-wake discriminability and whole-night retrospec-
tive sleep misperception of insomniacs.
Compared to good sleepers, we predicted that insomniacs
would demonstrate both reduced sleep-wake discriminability and
increased bias toward reporting prior wakefulness in the labora-
tory. These measures were expected to correlate with the degree
to which participants underestimated sleep time at home.
METHOD
Subjects
Participants were recruited through appeals in the media.
Selection was based on self-reported sleep at home as recorded
using a 7-day sleep-wake diary and confirmed with initial PSG.
All participants were required to be between 18 and 70 years of
age and medically fit. Participants were free from medication for
sleep, heart, blood pressure, thyroid, and psychiatric problems
and were excluded if they reported excessive consumption of
alcohol or caffeine or tobacco. In addition, participants were
excluded if they reported symptoms indicative of sleep apnea,
periodic limb movements in sleep or restless legs syndrome in a
sleep-habits questionnaire. Insomniacs who described a sleep
pattern that may be primarily attributable to poor sleep hygiene
or circadian factors were also excluded. Insomniac participants
reported patterns of sleep and daytime functioning consistent
with a diagnosis of chronic moderate or severe psychophysiolog-
ic insomnia.
7
Based upon responses from 7-day sleep-wake diaries, insom-
niacs (10F, 4M) reported a mean sleep onset latency (SOL) of
greater than 30 minutes and mean total time awake after sleep
onset (WASO) of greater than 40 minutes. Their resulting sleep
efficiency (percentage of time in bed spent asleep) was required
to be less than 80%. Controls (4F, 4M) reported mean SOLof less
than 20 minutes, mean WASO of less than 20 minutes, and a sub-
sequent sleep efficiency of greater than 90%. The mean (sd) age
of insomniacs was 58.14 (6.80) years, and was 58.38 (8.38) years
for controls. While the insomniac participants reported difficulty
both initiating and maintaining sleep, the primary complaint in
this sample is sleep maintenance insomnia, which is typical for
this age group.
The experimental protocol was approved by the Research and
Ethics Committee of the Daw Park Repatriation General
Hospital, South Australia. Informed consent was obtained from
all participants prior to commencing the study. All participants
were offered a gratuity of AU$290 for their participation.
Home Polysomnography
Home recordings were conducted using the Compumedics
(Melbourne, Australia) P series portable PSG recorder. This was
configured to record two (EEG) channels, two (EOG) channels,
one (EMG) channel and light level, which was used to calculate
lights-out time and total time available for sleep. Grass 10mm
gold-plated silver electrodes with connectors specific to this
recorder were used for all placements. Subjective reports of sleep
overnight were gathered by sleep diary.
Laboratory Polysomnography
Laboratory recordings were conducted in a sleep laboratory
consisting of four climate-controlled, sound-attenuated bedrooms
and a central control room. Sleep data was gathered using the
Compumedics S series PSG system, which allowed display of
on-line analysis of EEG frequencies by 30-second epochs in
addition to the recording and display of raw PSG data, from all
bedrooms. The EEG data were sampled at 250 Hz at a gain of 125
µV and low pass filter setting of 30kHz.
Output from a PC sound card could be directed to one of the
four bedrooms through a switching box that connected to a
speaker at the head of each bed. Material presented to partici-
pants through the speakers was recorded and replayed digitally
on a PC using Cool Edit 96 audio software from Syntrillium
Software, Scottsdale, AZ. An independent intercom system
allowed personnel in the control room to monitor verbal reports
from subjects. An infrared camera mounted in the ceiling of each
bedroom and connected to monitors in the control room facilitat-
ed visual monitoring of participant’s position and movement dur-
ing recording. Nicolet 10mm gold-plated silver electrodes were
used for all head placements. Subjects completed a morning
questionnaire immediately following each laboratory session.
PROCEDURE
Home recordings
Two overnight PSG recordings were conducted in the partici-
pants’home in order to compare subjective and objective sleep in
the home setting. Although reduced sleep time due to a first-night
effect has not been reported in home PSG,
8,9
these recordings
occurred with one to six nights of potential recovery sleep inter-
vening.
