Sleep, 14(5):448-453
© 1991 Association of Professional Sleep Societies
Enforced 24-Hour Recovery Following
Sleep I)eprivation
Leon Rosenthal, Lori Merlotti, Timothy A. Roehrs and Thomas Roth
Sleep Disorders & Research Center, Henry Ford Hospital, Detroit, Michigan, U.S.A.
Summary: The pattern of recovery sleep after slee:p deprivation was investigated in healthy young adults. Six
subjects experience:d three experimental conditions (0, 24, and 48 hr sleep deprivation) in a Latin Square design.
The recovery period consisted of a 24-hr enforced time in bed during which subjects were polysomnographically
recorded beginning at 0800. To assess the differentiall effects of the deprivation conditions, the total sleep time on
the 24-hr recordings was submitted to a six (4-hr block) by three (deprivation condition) multivariate analysis of
variance. Subjects slept more following the 24- and 48-hr conditions when compared to the O-hr condition. Across
conditions, subjects slept more during the first 4 hr when compared to the remaining five blocks. Importantly, there
was a significant interaction of sleep deprivation by 4-hr block. In block I sleep was differentially recovered between
each condition with more sleep being recorded following longer hours of deprivation. In block 2 subjects in the 24and 48-hr conditions slept comparable amounts and significantly more than those in the O-hr condition. In blocks
3 and 4 only the 48-hr condition exhibited significantly more sleep than the O-hr condition. However, significantly
less sleep was found in block 6 following the 48-hr condition. Overall, subjects recovered 72% and 42% of the total
amount of sleep lost during the 24- and 48-hr conditions, respectively. Key Words: Sleep-Sleep deprivationSleep duration-Sleep stages.
Research on recovery sleep in humans following deprivation (l,2) has either scheduled an extended time
in bed of 8-12 hr, or instructed subjects to rise only
after they have met their sleep quota (ad lib sleep).
Using these strategies, it is found that the loss of one
night of sleep results in lengthening a subsequent night's
sleep by approximately 10-20% (3,4). Thus, researchers have concluded that only a small fraction of the
total sleep time (TST) lost during deprivation is recovered (5)_
The composition of the recovery sleep has also bt~en
characterized. A selective enhancement of slow-wave
nonrapid eye movement (NREM) sleep (stage 3~) has
been documented following a night without sleep (6).
Selective slow-wave sleep recovery has not only been
found in human subjects but has been a consistl:!nt
finding across other species subjected to comparable
periods of wakefulness (7). When sleep deprivation is
prolonged to more than a single night, the recov1:!ry
sleep consists of enhanced stage 3~ NREM sleep and
REM sleep (8). For example, in a classical study by
Kales et al. (9) where four subjects were deprived for
Accepted for publication April 1991.
Address correspondence and reprint requests to Dr. Leon Rosenthal, HFH Sleep Disorders Center, 2921 West Grand Boulevard,
Detroit, Michigan 48202, U.S.A.
205 hr of sleep and then recorded for three consecutive
nights (12 hr the first night and 9 hr the subsequent
nights) subjects regained 79% of stage 4 NREM sleep,
33% of stage REM sleep, but only 9% of TST (l0).
However, it is important to note that the sleep efficiency on each of the three recovery nights was 100%_
The review of these findings raises important methodological considerations. It is possible that both strategies, ad lib sleep and slightly extended time in bed,
terminate the recovery period prematurely. The latter
underestimates recovery by scheduling a shorter than
required time in bed and ad lib sleep by confounding
the drive to sleep with the individual's motivation to
get out of bed_ Although some subjects might elect to
stay in bed, others might opt to curtail their sleep before
recovery sleep is completed_ An additional consideration in the ad lib methodology is the possible influence
of circadian factors. This would be especially important if motivation interacted with circadian rhythms.
Thus, allowing subjects to terminate their sleep period
may result in an equivocal assessment of recovery sleep.
It is hypothesized that these strategies have resulted
in a systematic underestimation of the duration of recovery sleep. In order to further determine the extent
of sleep recovery a 24-hr enforced time in bed was
utilized. Recovery sleep after 0, 24 and 48 hr sleep
448
ENFORCED 24-HOUR RECOVERY FOLLOWING SLEEP DEPRIVATION
deprivation was assessed utilizing a 24-hr enforced
bedtime.
