Sleep, 2(4):391-407
© 1980 Raven Press, New York
Timing of REM and Stages 3 + 4 Sleep
During Temporal Isolation in Man
Elliot D. Weitzman, Charles A. Czeisler, Janet C. Zimmerman, and
Joseph M. Ronda
Laboratory of Human Chronophysioiogy, Department of Neurology, Montefiore
Hospital and Medical Center, Albert Einstein College of Medicine, Bronx, New York
Summary: During nonentrained sleep-wake conditions in man, healthy adult
subjects spontaneously develop "long" biological days (> 35 hr) in addition to
the normal, approximately 25 hr day. The ratio of sleep to total time remains
constant (approximately 0.30), with long sleep episodes occurring approximately 180 out of phase with the short sleep episodes. The timing and amount
of REM sleep advance to an earlier time within the sleep episode during freerunning, whereas stage 3 + 4 sleep is related to the initiation and course of the
sleep process itself. The REM - NREM cycle length does not change, comparing entrained and nonentrained conditions. The study of the chronophysiology
of humans under nonentrained conditions may serve as a model of the
chronopathology of sleep-wake changes which occur in sleep disorders associated with depression, narcolepsy-cataplexy, sleep-wake dyssomnias,
delayed sleep phase insomnia, and aging. Key Words: Circadian-REMStages 3 + 4-Human sleep-Biological rhythms-Chronophysiology.
0
An important finding in biological rhythm research is that when organisms are
not entrained by 24 hr zeitgebers (time cues), they develop daily cycles which have
periods greater or less than 24 hr. Extensive research in animals utilizing a
rest-activity measurement has demonstrated that these "free-running" period
lengths are species-specific and genetically influenced (Pittendrigh, 1961; Pittendrigh and Daan, 1974; Daan and Pittendrigh, 1976; Pittendrigh and Daan, 1976a;
Pittendrigh and Daan, 1976b). When the animal is maintained in constant conditions, the new cycle lengths can be remarkably constant for months to several
years. Studies in normal man under isolated conditions have also demonstrated
similar free-running rhythms with period lengths in almost all instances greater
than 24 hr (Aschoff and Wever: 1962, 1976; Mills 1966, 1974; Aschoff, 1967, 1969,
1970; Mills et aI., 1974, 1976; Aschoff et aI., 1976). The "rest" segment has been
Accepted for publication April 1980.
Address reprint requests to Dr. Weitzman at Laboratory of Human Chronophysiology, Department of Neurology, Montefiore Hospital and Medical Center, 111 East 210th Street, Bronx, New
York 10467.
391
392
E. D. WEITZMAN ET AL.
assumed to be a sleep episode in previous studies, measurements made by either
"lights-out" or "bedrest" time. However, no systematic, long-term laboratory
studies of the temporal complexity of sleep states and waking have been reported,
although there have been polysomnographic recordings of free-running sleep
episodes in short laboratory studies (Webb and Agnew, 1974) and for intermittent
segments in underground cave studies (Chouvet et aI., 1974). Analysis of the
previously reported detailed sequential durations of the rest (self-selected lightsout) time indicates that a significant day-to-day variability is present, suggesting
that internal sleep stage amounts and timing may be related to such variability. In
addition, the assumption that rest is sleep cannot be made, since wake periods
prior to sleep onset, and during the lights-out episode, clearly occur, presumably altering the biological rhythm properties and perhaps influencing other correlative
measured periodic events such as body temperature and hormonal cycles.
Accordingly, we have made detailed and prolonged polygraphic measurements
of sleep stage characteristics of a group of normal adult men living in an environment free of time cures for prolonged periods (15-183 calendar days). These
results are part of a comprehensive multi variable study ofthe chronophysiology of
man living in a time-free environment with a nonscheduled daily pattern of living
(Czeisler, 1978; Weitzman et aI., 1979a).
A seminal event in the understanding of sleep was the recognition in the 1950s
that REM sleep represented approximately 20% of the total night's sleep, occurred in 90-100 min cycles and was associated with vivid dreaming and a wide
variety of autonomic activity changes (Dement, 1973). In normal adult subjects
habitually sleeping at night, a very consistent finding was the REM latency from
sleep onset of 80 to 100 min. Indeed, this reliable normal pattern was used to
identify specific pathological conditions of the sleep, namely, narcolepsy-cataplexy (Rechtschaffen et aI., 1963; Dement et aI., 1966; Richardson et aI., 1978) and
endogenous depression (Kupfer, 1976).
The development of a body of knowledge of the organization of the circadian
rest-activity rhythm in normal man during prolonged temporal isolation has provided a unique experimental approach to define the endogenous and sleep-related
timing systems of sleep stages without the confounding paradigm of sleep deprivation in most previous studies. We report here the results obtained regarding the
timing and organization of REM and stages 3 + 4 sleep in normal subjects during
entrained and free-running conditions.