Electrodes were placed according to the international 10 to 20
system. Two EMG electrodes were placed on the chin. Left and
right EOG electrodes were placed 1cm out and 1cm below the
left and right outer canthus of both respective eyes. Both EOG
electrodes were referenced to a single electrode placed 2 cm
above the nasion. Two pair of EEG electrodes were placed on the
scalp, with the C3-A2 configuration as the primary channel and
a backup pair in the C4 and A1 positions. Following cleaning and
light abrasion of electrode sites, electrode cups were filled with
conductive gel and cemented or taped into position as required.
Impedances between electrode pairs were checked using the
recorder’s on-board impedance meter and were required to be
below 5Kohm before recording commenced.
Recording was started prior to the experimenter leaving the
Insomniacs’ Perception of Wake Instead of Sleep—Mercer et al
SLEEP, Vol. 25, No. 5, 2002
560
participant’s home. Participants were instructed on how to dis-
connect and reconnect a single cable that attached the electrodes
to the main recording unit. This allowed participants to go to bed
at their normal time and get up as required overnight and upon
final awakening in the morning. While the laboratory recordings
employed a series of overnight probes of sleep-wake discrim-
inability, subjective sleep in the home sessions was recorded the
following day using a sleep diary. The sleep-wake diary was
completed as soon as possible following rising in the morning.
An experimenter arrived at the participant’s home at a prear-
ranged time in the morning to remove the electrodes and collect
the sleep diary and recorder.
Laboratory Recordings
Participants slept overnight in the laboratory once a week for
five weeks. They were instructed to arrive approximately one
hour prior to their usual lights-out time, after which they were
prepared for recording in the same manner as for the home stud-
ies. In addition to EEG, EOG, EMG and light measures, partici-
pants were fitted with a Vitalog respiratory effort band and
(ECG) electrodes in the Lead II position. Participants elected
their own lights-out time and wake-up time.
Prior to lights-out, participants were familiarised with the
questions that were to be asked overnight. Experimenters were
careful not to give feedback or otherwise influence participants’
expectations of the likelihood of the questions being delivered
from wake or sleep.
Probes of sleep-wake discriminability occurred throughout the
laboratory sessions. These consisted of a 200 Hz tone that was
delivered through the speaker at the head of the bed and contin-
ued until the participant verbally acknowledged hearing the tone.
A series of four recorded questions was then delivered through
the speaker, and the verbal responses were monitored and record-
ed by the experimenters. The questions were as follows: 1) Just
prior to hearing the tone, were you awake or asleep? 2) How sure
are you of that decision? Positive, pretty sure, not so sure? 3)
How long has it been in minutes since I last called you? 4) How
long in minutes have you been asleep since I last called you?
An initial probe was presented 5 minutes following lights out,
which was excluded from analysis. This provided a reference
point for responding to questions 3 and 4 for the following probe.
Each subsequent probe occurred in one of three conditions: 1)
non-REM: Five minutes following the onset of Stage 2 sleep,
uninterrupted by AASM
10
-criteria arousals. 2) REM: Five min-
utes of uninterrupted REM sleep following the first eye move-
ment. 3) Wake: Following five minutes of continuous wakeful-
ness without movement.
Non-REM and REM probes were alternated in order to pre-
serve sleep architecture and provide an even distribution of both
types of sleep probe across the night. This schedule typically
yielded 12 probes per night for each participant, approximately
four from each of the three conditions. Figure 1 illustrates the
placement of probes on a typical laboratory night.
Polysomnographic Scoring
Sleep data from both home and laboratory recordings were
analysed in 30-second epochs according to Rechtschaffen and
Kales
11
criteria by one of three trained scorers. Five percent of
sleep records were randomly allocated for reanalysis by a second
scorer to ensure interrater concordance remained above 90%.
Signal Detection Analysis
Responses to the first two questions presented overnight were
subjected to a SDT analysis as described by McNichol
12
and used
by Sewitch.
6
Responses to the first question (“Just prior to hear-
ing the tone, were you awake or asleep?”) were used to obtain a
non-parametric measure of discriminability P(A). This measure
reflects the ability to detect a signal (in this case PSG-defined
wake) in the presence of background noise (PSG-defined sleep).