,
2200
METHODS
I
,
2200
I
1
I
I
I
0800
2200
J
,
I
!
I
I
I
!
0800
o hr
0800
2JOO
0800
CONDITION
I
0800
0800
0800
0800
0000
24 hr CONDITION
Subjects
The subjects were six male normal sleeping (as defined below) volunteers, aged 20-30 yr. All subjects
had a history free of drug abuse, medical or psychiatric
disorders. The subjects signed an informed consent and
were paid for their participation.
!
I
449
,
2200
0800
,
I
!
I
,
,
I
2200
!
I
,
,
J
I
,
,
,
!
t
I
I
0800
,
,
48 hr CONDITION
FIG. 1. Schedule for each of the three experimental conditions.
The black bars indicate the time in bed periods.
Procedure
All subjects passed medical and sleep laboratory
screening. They were required to have a normal physical examination and a drug-free urine toxicology screen
prior to entering the study. The sleep screening consisted of having subjects spend 8 hr in bed (2330-0730
hr) while being polysomnographically monitored using
electrodes attached at standard placements to record
electroencephalogram (EEG), electrooculogram (EOG),
and electromyogram (EMG). Subjects also wore a nasal/oral thermistor to monitor air flow and EMG leg
electrodes to monitor leg movements. They all had a
screening polysomnogram with a sleep efficiency of at
least 88%, no evidence of periodic leg movements during sleep, sleep disordered breathing or indication of
other pathologies during sleep. For all the subsequent
sleep recordings electrodes were attached to record
EEG, EOG, and EMG only. The day following the
screening polysomnogram, subjects were tested for level of sleepiness using the standard multiple sleep latency testing (MSLT) procedures (11). Subjects were
required to have a mean sleep latency on the MSLT
of > 8 min in order to qualify for the study.
All subjects were exposed to the three experimental
conditions over a 4-wk period with at least 5 days
between conditions. Experimental conditions were assigned in a Latin Square design. Prior to each experimental condition subjects reported to the laboratory
at 2100 hr and were required to stay in bed for 10 hr
(2200-0800 hr) to minimize the possibility of a sleep
debt prior to each deprivation condition. The following
day, they were tested for level of sleepiness using the
standard MSLT procedure. An average sleep latency
on the MSLT of > 8 min was required to assure comparable levels of sleepiness at the beginning of each
experimental condition. Then, depending on the deprivation condition subjects were allowed a normal
night of sleep (2300-0700) or were kept awake for 24
or 48 hr. The recovery period started at 0800 and lasted
24 hr in each of the three conditions (Fig. 1). No caffeinated beverages or excessive exercise were allowed
during the deprivation period and subjects were monitored to assure compliance with the experimental protocol.
During the entire 24-hr recovery period, subjects
were required to remain in bed in a darkened bedroom
with provisions for toileting and meals made according
to the following protocol. Food was offered at three
specified times and only if subjects were awake (07000900, 1200-1430, and 1800-2030). Subjects had to
awaken spontaneously and stay awake for 10 min before food was offered. If subjects were asleep the meal
was delayed until the next specified meal period. The
food offered at each meal consisted of a sandwich, fruit
and a noncaffeinated beverage. Thirty minutes were
allowed for the consumption of food with the exception
of the third period, which was extended to 45 min to
allow for the replacement of electrodes. Subjects remained in their rooms during the consumption of their
meal, but were allowed to get out of bed, sit at a table
and use the rest room. Subjects remained in bed with
lights off and were polysomnographically recorded the
rest of the time. Subjects were not allowed watches or
alarm clocks in their rooms and their interaction with
laboratory personnel was limited to the meal periods.
An additional 15-min bathroom break, between 2300
and 0100 was provided. At other times use of the
bathroom was allowed only if the subject had been
awake for 10 min or longer. Polysomnographic recordings were scored in 30-sec epochs according to the
standards of Rechtschaffen and Kales (12) by scorers
blind to the experimental condition. The interrater reliability was maintained at 90% or higher.