METHODS
A special environment was established where the individual subjects lived for
many weeks (Czeisler, 1978; Weitzman, et aI., 1979a). A three-room apartment
(study, bedroom, and bathroom) was arranged without windows; the walls were
sound attenuated, and a double-door entrance separated the temporal isolation
facility. A closed-circuit TV system and voice intercom monitored the subject's
activities.
Ten male subjects were individually studied. The first group (3 SUbjects: AA,
AB, and AC) was studied for 15 calendar days, and the second group (6 subjects:
Sleep, Vol. 2, No.4, 1980
TIMING OF REM AND STAGES 3 + 4 SLEEP IN MAN
393
AD, AE, AF, AG, AI, and AJ) for 25 calendar days. A single subject (CA) was
studied for an extended stay of 105 calendar days. No subject had significant
psychopathology or medical illness; none was on drugs of any kind. Each subject
kept a written daily diary of sleep times for at least 2 weeks and maintained a
regularly scheduled bedrest-activity cycle in accord with their usual habits. After
entry in the temporal isolation facility, an entrained condition of three or four
scheduled 24 hr bedrest-activity cycles preceded the nonscheduled free-running
portion of the study. The entrained clock times were determined by the subject's
recorded habitual lights-off-lights-on times at home. The subject was told that
his sleep time would be scheduled for certain portions of the study but was not
advised of the clock times nor the duration. Following the entrained portion, each
subject was told that he could choose to go to sleep and awaken at any time he
wished. He was not allowed to "nap." A decision to go to sleep, therefore,
represented the sleep episode for that biologic "day." Food was available to the
subject on demand as breakfast, lunch, dinner, and a "snack." The subject could
request any meal type at any time. A set of buttons were available which when
pushed was coded on a paper punch tape and indicated the behavior the subject
was about to initiate and the elapsed time (to the nearest minute) from the beginning of the study. These behaviors included meal and type, sleep time, awake
time, urination, take shower, defecation, blood sample, and exercise. The paper
punch tape structured the entire time series of each study.
The subject was totally isolated from contact with all nonlaboratory persons but
communicated by intercom and direct discussion with selecte'd laboratory staff.
The supervising staff members were scheduled on a random basis as to time of day
and duration of work shift to prevent the subject from obtaining a time cue.
The following measurements were made for each subject.
1. Polygraph sleep recording. The interval between the subject's decision to
sleep and lights-out with full electrode application was less than 15 min. All
polygraphic records were scored by standard methods (Rechtschaffen and Kales,
1968).
2. Rectal temperature. A rectal thermistor probe was maintained by each subject
throughout the entire study except for brief daily periods of defecation. The temperature was automatically recorded every minute on the punch paper tape and a
printout.
3. Plasma cortisol and growth hormones. During this study a chronic catheter
inserted into the arm vein of the subjects was used to obtain plasma at approximately 20 min intervals for subsequent assay for cortisol and growth hormone.
4. Polygraphic data scoring. All scored data was transferred to a computercompatible format and analyzed for total sleep, lights-out, and all sleep stages for
each lights out- sleep episode. The pattern of sleep stage sequences was visualized by a special display program. A quantitative determination was made for a
set time period of the percentage of each sleep stage and waking. The results of
that analysis was also displayed utilizing a computer plotting technique.
5. Special mathematical techniques and computer algorithms. In addition to the
usual statistical method of analysis and computer plotting and display routines,
several mathematical techniques were created to assist in the analysis of the data
Sleep. Vol. 2, No.4, 1980
394
E. D. WEITZMAN ET AL.
(Czeisler, 1978; Weitzmann et aI., 1979a). These include (a) estimate of period
length using a minimum variance fit, (b) wave-form eduction, and (c) averaged
timed event relationship.
RESULTS
Sleep- Wake Cycle Pattern
Each of the 10 subjects developed a free-running (FR) sleep-wake cycle following the entrained (EN) base-line condition. In each case the mean period
length was longer than 24 hr. The subject population was divided into two types
(excluding the tenth subject,CA). In type A (6 subjects: AA, AB, AE, AF, AG,
and AI), the period lengths during free-running averaged between 24.4 and 26.2 hr,
whereas type B (3 subjects: AC, AD, and AJ) had consistently long periods
greater than 37 hr. The bedrest episode for the type B subjects ranged from 8 to 20
hr, with an average of 14 hr. Linear regression analysis through mid-bedrest
demonstrated a very stable period length (r 2 > 0.99 for each). Short bedrest
episodes recurred at a regular phase of the circadian cycle with a period slightly
longer than 24 hr. The long bedrest episodes began at a phase angle approximately
180 shifted from that of the short sleep episodes. Variations in sleep lengths were
related to the phase of the ongoing circadian oscillation at which the sleep episode
occurred. In addition, when prior wakefulness lasted more than 1,440 min, there
was a clear increase in sleep length with sleep episode duration lasting 600-1,200
min (Czeisler, 1978; Weitzman et aI., 1979a).