The allocation of wake as the signal is somewhat arbitrary, but as
in Sewitch’s
6
case, the focus of the work is documenting the error
of judging PSG-defined sleep to be wakefulness. Reporting being
awake during PSG-defined sleep is therefore considered a false
alarm according to Table 1.
The proportion of hits and false alarms were used to obtain
P(A) measures in the manner described by McNichol
12
for both
insomniac and good sleeper groups. In SDT analyses, P(A) varies
between 1.0 (perfect discriminability) and 0.5 (chance level, no
discriminability).
Insomniacs’ Perception of Wake Instead of Sleep—Mercer et al
SLEEP, Vol. 25, No. 5, 2002
561
Figure 1—Placement of probes of sleep-wake discriminability on a typical laboratory night. Sleep probes followed 5 minutes of uninterrupted Stage 2 and REM sleep
alternately. Wake probes followed 5 minutes of wake without movement.
RESULTS
Perception of Sleep and Wake: The Signal Detection Theory
Analysis
Table 2 shows the mean hit rate, false alarm rate and P(A) val-
ues of the insomniac and good sleeper groups derived from the
laboratory sessions. Independent samples t-tests indicated that
insomniacs overall had a significantly higher false alarm rate
(probability of reporting being awake just prior to probes from
sleep) than good sleepers. No significant difference between
groups existed for hit rate (probability of reporting being awake
just prior to probes from wake). The non-parametric measure of
discriminability P(A), which is derived from both these mea-
sures, was significantly lower for the Insomniac group.
Response Bias
Responses to the first question in conjunction with the second
yield a six-point scale of confidence in identifying being awake
just prior to hearing the tone. This ranges from a very high degree
of certainty of being awake (“wake, positive”), to a very low
degree of certainty of being awake (“sleep, positive”).
Probabilities of responding in one of the six possible ways to
probes in both PSG-defined sleep and wake are presented in
Table 3. Response probabilities are cumulative starting from
response 1 to response 6, so that category 6 incorporates all pos-
sible responses and, as such, has a probability of 1. Data within
the two groups is combined to yield a representative cumulative
probability distribution for each group.
A receiver operating characteristics (ROC) curve can be plot-
ted using these data for each group as a whole (Figure 2). The
pairs of numbers for the six levels of certainty in the tables above
are plotted as hits (responses from a wake probe) on the y-axis
and false alarms (responses from a sleep probe) on the x-axis.
The area beneath the curves represents the P(A) measure of dis-
criminability and shows the decreased value of P(A) and thus
reduced discriminability for the insomniac group. The points on
the insomniacs’ ROC curve are distributed further to the right of
the graph compared to controls. This is indicative of an increased
response bias toward reporting prior wakefulness. As with
Sewitch’s study, calculation of a non-parametric measure of
response bias is precluded due to the high hit rates in both
groups.
12
However, insomniacs’ cumulative probabilities of
reporting prior wake during sleep (cells 1, 2, and 3 of Table 3) are
more than twice that of Controls, which is a clear indication of
increased response bias.
Differences in false alarm rate alone contributed to the signif-
icantly lower P(A) coefficient for the insomniac group.
Therefore, false alarm rate is in this case an indicator of discrim-
inability, as the probability of accurately detecting the waking
state is the same between groups and displays low variability.
The accuracy in detecting prior sleep upon awakening is the
major focus of the present work and, therefore, a more detailed
analysis of false alarm rates is warranted. False alarm rate data
allows further analysis of between-group differences in the cate-
gories of Sleep stage and time of night where subdivision of the
data precludes P(A) calculations.
Reporting Prior Wake at Probes Occurring in Non-REM and
REM Sleep
Probes conducted during sleep occurred alternately in non-
REM and REM sleep. As the nature of mentation in REM sleep
should be more conducive to detecting the sleeping state, the
probability of reporting already being awake was analysed sepa-
Insomniacs’ Perception of Wake Instead of Sleep—Mercer et al
SLEEP, Vol. 25, No. 5, 2002
562
Table 1—Signal detection theory contingency table: The
sleep-wake state when probed vs subjective judgment of prior
state.