Data were analyzed using the general linear model
analysis procedure (SAS Institute) with one withingroup variable being sleep deprivation condition and
the other within-group variable being hour block (six
4-hr blocks). The probabilities reported are corrected
by the Greenhouse-Geisser method. Post hoc contrasts
were done comparing the three conditions and the six
4-hr blocks.
Sleep, Vol. 14, No.5, 1991
450
L. ROSENTHAL ET AL.
TABLE 1. Mean (standard deviation) of total sleep time for each 4-hr block of the recovery periodfollowing 0,24 and 48
hr of sleep deprivation (sip dep)
Block (timer
Sip dep
1 (0800-1159)
2 (1200-1559)
3 (1600-1959)
4 (2000-2359)
5 (2400-0359)
6 (0400-0759)
Total
o hr
195 (33)
229 (7)*
237(3)t
41 (44)
181 (62)*
224(21)*
43 (54)
101 (61)
146(81)*
21 (45)
110 (103)
124(64)*
145 (90)
116 (80)
131(83)
208 (34)
204 (39)
131(77)*
653 (103)
941 (180)
992 (109)
24 hr
48hr
a
Post-hoc comparisons of significant interaction (p < 0.05): * vs. O-hr condition; t vs. both 0- and 24-hr conditions.
RESULTS
Baseline sleep-wake function
The total sleep times (TSTs) on the nocturnall0-hr
recordings, prior to each experimental condition, were
comparable (564 min ± 34 for the O-hr condition' 545
min ± 40 for the 24-hr condition; and 544 min ±
34.06 for the 48-hr condition). The mean sleep latency
on the MSLT the following day was also comparable
among the three experimental conditions (13 min ±
6 for the O-hr condition, 13 ± 4 for the 24-hr condition,
and 12 ± 4 for the 48-hr condition).
Recovery sleep time
To evaluate the pattern of recovery sleep the 24-hr
recovery period was divided into six (4-hr) blocks. Thus,
the blocks represent clock hours 0800-1159 (block 1),
1200-1559 (block 2), 1600-1959 (block 3),2000-2359
(block 4),2400-0359 (block 5) and 0400-0759 (block
6). To assess the differential effects of the sleep deprivation conditions, the total sleep time on the 24-hr
recordings were submitted to a six (4-hrblock) by three
(deprivation condition) multivariate analysis of variance (MANOYA). The results of this analysis showed
TABLE 2. Mean (standard deviation) of sleep stages (minutes and percentages) during the recovery sleep period following 0, 24 and 48 hr of sleep deprivation
Condition
Minutesa .b
Percent'
Stage 1
0
24
48
21 (5)
14 (5}i
13 (3)i
Stage 2
0
24
48
136 (48)
137 (66)
133 (37)
330 (64)
Sleep
Stage 3/4
0
24
48
REM stage
495 (105)d
548 (72)d
51 (40)
112 (66)d
123 (56)d
0
135 (29)
24
197 (55)
48
188 (51)
a Significant main effect for stage 2 (p < 0.01).
b Significant main effect for stage 3/4 (p < 0.01).
, Significant main effect for stage 1 (p < 0.05).
d Post-hoc comparisons (p < 0.05): vs. O-hr condition.
Sleep, Vol. 14, No.5, 1991
51 (8)
53 (6)
55 (4)
8 (6)
13 (8)
13 (6}
21 (3}
21 (2}
19 (4)
a significant main effect of sleep deprivation (df 2, 10;
F 17.41; P < 0.01). Post-hoc comparisons showed that
subjects slept significantly more following the 24- and
48-hr sleep deprivation conditions (941 ± 180 min
and 992 ± 109 min) when compared to the O-hr condition (653 ± 103 min). There was no statistically
significant difference between the 24- and 48-hr conditions (see Table 1). A main effect of hour block was
also found (df 5,25; F = 7.53; p < 0.01). Post-hoc
comparisons showed that subjects slept significantly
more in block 1 when compared to all other blocks.
Also, subjects slept more in blocks 2 and 6 when compared to blocks 3 and 4. Importantly, there was a significant interaction of sleep deprivation by hour block
(df 10,50; F = 5.34; P < 0.01). The post-hoc comparisons showed that recovery sleep in blocks 1, 2, 3, 4
and 6 differed as a function of deprivation condition.