Subject CA lived under free-running conditions for 80 calendar days and demonstrated several important features (Table 1). He maintained a regular freerunning period length of approximately 25 hr for the first 30 bedrest-activity
cycles. He then developed a bedrest-activity pattern consisting of alternating
long cycles (>36 hr) with a series of shorter cycles (approximately 25 hr). This
alternating pattern persisted until it was interrupted by a special light-dark entrainment protocol on calendar day 84. The sleep episodes continued on an approximately 25 hr period length in spite of the interruption by very long noncircadian periods. These approximately 25 hr self-selected sleep-wake times were
therefore entrained to an internal periodic process, which can be considered an
"internal zeitgeber." The long sleep episodes (>600 min) occurred at a phase
angle approximately 180 shifted from the short sleep episodes but in parallel with
0
0
TABLE 1. Sleep stage characteristics jOr CA during entrained, freerunning, and re-entrained experimental conditions
Condition
n
Total sleep
time
(min)
Entrained
Free-running
Re-entrained
4
67
14
426
478
449
REM sleep
Stages 3 + 4
min
%TST
min
%TST
100
100
91
23.6
20.9
20.2
115
132
157
26.9
27.7
35.0
TST, total sleep time.
Values shown represent the means of the number of days indicated.
Sleep. Vol. 2. No.4. 1980
TIMING OF REM AND STAGES 3 + 4 SLEEP IN MAN
395
the same period-length midsleep regression line. Analysis of the relationship between length of bedrest episodes and length of prior activity demonstrated that for
only 7 out of 16 activity episodes lasting greater than 20 hr, did the subsequent
bedrest episodes exceed 12 hr in length. However, as was the case for the other
subjects, no long bedrest episode was preceded by an activity episode less than 20
hr in length (Czeisler, 1978; Weitzmann et aI., 1979a).
There was a rapid phase delay of lights-out and sleep onset of at least 6 hr within
48 hr of the onset of the free-running condition for 8 of the 9 subjects. The other
subject (AG) delayed his sleep onset by 5 hr on the third biological day. In
addition, all 6 subjects in group A had a characteristic "scalloped" appearance of
the timing of bedrest onset, with a variable cycle of 3-4 days. This could not be
explained as a "transient" process related to the onset of free-running, since it
clearly continued throughout the free-running condition in 4 subjects (AB, AE,
AG, and AI).
The bedrest-dark episode in general corresponded with the sleep episode for
each subject and for each night. However, it was found at times that there was a
short delay from lights-out-bedrest onset to sleep onset. The 2 older subjects (AI,
50 and AJ, 51 years old) consistently interrupted their sleep episodes by awakening for short sequences during the bedrest episode and remained awake in the dark
for sequences up to 1 hr after awakening and prior to activity onset. These waking
interruptions were also present during the entrained segment. These findings emphasize the importance of defining sleep stages polygraphically when measurements of biological rhythm variables are made.
Sleep Stages
There was considerable variability in the mean total sleep time (TST) during
free-running, with 2 subjects averaging 13.8 and 12.8 hr (AC, AD), whereas the
others were approximately 8 hr. In spite of this variability in TST per sleep
episode during free-running, the ratio of sleep time to period length only varied
between 0.24 and 0.35 across subjects, with an average of 0.29. This compared
with 0.30 during the entrained condition. When the entrained ratio was compared
to the free-running ratio for each subject, it was noted that two subjects with high
entrained ratios (long sleepers) (0.31 and 0.32) increased the value to 0.35 and
0.34, respectively, during free-running, whereas 4 subjects with the lowest entrained ratios (0.27, 0.28, 0.29, and 0.29) (shorter sleepers) all decreased the ratio
to 0.25,0.25,0.24, and 0.25, respectively, during free-running. The 3 other subjects
with intermediary entrained values had little change during free-running.
The sleep stage characteristics for all subjects were compared as a function of
sequential experimental bedrests during the three experimental conditions (entrained, free-running and re-entrained). The values of REM percent of TST were
remarkably constant throughout and did not differ significantly as a function of
experimental conditions. The stages 3 + 4 percent of TST did increase to a small
extent from the entrained (27.8%) to the free-running (29.8%) condition, especially
during the last six free-running bedrests. A small average increment occurred
during the five re-entrainment bedrests (32.2%) for 5 subjects.
Sleep. Vol. 2. No.4. 1980
396
E. D. WEITZMAN ET AL.
An interesting result was obtained when comparisons were made for REM
percent of TST by subject and by experimental condition. There was considerable
variability in REM sleep across subjects (range, 15-30%) during the entrained
segment. However, the intrasubject variability was very small as a function of
experimental conditions. This was not the case for stages 3 + 4, since both the
inter- and intrasubject variability was similar in all three experimental conditions.