Subject's response
"Awake" "Asleep"
PSG-defined state Wake (Signal) Hit Miss
immediately before Sleep (Noise) False Alarm Correct
hearing the tone Rejection
Table 2—Comparison of insomniacs and controls on the mean
(SD) signal detection theory analysis parameters of hit rate
(probability of accurately identifying prior wake), false alarm rate
(probability of identifying prior sleep as wake) and P(A) (nonpara-
metric measure of sleep-wake discriminability).
Insomniac Good sleeper Group p
difference t
Hit Rate .967(.07) .968 (.06) .043 ns
False Alarm
Rate .727 (.23) .288 (.14) 4.87 <.001
P(A) .745 (.11) .911 (.04) 3.88 <.001
Table 3—Probabilities of responding in one of six ways to probes presented during wake or sleep, cumulated from response 1 to 6.
Insomniacs
Response: 1 2 3 4 5 6
wake, positive wake, wake, sleep, sleep, sleep,
pretty sure not so sure not so sure pretty sure positive
Stimulus: Signal (wake) .918 .975 .975 .984 .992 1.00
Noise (sleep) .378 .592 .700 .748 .929 1.00
Controls
Stimulus: Signal (wake) .852 .975 1.00 1.00 1.00 1.00
Noise (sleep) .140 .222 .284 .385 .634 1.00
rately for the two types of sleep probe. Figure 3 illustrates the
mean False alarm rates for insomniacs and controls when probed
from both non-REM and REM sleep.
False alarm rates for both non-REM and REM sleep probes
were incorporated into a two-way analysis of variance comparing
Insomniacs and controls.
Both insomniacs and controls were more likely to report
already being awake when probed during non-REM sleep com-
pared to REM (F = 34.6, p < .001) There was no interaction
between group (insomniac vs control) and type of sleep probe
(Non-REM vs REM), F = .446, n/s.
False-Alarm Rates in the First Half of Laboratory Nights
Compared to Second Half
A two-way ANOVA compared false alarm rates of both
insomniacs and controls in the first half or second half of labora-
tory nights. The first half of the night was defined as before 3am.
No significant main effect of time of night was observed (first
half vs second half, F = .114, n/s), nor was there any interaction
with group (F = .131, n/s).
Change in False-Alarm Rate Across Laboratory Sessions
To investigate whether sleep-wake discriminability for either
insomniacs or controls improved across laboratory sessions, a
two-way ANOVA compared averaged false alarm rates from
nights 1 and 2 with those from nights 4 and 5.
No main effect of laboratory night was observed (F = .241, p
= .629, n/s), nor was any interaction between Lab Night and
Group observed (F = .382, p = .545, n/s).
Home Studies Before Lab Sessions Vs. After
Six subjects had their home sleep recorded before the labora-
tory sessions, and 16 subjects had home recordings following
completion of the laboratory protocol. A one-way analysis of
variance was conducted to investigate whether the order of data
collection influenced either laboratory or home variables. There
were no significant differences in the major laboratory or home
measures as a result of whether home recordings were conducted
before or after the laboratory sessions.
Sleep Between Probes
Differences between groups in the quality or quantity of PSG
measured sleep between each probe may account for differences
in discrimination of sleep state. Table 4 presents mean time (SD)
spent awake and asleep between probes in the laboratory. Wake
time is expressed as total wake time, SOL, and WASO. Sleep
time is expressed as a total in addition to a breakdown by sleep
stage. There were no group differences in any of the sleep-wake
measures obtained between probes.
Objective Vs. Subjective Total Time and Total Sleep Between
Probes
In order to investigate whether global underestimation of
Insomniacs’ Perception of Wake Instead of Sleep—Mercer et al
SLEEP, Vol. 25, No. 5, 2002
563
0 0.2 0.4 0.6 0.8 1.0
False Alarm Cumlative Probability
1.0
0.8
0.6
0.4
0.2
0
Hit Cumulative Probability
Insomniac ROC curve
Control ROC curve
Figure 2—Receiver operating characteristics (ROC) curves for insomniacs and
Controls. Smoothed curves incorporate a hypothetical zero point and extend to a
cumulative probability of 1.0, which incorporates all possible responses.
Non-REM REM
1
0.8
0.6
0.4
0.2
0
Mean Insomniac
False Alarm Rate
Mean Control
False Alarm Rate
Figure 3—Mean false alarm rates (probability of identifying prior sleep as wake)
for Insomniacs and Controls, for probes occurring in both non-REM and REM
sleep, incorporating SD error bars.