In block 1 the TST for each experimental condition
differed significantly (195 ± 33 min following the O-hr
condition, 229 ± 7 following the 24-hr condition and
237 ± 3 following the 48-hr condition). In block 2 the
TST in the O-hr condition (41 ± 44 min) was significantly lower than the 24- and 48-hr conditions, whereas the 24- and 48-hr conditions were comparable (181
± 62 min and 224 ± 21 min, respectively). The pattern
of recovery sleep was similar for blocks 3 and 4 with
TST in the O-hr condition being significantly less (43
± 54 min and 21 ± 45 min) than after 48 hr of deprivation (146 ± 81 and 124 ± 64 min). TST after 24
hr deprivation was intermediate and not different from
the other two conditions (101 ± 61 and 110 ± 103
min). In the last block (hours 21-24) there was a reversal of this pattern. TST in the O-hr condition was
the highest (208 ± 34 min) whereas TST in the 48-hr
condition was the lowest (131 ± 77 min). The 24-hr
condition was not significantly different from the
0- and 48-hr conditions (TST of 204 ± 39 min).
Recovery sleep staging
Although the primary purpose of this experiment
was to study the amount of recovery sleep following
deprivation, analyses of sleep stages was also carried
out. However, not all sleep stages were present in each
4-hr block. Thus a one-factor (sleep deprivation con-
451
ENFORCED 24-HOUR RECOVERY FOLLOWING SLEEP DEPRIVATION
TABLE 3. Sleep stage percent (mean and standard deviation) for each 4-hr block of the recovery period following 0, 24,
and 48 hr of sleep deprivation (sip dep)
Block (time)a
<-
Sip dep
1 (0800-1159)
2 (1200-1559)
3 (1600-1959)
4 (2000-2359)
5 (2400-0359)
6 (0400-0759)
Stage 1
0
24
48
20 (5)
8 (2)
5 (4)
30 (38)
22 (13)
8 (3)
55 (41)
23 (II)
18 (9)
20 (40)
7 (6)
35 (25)
13 (7)
14 (8)
30 (35)
20 (5)
15 (9)
28 (24)
Stage 2
0
24
48
51 (10)
45 (12)
41 (14)
38 (38)
54 (13)
66 (4)
17 (21)
54 (8)*
59 (7)*
9 (23)
43 (34)
49 (23)*
48 (28)
45 (25)
48 (27)
52 (10)
52 (8)
53 (16)
Stage 3/4
0
24
48
2 (3)
30 (17)*
38 (15)t
5 (10)
4 (7)
4 (5)
9 (14)
1 (2)
1 (.1)
1 (3)
1 (3)
2 (5)
11 (15)
10 (14)
9 (8)
6 (10)
11 (14)
6 (13)
REM stage
0
24
48
26 (7)
17 (4)*
16 (9)*
11 (16)
20 (11)
22 (2)
3 (6)
22 (12)*
23 (10)*
3 (8)
16 (13)
14 (14)
11 (8)
14(13)
13 (10)
22 (6)
22 (8)
13(11)
Sleep stage
a Post-hoc comparison of significant interaction (p < 0.05): *
VS.
dition) MANOVA was carried out for each stage of
sleep. First, the total number of minutes spent in each
stage during the 24-hr recovery was submitted to this
analysis. There was no significant main effect on the
total number of minutes spent in stage 1 NREM sleep
(136 ± 48, 137 ± 66 and 133 ± 37 for the 0-,24- and
48-hr conditions, respectively) and REM stage sleep
(135 ± 29, 197 ± 55 and 188 ± 51 for the 0-, 24- and
48-hr conditions, respectively). However, there was a
main effect for stage 2 NREM sleep (df2,10; F= 14.18;
p < 0.01). Post-hoc comparisons showed that the total
number of minutes spent in stage 2 was significantly
lower following 0 hr of sleep deprivation (330 ± 64)
than after 24 and 48 hrofdeprivation (495 ± 105 and
548 ± 72, respectively). No statistically significant difference was found between the 24- and 48-hr conditions. There was also a main effect of condition for
stage 3/4 NREM sleep (df2,1O; F= 15.82; p < 0.01).