These results indicate that each subject maintained an individual control of REM
percent of sleep time which was independent of the entrained or free-running state.
This does not appear to be the case for stages 3 + 4 sleep.
Three subjects (AC, AD, and AJ) consistently had long sleep episodes associated with long sleep-wake cycle lengths. There were a total of 27 sleep
episodes lasting 10 hr or longer. These long sleep episodes differed from the short
sleep episodes. As reported elsewhere, the timing of the onset of these long sleep
episodes occurred at a different phase of the subjects' circadian temperature
rhythm (130°-270°, 0° = midtrough) than the onset of the short sleep episodes
(270°-120°) (Czeisler, 1978; Weitzman et aI., 1979a). In addition, during the
long sleep episodes, sustained stages 3 + 4 sleep would characteristically
occur between 12 and 18 hr after sleep onset. However, the first 4 hr of the long
sleep episodes did not differ significantly in regard to the characteristic timing and
amount of stages 3 + 4 sleep seen under entrained conditions. Thus, despite
normal amounts of stages 3 + 4 sleep present at the onset of these long sleep
episodes, stages 3 + 4 would reappear after 12-16 hr of sustained sleep. Although
occasional awake episodes interrupted these long sleep times (especially for subject AJ), they were not sufficiently long to explain the reoccurrence of stages 3 + 4.
Sleep Stage Organization for Sleep Episodes
The mean percentage of time of REM sleep per sequential hour of sleep during
the entrained condition demonstrated a difference when comparing the two subject groups (A and B). There was more REM during hours 2-8 of sleep for group
B than for group A (Fig. 1). The sequential hourly pattern of the mean percentage
of time of stages 3 + 4 sleep also differed during the entrained conditions (Fig. 2).
There was more stage 3 + 4 initially for group A (60% in first hour) than for group
B (42% in 1st hour). The decrease was then progressive for group A (almost linear
to the 6th hour). In group B, the stage 3 + 4 percent initially increased, reaching a
peak at the 3rd hour (51%) and then showing a sharp fall at the 4th hour of the
sleep episode with a small percentage rise again at the 5th and 6th hours after
bedrest onset. The mean percentage of time awake in the entrained condition also
differed during the 1st hour after bedrest onset, with group A showing significantly
less (15%) than group B (28%). There was a tendency for more wakefulness to
occur later in the sleep episode (specifically hours 4, 5, and 7) for group B.
During the free-running as compared to the entrainment condition for groups A
and B, there were major changes in the sequential hourly timing of REM sleep
during the course of the subjects' sleep episodes (Tables 2 and 3; Figs. 1 and 3).
After the first hour of the sleep episode there was a sharp increment in the amount
of stage REM during free-running for group A, reaching a mean of 30% on the
Sleep, Vol. 2, No.4, 1980
TIMING OF REM AND STAGES 3 + 4 SLEEP IN MAN
~
w
397
Entrained
60
Free running
II:
W
l!)
~
50
••
••
••
••
Ul
~
w
40
~
••
••
j::::
u..
0
30
ui
+1
,
20
I-
Z
w
<..l
II:
•
,,
ui
"'t~l--r
--'
10
W
a.
z
«
w
2
~
6
4
,,
\J
'
10
8
12
HOUR OF SLEEP EPISODE
FIG. 1. Change in the mean (±SE) percentage of sleep time in REM for each sequential hour of the
bedrest episode for subject group A during entrained and free-running segments (see text). Asterisks
denote statistical significance (t-test): *p < 0.05; **p < 0.02; ***p < 0.01; ****p < 0.001.
second hour and then gradually falling to a value of 15% on the 8th hour after
lights-out. (During entrainment, the 8th hour had a mean value of 40%). Indeed,
the curve of REM sleep percent for the sleep episode is a falling one during
free-running compared to a rising one during entrainment. During free-running,. a
progressive decrement of the percentage of stages 3 + 4 was clearly present for
groups A and B, similar to that found in the entrained condition (Figs. 2 and 4).
70
Entrained
Free running
<t
+
'"
w
60
C!>
«
l-
(/)
~
50
w
::0
...i=
40
FIG. 2. Change in the mean (±SE)
percentage of sleep time in stages 3 + 4
for each sequential hour of the bedrest
episode for subject group A during the
entrained and free-running segments.
0
W
vi
+,
30
IZ
W
u
a:
w
20
a.
z
«
10
w
::0
•
2
4
6
8
10
12
HOUR OF SLEEP EPISODE
Sleep, Vol. 2, No.4, 1980
398
E. D. WEITZMAN ET AL.
TABLE 2. Comparison of REM latency and amount (in 1st 180 min
after sleep onset) for the entrained and free-running
condition for subjects AA -AJ
Entrained
n
REM latency
REM amount (in 1st 180
min after sleep onset)
Sleep length
" p
b
Free-running
min
86.9 ± 34.5
16.4 ± 9.76
41
41
41
min
n
107
107
62.9 ± 20.81"
30.0 ± 15.42"
107
558 ± 216.5 b
475 ± 52.31
< 0.001.