Table 4—Mean (SD) sleep and wake time in minutes between laboratory probes for good sleepers and insomniacs.
Total sleep Total wake SOL Stage 1 Stage 2 Stage 3 Stage 4 REM WASO
Insomniac 46.9 17.4 11.3 3.20 25.9 8.35 2.28 7.09 6.01
(11.7) (8.84) (7.61) (1.28) (8.52) (4.21) (2.66) (2.72) (3.49)
Good sleeper 41.9 15.5 7.49 3.24 20.7 7.62 3.20 7.14 7.97
(5.52) (7.05) (3.55) (0.91) (4.78) (1.93) (2.36) (2.38) (3.95)
t 1.11 .515 1.34 -.069 1.59 .464 -.811 -.049 -1.21
p ns nsnsnsnsnsns ns ns
sleep in insomniacs is a product of the overestimation of elapsed
time or the underestimation of sleep time, objective sleep and
elapsed time measures were compared to subjective measures
gained from questions 3 and 4 at each probe. These questions
were “How long has it been in minutes since I last called you?”
and “How long have you been asleep since I last called you?”
Table 5 presents mean (sd) subjective estimates of total sleep
and total time between probes compared with PSG-derived sleep
and elapsed time measures. Subjective and objective sleep time
was then expressed as a percentage of subjective and objective
total elapsed time for both groups.
Both groups significantly underestimated total time between
probes (Main Effect of Objective vs Subjective Time, F = 9.12, p
= .007) to a comparable extent (Interaction, F = .075). Both
groups also significantly underestimated sleep time (F Objective
vs Subjective Sleep = 37.5, p <. 001). However, Insomniacs dis-
proportionately underestimated sleep time, despite obtaining
slightly more PSG-defined sleep (Interaction F = 8.71, p = .008).
A corresponding significant difference in objective and subjec-
tive % time asleep was observed (F = 17.6, p < .001), with insom-
niacs reporting a disproportionately lower percentage of estimat-
ed time asleep (Interaction F = 22.5, p < .001).
Sleep During Home PSG Recordings
Data collected from the home recordings was averaged over
the two nights for analysis. Table 6 shows that insomniacs’ PSG-
defined sleep varied from controls in only a few respects. The
lower amount of TST and longer SOLfor insomniacs approached
significance. Insomniacs demonstrated significantly less Stage 2
and REM sleep compared to controls, and a greater amount of
WASO. Diary reports of Total Sleep Time, WASO and sleep effi-
ciency were compared with their equivalent PSG-derived values
for both insomniacs and controls using two-way ANOVAs.
Insomniacs underestimated TST, whereas Controls overestimated
the amount of sleep obtained (Interaction F = 9.37, p = .006).
Insomniacs overestimated WASO in contrast to Controls’ under-
estimation (F = 7.51, p = .013). Consequently, Insomniacs dis-
proportionately underestimated Sleep Efficiency (F = 7.52, p =
.013).
Correlation Between Perception of Sleep During Home PSG
and Laboratory Measures of Sleep Discriminability
The SOL, TST and WASO measures averaged over the two
nights of home PSG recordings were compared to sleep-diary
responses completed the following morning. Objective measures
were subtracted from corresponding subjective measures to pro-
duce discrepancy scores for SOL, TST, and WASO. Subjective
and objective sleep efficiency scores were calculated from these
measures for each subject to express the percentage of time in
bed spent asleep. Objective sleep efficiency was subtracted from
subjective sleep efficiency to calculate a sleep efficiency discrep-
ancy score.
Table 7 presents Pearson correlation coefficients of these dis-
crepancy scores and the SDT measures of total hit rate, total false
alarm rate, and P(A) derived from the laboratory sessions.
A decrease in P(A) value and an increase in false alarm rate
both indicate reduced ability to discriminate between sleeping
and waking states in the laboratory. Both of these measures are
significantly correlated with overestimations of WASO and
underestimations of TST and sleep efficiency in the home record-
ings.