Post-hoc comparisons showed less stage 3/4 NREM
following the O-hr condition (51 ± 40) when compared
to the 24- and 48-hr conditions (112 ± 66 and 123 ±
56, respectively). Here again, the recovery of stage 3/4
NREM sleep was not different between the 24- and
48-hr conditions (see Table 2).
The recovery of sleep stages was also analyzed as
percentages of TST during the 24-hr recovery period.
The percent of each sleep stage was derived and submitted to a one-factor (sleep deprivation condition)
MANOVA. The results only showed a main effect of
condition for stage 1 NREM sleep percent (df2,10; F
= 8.44; P < 0.05). The post-hoc comparisons showed
a higher percent of stage 1 NREM sleep after 0 hr of
deprivation (21 ± 5) when compared to the 24- and
48-hr conditions (14 ± 5 and 13 ± 3, respectively).
No significant difference was found between the 24and 48-hr conditions.
Finally, the recovery pattern of the different sleep
O-hr condition;
t vs. both 0- and 24-hr conditions.
stages was analyzed. The data was submitted to a six
(4-hr block) by three (deprivation condition) MANOVA. The reader is reminded that not all sleep stages
were present in each 4-hr block and thus, the results
of this analysis should be interpreted with caution.
Table 3 shows the main sleep stage percentages for
each 4-hr block across the three sleep deprivation conditions. Significant interactions of sleep deprivation by
4-hr block occurred for stage 2, stage 3/4 and REM
stage sleep [df 10,50; F = 3.36 (p < 0.05), F = 9.12
(p < O.01)andF= 2.51 (p < 0.05), respectively]. Sleep
deprivation was associated with significant increases
in percent stage 3/4 NREM and REM in block 1, percent stage 2 NREM and REM in block 3 and percent
stage 2 NREM in block 4.
Relation of sleep recovered to sleep lost
In order to estimate the amount of TST recovered
vis a vis the amount lost during each of the deprivation
conditions a percentage of recovery sleep was derived.
The percentage was derived by subtracting the amount
of total sleep after the 24- or 48-hr condition from the
amount subjects had in the O-hr condition and then
dividing this number by the TST on the nocturnal eight
hours prior to the O-hr condition (x = 402 ± 33 min)
(10). For the 48-hr condition the denominator was
multiplied by 2 as subjects remained awake for two
nights. The TST recovered was 72% following the 24hr sleep deprivation condition and 42% following the
48-hr condition. Similar estimates of the recovery of
specific sleep stages were derived. As it can be observed
in Table 4, there was an almost completely recovery
of stage 3/4 (98%) after the 24-hr condition, which was
drastically reduced to 59% after 48 hr of deprivation.
A similar pattern was observed for stage 2 NREM and
REM sleep (see Table 4).
Sleep, Vol. 14, No.5, 1991
452
L. ROSENTHAL ET AL.
TABLE 4. Amount of recovery sleep, overall and by stage, expressed as a percent of lost sleep
Sleep stages (min)
O-hr condition
24-hr condition
48-hr condition
2
3/4
REM
Total
52
61
112
98
122
124
59
134
78
197
81
402
941
156
187
34
804
992
42
136
332
Amount lost
Recovery sleep
% Recovery
54
138
4
210
495
78
Amount lost
Recovery sleep
% Recovery
108
133
-3
420
549
52
DISCUSSION
653
72
vation, is intriguing. It is possible that subjects satiated
their homeostatic sleep process by recovering more
slow-wave sleep as it has been suggested previously
(13,14). Alternatively, it could be argued that only a
finite amount of sleep can be recovered in a 24-hr
period. Extension of the scheduled time in bed, beyond
24 hr, would be necessary to evaluate this possibility.
This hypothesis could also explain the lack of differences in TST following the 24- and 48-hr sleep deprivation conditions.
The results of this experiment show that a greater
amount of sleep is recovered following sleep dep:rivation than previously reported. This was the result
of utilizing an enforced 24-hr time in bed and thus
allowing for an unbiased measurement of the sleep
recovery process. By minimizing the impact of possible
competing motivators the extent of sleep recovery was
allowed to be expressed. To maximize this respons(~ a
concerted effort was made to keep all other possible
Acknowledgement: This work was supported by NIMH
behavioral motivators neutral. For example, meals were
provided at times when subjects were awake so hunger grant MH-00860, awarded to Dr. Leon Rosenthal.
would not interfere with the sleep recovery process.