P < 0.01.
Means ± SD are given.
TABLE 3. REM latency and amounts during entrained
and free-running for subject CA
Entrained
REM latency
REM amounts (in
(1st 180 min)
Sleep length
Free-running
n
min
n
min
5
88.0 ± 20.66
16.0 ± 3.54
68
68
48.0 ± 34.33"
30.6 ± 15.14"
5
445 ± 3.7
4
528.3 ± 135
68
< 0.025 «(-test).
Means ± SD are given.
" p
~
Entrained
a::
Free running
w
w 70
(!)
~ 60
en
z
50
w
i= 40
u.
o
o 30
~
--r~ __ -J----1J:1-' \.
," "
~
a::
w
a..
___ J_
10
z
«
w
~
2
4
6
,
8
10
12
/j
If
,'..
,
-- - oII'f-'---"
20
z
~
•
•
**
~
t
- --
,""
14
16
..
18
• •
20
22
HOUR OF SLEEP EPISODE
FIG. 3. Change in the mean (±SE) percentage of sleep time in REM for each sequential hour of the
bedrest episode for subject group B during the entrained and free-running segments.
Sleep. Vol. 2. No.4. 1980
TIMING OF REM AND STAGES 3 + 4 SLEEP IN MAN
<t
399
Entrained
+
rt)
LU
Free running
70
(!)
<l:
I-
CJ)
60
z
50
LU
::::E
i= 40
u.
0
,,
30
0
LU
en
+,
20
I-
- ---
Z
LU
u
10
0::
LU
0..
Z
<l:
2
4
LU
::::E
6
•
-- -r---l
l
/}
--!--'-- -~'{<\--r"
• - ',+- - -- 10
8
12
14
16
18
•
20
22
HOUR OF SLEEP EPISODE
FIG. 4. Change in the mean (±SE) percentage of sleep time in stages 3 + 4 for each sequential hour of
the bedrest episodes for subject group B during the entrained and free-running segments.
The 2nd hour after bedrest onset, however, did not differ between the two conditions (entrained, 45%; free-running, 33%) for group A. This difference at the 2nd
hour was presumably related to the major increase of REM percent during freerunning which occurred at this time. There was also a shift of the waking time to
the last third of the sleep episode for group A. The subjects showed only 9% of
time awake in the 1st hour during free-running compared to 16% when entrained.
However, during the 7th through 9th hours after bedrest onset, the percentage of
time awake increased dramatically (i.e., 7th hour-4% EN, 26% FR; 8th
hour-ll% EN, 37% FR; 9th hour-21% EN, 43% FR).
Another characteristic difference between the long and short sleep episodes was
the timing and amount of REM sleep within the first 3 hr after sleep onset (Tables
4 and 5). All of the sleep episodes which had a very small REM latency «20 min)
were short sleep episodes during the free-running conditions. The mean REM
TABLE 4. Comparison of REM amount in 1st 180 min after sleep
onset and sleep length for sleep episodes with more than 30 min
and less than 30 min of REM sleep (in 1st 180 min
from sleep onset) for subjects AA -AJ
REM <30 min
Sleep length
REM amount
(in 1st 180 min of sleep)
REM >30 min
n
min
n
min
50
50
492 ± 176"
43.5 ± 10.4
57
57
607 ± 234
18.3 ± 7.4
" p < 0.001 (t-test).
Means ±. SD are given.
Sleep, Vol. 2, No.4, 1980
E. D. WEITZMAN ET AL.
400
TABLE 5. Comparison of REM amounts (in 1st 180 min
after sleep onset) and REM latency during free-running
for sleep periods greater than 600 min and less than 600
min in length for subjects AA -Ai
Sleep period >600 min
n
REM amount
REM latency
27
27
min
IS.7 ± 11.6"
73.0 ± 21
Sleep period <600 min
n
min
SO
SO
33.S ± 14.7
59.6 ± 20
" p < 0.001 (I-test).
Means ± SD are given.
latency (sleep onset to onset of first REM episode) clearly decreased for 9 of the
10 subjects (AI was the exception) comparing entrained to the free-running condition. A partial recovery took place during the re-entrainment conditions. In addition, the mean total minutes of sleep in the first 3 hr of sleep increased for 8 of the
9 subjects (subject AI excepted) between entrainment and free-running. However,
during re-entrainment these values did not return to base line. The timing and
amount of REM sleep during the first 3 hr after sleep onset in subjects AA - AJ and
CA were determined during the free-running condition when they had the alternating long and short sleep-wake cycles (Tables 3-5). All the bedrests with a short
REM latency « 10 min) and the bedrest with more than 30 min of REM sleep in
the first 3 hr except for one occurred within 90° of the nadir (0°) of the circadian
temperature rhythm. In addition, for 12 REM onsets which occurred within 10
min of sleep onset, 11 occurred within 60° of a specific phase (midtrough) of the
circadian temperature rhythm (Czeisler et aI., 1980).