DISCUSSION
The present investigation demonstrated that when insomniacs
are awakened from PSG-defined sleep, they are much more like-
ly than good sleepers to report having been already awake. This
effect occurs in awakenings from REM sleep and particularly
from Stage 2 sleep, irrespective of whether they occur early or
late in the sleep period.
The SDT analysis provides measurements of the two con-
structs of discriminability and response bias. An SDT analysis
was applied to participants’judgement of sleep-wake state for the
period just prior to hearing the tones. When compared with good
sleepers, insomniacs were found to have reduced sleep-wake dis-
Insomniacs’ Perception of Wake Instead of Sleep—Mercer et al
SLEEP, Vol. 25, No. 5, 2002
564
Lab Nights 1 & 2 Lab Nights 4 & 5
1
0.8
0.6
0.4
0.2
0
Mean Insomniac
False Alarm Rate
Mean Control
False Alarm Rate
Figure 4—Comparison of mean false alarm rates (with standard deviaton)
across laboratory sessions; First and second nights combined and compared
with fourth and fifth night combined.
Table 5—Mean (SD) objective vs. subjective sleep time and total time in minutes between laboratory probes, with percentage of total
time spent asleep (% Asleep). ANOVAs compare objective PSG measures with subjective estimates for both insomniacs and good
Sleepers.
Sleep Time Total Time % Asleep
Objective Subjective Objective Subjective Objective Subjective
Insomniac 46.9 (11.8) 19.5 (15.1) 64.2 (14.3) 54.4 (14.7) 69.2 (9.94) 33.9 (20.6)
Good Sleeper 41.9 (5.52) 32.4 (12.3) 57.4 (11.1) 45.6 (13.1) 68.1 (7.43) 70.2 (13.5)
F Insomniac vs.
Good sleeper .755, p = .395 2.52, p = .128 11.9, p = .003
F Objective vs Subjective 37.5, p < .001 9.12, p = .007 17.6, p < .001
Interaction 8.71, p = .008 .075, p = .787 22.5, p < .001
criminability as well as increased bias toward reporting being
awake, which is consistent with the findings of Sewitch.
12
The SDT predicts a shift in response bias as a result of expec-
tations of the frequency of signals. If prior wakefulness is expect-
ed, there will be a response bias toward reporting wakefulness.
This bias toward reporting prior wake can be explained by
greater previous experience of nocturnal wakefulness at home.
The present study has shown that this is not simply a retrospec-
tive exaggeration of the insomniacs’ problem but a perceptual
bias that can be demonstrated when people are questioned
overnight.
The inability of insomniacs to discriminate between sleeping
and waking states, which can be calculated independently of a
response bias, is indicative of a sleep-wake perceptual deficit in
poor sleepers. As laboratory sessions occurred in an environment
free of time cues, participants rely upon internal cues to deter-
mine if they were awake or asleep just prior to hearing the tone
presented overnight. It is established that mentation persists into
PSG-defined sleep.
13
Insomniacs’ reduced sleep-wake discrim-
inability may be caused by either a greater amount of mentation
during sleep, mentation that more closely resembles awake men-
tation, or a misattribution of normal nocturnal mentation as
wakeful cognitive activity. Therefore, the qualitative and quanti-
tative differences in sleep mentation between insomniacs and
good sleepers warrants further investigation. Prior sleep was
more accurately identified by both groups when they are probed
during REM sleep. It is likely that the recollection of a dream fol-
lowing REM probes served as a cue for prior sleep. Mentation
during Stage 2 sleep may be less distinguishable from nocturnal
wake mentation and results in reduced discrimination between
these states.
If insomniacs experience higher cortical activation during
sleep (ie, hyperarousal), they would be expected to experience
more mental activity during sleep, increasing the difficulty of dis-
criminating between sleep and wake.
Reduced sleep-wake discriminability observed in insomniacs
may potentially be explained by qualitative differences in their
sleep compared to good sleepers. If insomniacs experienced less
sleep or had significantly lighter sleep between probes than did
controls, they may be expected to demonstrate less discrim-
inability. The schedule of probes presented during sleep, howev-
er, ensured that no difference between good and poor sleepers
was observed in the amount of the various stages of sleep gained
between probes. Differences in sleep quality between probes
therefore do not appear to account for the group differences in
sleep-wake discriminability.