Relevant to this methodology is the lack of complaints
REFERENCES
by the participants who had no difficulty completing
1. Gulevich G, Dement W, Johnson L. Psychiatric and EEG obthe protocol.
servations on a case of prolonged (264h) wakefulness. Arch Gen
The pattern of sleep during the 24 hr showed that
Psychiatry 1966;15:29-35.
,
much of the recovery took place during the first 8 hr 2. Akerstedt T, Gi1Jberg M. A dose-response study of sleep loss
and spontaneous sleep termination. Psychophysiology 1986;23:
of the recording. However, the enforced time in bed
293-7.
methodology showed statistically significant differ- 3. Webb WB, Agnew HW. The effects on subsequent sleep of an
ences in recovery sleep beyond this period. In fact,
acute restriction of sleep length. Psychophysiology 1975; 12:36770.
significant recovery continued through the fourth 4-hr
4. Benoit 0, Foret J, Bouard G, Merle B, Landau J, Marc ME.
block. Such a prolonged recovery would not have been
Habitual sleep length and patterns of recovery sleep after 24
detected by previous methodologies evaluating recovhour and 36 hour sleep deprivation. Electroencephalogr Clin
NeurophysioI1980;50:477-85.
ery sleep.
5. Horne J. Waking awareness, subsequent sleep and cerebral 'resThe results highlight the importance of methodtitution.' In: Horne J. Why we sleep. New York, Oxford Uniological considerations when discussing the extent of
versity Press, 1988:142-79.
sleep recovery. This process is inherently limited by 6. Dement W, Greenberg S. Changes in total amount of stage four
sleep as a function of partial sleep deprivation. Electroencephathe time in bed which is utilized in any given study.
logr Clin NeurophysioI1966;20:523-6.
In view of the present results previously held beliefs 7. Borbely AA, Neuhaus HU. Sleep deprivation: effects on sleep
and EEG in the rat. J Comp PhysioI1979;133:71-87.
about the extent of sleep recovery need to be reas8. Mullaney DJ, Johnson LC, Naitoh P, Friedman JK, Globus
sessed. Future research should consider the possible
GG. Sleep during and after gradual sleep reduction. Psychobias introduced by utilizing ad lib sleep in a variety of
physiology 1977;14:237-44.
experimental protocols. As to the 24-hr enforced time 9. Kales A, Tjiaw-Ling T, Kollar EJ, Naitoh P, Preston TA,
Malmstrom EJ. Sleep patterns following 205 hours of sleep dein bed methodology, future research is needed to assess
privation. Psychosom Med 1970;32:189-200.
the effects of initiating the recovery sleep period at 10. Horne J. Sleep deprivation. In: Horne J. Why we sleep. New
York, Oxford University Press, 1988:13-78.
different times of the circadian rhythm.
II. Carskadon MA, Dement W, Mitler MM, Roth T, Westbrook
An unexpected finding in this experiment was the
PR, Keenan S. Guidelines for the multiple sleep latency test
(MSLT): a standard measure of sleepiness. Sleep 1986;9:519actual reversal in the pattern of TST among the three
24.
experimental conditions during the sixth 4-hr block.
12. Rechtscaffen A, Kales A. A manual ofstandardized terminology,
The significantly decreased amount of sleep recorded
techniques and scoring system for sleep stages ofhuman subjects.
in the last 4-hr block, following 48 hr of sleep depriWashington, DC: U.S. Government Printing Office, 1968.
Sleep. Vol. 14. No.5. 1991
ENFORCED 24-HOUR RECOVERY FOLLOWING SLEEP DEPRIVATION
13. Brunner DP, Dijk DJ, Tobler!, Borbely A. Effect of partial sleep
deprivation on sleep stages and EEG power spectra: evidence
for non-REM and REM sleep homeostasis. Electroencephalogr
Clin NeurophysioI1990;75:492-9.
453
14. Akerstedt T, Gillberg M. Sleep duration and the power spectral
density of the EEG. Electroencephalogr Clin Neurophysiol
1986;64: 119-22.
Sleep, Vol. 14, No.5, 1991