In order to compare the sequential hourly pattern of sleep stages and waking of
short «600 min) compared to long (>600 min) sleep for all subjects (AA - AJ, CA)
during free-running, we determined the mean percentage of time in stages REM, 3
+ 4 sleep, and waking for each sequential hour.
The graphs of mean percentage of stages 3 + 4 were remarkably similar for the
long and short sleep episodes (Fig. 5). The 2nd, 3rd and 4th hours demonstrated a
small decrease for the short sleep episodes, reflecting the increase at those times
of REM sleep (see below) for the short sleep episodes. Of considerable interest is
the finding that there was an increase in stages 3 + 4 sleep after 14 hr of sleep, for
the long nights, on the 14th, 17th, 18th, and 20th hr. The values rose to 7,6,9, and
21 %, respectively, for the few sleep episodes which lasted that long.
There was also a major difference in the REM percent comparing long and short
sleep episodes (Fig. 6). REM sleep was higher by approximately twofold for hours
2-4 for the short sleep episodes. However, for the 8th through 10th hours, the
percentage of REM was higher for the long sleep episodes. Indeed, the usual
polarity of REM sleep during the daily sleep episode was absent for the >600 min
group, with the REM percent remaining at approximately 20% from the 2nd
through 18th hours. On the 19th and 22nd hours, recorded in 4 and 1 sleep
episodes, the REM percent rose to 45 and 48%, respectively.
Sleep, Vol. 2, No.4, 1980
TIMING OF REM AND STAGES 3 + 4 SLEEP IN MAN
v
•
'"w
\!)
;:!
70
Greater than 600 mins.
Less than 600 mins.
60
*
If)
~
401
50
w
::E
l-
40
I>.
0
W
30
If)
+1
IZ
20
•
W
u
II:
10
•
W
•
0-
Z
<t
W
::E
2
4
6
6
10
14
12
16
16
20
22
HOUR OF SLEEP EPISODE
FIG. 5. Change in the mean (:;:SE) percentage of sleep time of stages 3 + 4 for all subjects during
free-running for each sequential hour for all bedrest episodes shorter than and longer than 600 min.
The percentage of time awake was not different for the long sleep episodes for
the first 6 hr but, of course, remained small until the 11th hour. After that there
was progressive increment of mean waking time, largely reflecting the variability
in sleep length that lead to a smaller number of sleep episodes with which to derive
the average.
::E
w
Greater than 600 mins.
Less than 600 mins.
II:
W
60
\!)
;:!
If)
50
~
~
lI>.
0
30
•
•
w 40
**
**
***
'r' 'l--:·'~~-f---T"
**
-1r'\ ,- /" f. . '1
--,,)..
' .J J.
,',,, t
'
W
iii 20
•
+I
I-
• - .T_ - - - - - -
't
'
,
'I
I
'- '-'---", +"
,-
Z
w 10
U
II:
• •
I
~-
W
0-
Z
<t
W
::E
2
4
6
6
10
12
14
16
16
20
22
HOUR OF SLEEP EPISODE
FIG. 6. Change in the mean (±SE) percentage of sleep time of REM for all subjects during freerunning for each sequential hour for all bedrest episodes shorter than and longer than 600 min.
Sleep. Vol. 2. No.4. 1980
402
E. D. WEITZMAN ET AL.
REM - NREM Cycle Length
An analysis was made of the REM - NREM cycle length during the different
experimental conditions (Fig. 7). The latencies in minutes from sleep onset to first
mid-REM episode, first mid-REM to second mid-REM, etc., were determined. It
was found that except for a shortened latency from sleep onset to the first midREM episode during free-running, there was no difference in cycle lengths as a
function of experimental condition. There was a consistent decrease in cycle
length for the fourth and fifth cycle for each condition. This REM - NREM cycle
length remained stable (x = 85 min) for up to 11 cycles during the long sleep
episodes (> 10 hr). Thus, there is no evidence that sleep stage cycle length is
altered by the increased sleep-wake period length during free-running conditions.
In addition, previously reported results (Feinberg, 1974) of a stable but slightly
reduced cycle length when sleep is extended are confirmed by these data for those
long sleep episodes which extend from 10 to 20 hr.
DISCUSSION
The shift of REM sleep to an earlier time within the circadian "daily" sleep
episode during free-running supports the concept that the timing of REM sleep is
strongly influenced by an underlying endogenous biological rhythm oscillator.