Increased reporting of wake time overnight may be
attributable to the misperception of time awake rather than an
Insomniacs’ Perception of Wake Instead of Sleep—Mercer et al
SLEEP, Vol. 25, No. 5, 2002
565
Table 6—Mean (SD) subjective and objective sleep during home recordings for insomniacs and good sleepers (averaged over two
nights of home recording).
Insomniac Good sleeper F p (1 tailed)
Total PSG sleep 351.2 (76.7) 396.0 (38.8) 2.45 .067
Total diary sleep 246.8 (96.1) 418.1 (58.0) 20.8 <.001*
PSG sleep efficiency 78.8 (12.7) 88.7 (5.4) 4.38 .025*
Diary sleep efficiency 57.1 (22.1) 91.9 (5.6) 18.7 <.001*
SOL 30.4 (27.7) 16.3 (7.6) 1.95 .089
Stage 1 23.1 (10.1) 29.1 (12.5) 1.51 .116
Stage 2 164.7 (56.3) 204.8 (36.0) 3.26 .043*
Stage 3 74.9 (37.6) 60.5 (29.0) .874 .181
Stage 4 20.1 (30.4) 14.2 (21.2) .234 .318
REM 68.4 (26.2) 88.4 (18.8) 3.57 .037*
WASO 61.7 (32.9) 36.4 (26.3) 3.45 .039*
Diary WASO 129.0 (73.9) 20.0 (16.0) 16.6 .001*
Table 7—Correlations (r) between subjective and objective discrepancies in home recordings vs laboratory-derived signal detection
theory measures.
Subjective vs. Objective Discrepancy in Hit Rate False Alarm Rate P(A)
Home Sleep (Probability of accurately (Probability of identifying (Non-Parametric measure of
identifying prior wake) prior sleep as wake) sleep-wake discriminability
incorporating Hit Rate and
False Alarm Rate)
Sleep Onset Latency .084 .350 .011
p = .356 p = .055 p = .482
Total Sleep Time
-.134 -.703 .548
p = .276 p < .001 p = .004
Wake Time After Sleep Onset .256 .827 -.685
p = .126 p < .001 p < .001
Sleep Efficiency
-.206 -.761 .528
p = .179 p < .001 p = .006
underreporting of sleep time. If insomniacs have a tendency to
exaggerate the passage of time overnight, the perceived propor-
tion of time spent asleep would be reduced regardless of sleep
perception. However, insomniacs did underestimate the amount
of sleep between probes compared to good sleepers, while accu-
racy of estimated elapsed time between probes did not vary
between groups. This illustrates that insomniacs’ overestimation
of wake time is the product of perceiving time asleep to be wake
time rather than a distortion of time perception.
These results are in contrast to those found by Coates et al,
5
who concluded that sleep-state misperception in insomnia occurs
only at the initial sleep onset and at spontaneous arousals, not
from experimental awakenings from REM or Stage 2 sleep.
Furthermore, their insomniacs were found to underestimate both
sleep time and elapsed time in association with this mispercep-
tion. The present study employed a slightly different awakening
protocol than did the Coates
5
study and involved five nights
instead of one, which resulted in many more data points and
opportunities to sample sleep-wake perception. Additionally,
potential experimenter feedback to subjects was controlled in the
present study using prerecorded questions and specific instruc-
tions for experimenters to avoid giving feedback on the likeli-
hood of probes occurring in wakefulness or sleep (something
known to effect response bias). These differences in methodolo-
gy may account for the divergent results. Furthermore, the
insomniac population in the Coates
5
study consisted primarily of
sleep- onset insomniacs, with a lower average age than the pre-
sent sample. It is possible that sleep-onset insomniacs will
demonstrate greatest sleep misperception at initial sleep onset,
whereas sleep- maintenance insomniacs’ misperception persists
throughout the night.
There may be some differences between spontaneous awaken-
ings at home, typically, toward the end of the sleep cycle, and our
protocol of forced awakenings in the laboratory. However, prior
sleep should, if anything, be easier to identify if the awakening is
in response to an external stimulus. A protocol could be devel-
oped to exploit spontaneous arousals and compare them to forced
awakenings to resolve this issue.