That is, the decision to go to sleep during the free-running condition was followed
by a very different temporally organized sleep pattern than when the subjects
were told to go to sleep based on their habitual entrained daily bedrest times.
However, the timing of stages 3 + 4 sleep within the sleep episode did not differ
substantially between entrained and free-running, indicating that these sleep
stages are primarily determined by the initiation and course of the sleep process
itself when it occurs within a daily circadian sleep-wake cycle.
The clear shift of waking time to the latter half of the daily sleep episode is also
of considerable interest, since this change is very similar to what is found in
patients who complain of an "early morning arousal" insomnia (Kupfer, 1976).
The characteristic pattern is to have a very short latency to sleep onset, but then
awaken spontaneously "too early," followed by either fragmented sleep-wake
pattern or frank full arousal with an inability to fall asleep again. The demonstration of a phase advance of waking time within the circadian sleep - wake cycle
during free-running in normal subjects strongly suggests that there is a similar
phase shift of the endogenous circadian "waking rhythm" in relation to the
"sleeping rhythm."
These additional results comparing long and short sleep episodes in a large
group of subject-nights strongly support the concept that the timing and amount of
REM sleep during free-running are primarily determined by the phase angle of an
underlying endogenous oscillator and that this process is closely associated with
the duration of total sleep. Recent studies by our group have also indicated that
the cumulative REM minutes, but not REM density (number of eye movements/
min), is phase-advanced relative to sleep onset during free-running conditions
(Zimmerman et al., 1980). We therefore conclude that the initiation and subsequent sleep process itself is not a major determinant of the timing of the under-
Sleep. Vol. 2. No.4. 1980
TIMING OF REM AND STAGES 3 + 4 SLEEP IN MAN
110
~
ENTRAINED
0
FREE-RUNNING
403
N = NUMBER OF SUBJECTS
100
90
(J)
W
t-
::::>
z
80
~
70
Latency
(Mid REM)
N=9 I N=9
N=9
N=9
N=8 N=9N=5 N=9
N=6
REM-REM EPISODES (MID REM)
FIG. 7. Mean sequential REM - NREM cycle lengths (min) during the sleep episodes of subjects
AA - AJ (n = 9) under entrained and free-running conditions.
lying brain stem mechanism controlling the amount of REM sleep. The timing and
amount of stages 3 + 4 sleep, however, do appear to be directly linked to the
initiation and course of the sleep process (for the subsequent 9 hr at least) and are
not dependent on the total duration of sleep. It should be emphasized that a small
amount of stages 3 + 4 sleep was present for each hour even up to the 18th to 20th
hours if the sleep episode persisted.
In a recent analysis of the timing of REM sleep, Moses et al. (1977) have
hypothesized that the REM - NREM cycle is a "sleep-dependent" process, i.e.,
that the initiation of the sleeping process sets the subsequent timing and amount of
REM sleep within the confines of a 90-110 min REM cycle period. However, the
results of the present and other studies do not support that concept. Previous
studies by our group as well as others have indicated that when the timing and
organization of sleep within the daily 24 hr sleep-wake cycle is changed, the
latency and amount of REM sleep after sleep onset can be radically different
(Globus, 1966; Weitzman et aI., 1974, 1979a). Sleep-onset REM episodes and a
shortened REM latency consistently occurred when we acutely inverted the
sleep-wake cycle by 180 (Weitzman et aI., 1974). REM-onset sleep was also
0
Sleep. Vol. 2. No.4. 1980
404
E. D. WEITZMAN ET AL.
shown to cluster at specific clock times in a study of daytime naps (Globus, 1966)
and when sleep onset was delayed past 2 AM, the REM cycle was shortened
(Bfezinova, 1974). In addition, Schulz et ai. (1975) have reported a gradual advance of REM sleep latency when subjects were studied for up to 30 sequential
nights.
In the study of an imposed 10 day, 3 hr sleep-wake cycle (2 hr wake, 1 hr sleep)
previously reported by us (Weitzman et aI., 1974), we found that in spite of an
approximate 40% sleep deprivation, total sleep and REM sleep continued to occur
as a circadian rhythm, although phase shifted by approximately 4-5 hr. REM
occurred during a very restricted phase of the rhythm, 75% occurring during the 3
hr available for sleep from 6 AM -noon (6-7 AM, 9-10 AM, and 11 AM -noon). In
addition, "early" REM onsets occurred 52% of the time. A sleep-onset REM
episode occurred 30 times (out of a total of 196 REM episodes). These REM
onsets also clustered about the 6 AM -noon clock time.
CONCLUSIONS
We confirm previous studies that biological rhythms of human beings free-run
at period lengths greater than 24 hr, typically at approximately 25 hr, but with
individual variability (Aschoff and Wever, 1962, 1976; Siffre, 1965; Mills, 1966,
1974; Siffre et aI., 1966; Aschoff, 1967, 1969, 1970; Chouvet et aI., 1974; Mills et
aI., 1974, 1976; Jouvet et aI., 1974; Webb and Agnew, 1974; Aschoff et aI., 1976;
Wever, 1979). After a variable time of free-running, many normal humans will
spontaneously develop "long" biological days (35 hr), and often these will alternate with "short" days (approximately 25 hr) (Weitzman et aI., 1979a).