Insomniacs showed a significantly greater discrepancy
between objective PSG-defined sleep and subjective sleep-diary
reports during home recordings compared with good sleepers.
This objective and subjective discrepancy correlated significant-
ly with laboratory measures of sleep-wake discriminability and
the probability of reporting prior wakefulness when probed dur-
ing sleep. The correlation between laboratory and home-study
measures suggests that insomniacs’ inaccurate retrospective
reports of TST across the night reflects a deficiency in overnight
sleep-wake perception.
Insomniacs perceive that they have not been asleep upon wak-
ing overnight and assume that they have been awake for an
extended period. This increases their expectation of experiencing
wakefulness, reinforcing the bias toward perceiving wakefulness.
Furthermore, subsequent sleep onset may be delayed because of
insomniacs’ negative reaction to what is perceived to be an
extended period of wakefulness overnight. The frustration and
increased activation that accompanies the perceived inability to
sleep serves to compound the problem.
Cognitive therapy may be useful for correcting insomniacs’
bias toward reporting wakefulness. This would consist of provid-
ing information and reassurance that mental activity persists dur-
ing sleep and extended periods of wakefulness overnight are
often interspersed with undetected sleep periods. This therapy
would, of course, be more effective if PSG recordings were avail-
able to support this information.
Sleep-wake discriminability may be improved through per-
ceptual retraining using immediate accurate feedback following
awakening. If insomniacs’discriminability is improved, it should
result in a more accurate assessment of time awake and time
asleep overnight and has the potential to break the proposed cycle
of insomnia perpetuated by sleep-state misperception.
REFERENCES
1. Carskadon MA, Dement WC, Mitler MM, Guilleminault C,
Zarcone VP, Spiegel R. Self-reports versus sleep laboratory findings in
122 drug-free subjects with complaints of chronic insomnia. Am J
Psychiatry 1976;133(12):1382-1388.
2. Edinger JD, Fins AI. The distribution and clinical significance of
sleep time misperceptions among insomniacs. Sleep 1995;18(4):232-
239.
3. Frankel BL, Coursey RD, Buchbinder R, Snyder F. Recorded and
reported sleep in chronic primary insomnia. Arch Gen Psychiatry
1976;33:615-623.
4. Borkovec TD, Lane TW. VanOot PH. Phenomenology of sleep
among insomniacs and good sleepers: Wakefulness experience when
cortically asleep. J Abn Psychol 1981;90(6):607-609.
5. Coates TJ, Killen JD, Silverman S, George J, Marchini E, Hamilton
S, Thoresan CE. Cognitive activity, sleep disturbance, and stage specif-
ic differences between recorded and reported sleep. Psychophysiol
1987;20(3):243-250.
6. Sewitch DE. The perceptual uncertainty of having slept: The inabil-
ity to discriminate electroencephalographic sleep from wakefulness.
Psychophysiol 1984;21:243-259.
7. American Academy of Sleep Medicine International classification
of sleep disorders, Revised. Rochester: Minnesota: American Academy
of Sleep Medicine, 2000.
8. Drewes AM, Nielsen KD, Taagholt SJ, Svensen L, Bjerregard K,
Nielsson L, Kristensen L. Ambulatory polysomnography using a new
programmable amplifier system with on-line digitization of data: techni-
cal and clinical findings. Sleep 1996;19(4):347-54.
9. Edinger JD, Fins AI, Sullivan RJ, JR, Marsh GR, Dailey DS, Hope
TV, Young M, Shaw E, Carlson D, Vasilas D. Sleep in the laboratory and
sleep at home: comparisons of older insomniacs and normal sleepers.
Sleep 1997;20(12):1119-26.
10. American Sleep Disorders Association Atlas Task Force. EEG
arousals: scoring rules and examples. Sleep 1992;15:174-184.
11. Rechschaffen A, Kales A. A manual of standardised terminology,
techniques and scoring system for sleep stages of human sleep subjects.
Washington DC: US Government printing office, 1968.
12. McNicol D. A primer of signal detection theory. London: George
Allen & Unwin: 1972.
13. Foulkes D. The psychology of sleep. New York: Scribners, 1966.
Insomniacs’ Perception of Wake Instead of Sleep—Mercer et al
SLEEP, Vol. 25, No. 5, 2002
566