During free-running, although the ratio of sleep to total time remains remarkably constant (about 0.30), short sleep episodes «10 hr) occur at a specific phase
angle of an internal circadian rhythm (e.g., body temperature), whereas long sleep
episodes (> 12 hr) take place approximately 180 out of phase with the short sleep
episodes but maintain the same period length. Sleep stage organization changes
during free-running such that REM sleep advances to an earlier time during sleep,
with a shortened REM latency (occurring at times less than 10 min after sleep onset)
and increased amounts during the first 3 hr of sleep. The total REM amount and
percentage for the entire sleep episode, however, remain constant. The timing and
amount of REM sleep following sleep onset also occurred preferentially at a
specific phase of the circadian temperature cycle, strongly supporting the concept
that certain sleep processes in the brain are endogenous biological rhythms. The
distribution of sleep stages 3 + 4 remains essentially the same during the three
experimental conditions. During the long sleep episodes (> 12 hr), stages 3 + 4
recur following 14-16 hr of sleep, indicating that these stages are not dependent
on the length of prior awake time but may be related to length of prior elapsed
time.
A different temporal distribution of REM sleep as well as of waking for the long
(>600 min) compared to short «600 min) sleep episodes indicates the importance
of an endogenous oscillator in the control of these processes. The pattern of stages
3 + 4 was very similar for long and short sleep episodes, however, indicating that
0
Sleep, Vol. 2, No.4, 1980
TIMING OF REM AND STAGES 3 + 4 SLEEP IN MAN
405
the neural mechanisms timing "slow wave" sleep differ from those controlling
REM sleep.
The absence of differences in the REM - NREM cycle during entrained, freerunning, and long and short sleep episodes also strongly suggests that the central
nervous system timing of these short-term events is not directly linked to the
control of the circadian bedrest-activity cycle. The finding that "REM density is
not significantly altered during the free-running sleep-wake condition" (Zimmerman et ai., 1980) supports the above concept.
It is therefore clear that sleep duration and REM timing in man is strongly
determined by the timing of the circadian system, whereas stages 3 + 4 sleep and
the REM-NREM cycle by the sleep process itself. These results support the
concept that sleep-wake cycle disorders in man are abnormalities of biological rhythm functions and that the changes seen during nonentrained (temporal
isolation) conditions may be a model of the changes which accompany sleep
disorders such as occur in depression, narcolepsy-cataplexy, sleep-wake dyssomnias, and delayed sleep phase insomnia (Czeisler et ai., 1979; Weitzman et ai.,
1979b), and aging.
ACKNOWLEDGMENT
Supported in part by NIH Grants MH-28460 and AG-00792, and by ONR Contract 00012-76-C-1071. C.A.C. was supported by the NIH Medical Scientist
Training Program (GM-07365) at Stanford Medical School and J .C.Z. by NIMH
Training Grant MH-06418.
REFERENCES
Aschoff J. Human circadian rhythms in activity, body temperature and other functions. In: AH Brown
and FG Favorite (Eds), Life Science and Space Research, Vol. 5, North Holland, Amsterdam,
1967, pp 159-173.
Aschoff J. Desynchronization and resynchronization of human circadian rhythms. Aerospace Med
40:844-849, 1969.
Aschoff J. Circadian rhythm of activity and of body temperature. In: JD Hardy, AP Gagg, and JA
Stolwijk (Eds), Physiological and Behavioral Temperature Regulation. Thomas, Springfield, III,
1970, pp 905-919.
Aschoff J, Gerecke U, and Wever R. Phase relations between circadian activity periods and body
temperature in man. Eur J Physiol 295: 175-183, 1976.
Aschoff J and Wever R. Spontanperiodik des menschen dei ausschulub aller zeitgeber. Naturwissenschaften 49:337-342, 1962.
Aschoff J and Wever R. Human circadian rhythms: A multi-oscilatory system. Fed Proc
35:2326-2332, 1976.
Bfezinova V. Sleep cycle content and sleep cycle duration. Electroencephalogr Clin Neurophysiol
36:275-282, 1974.
Chouvet G, Mouret J, Coindet J, Siffre M, and Jouvet M. Periodicite bicircadienne du cycle veillesommeil dans des conditions hors du temps. Electroencephalogr Clin Neurophysiol 37:367-380,
1974.
Czeisler CA. Human Circadian Physiology: Internal Organization of Temperature, Sleep-Wake and
Neuroendocrine Rhythms Monitored in an Environment Free of Time Cues. PhD dissertation,
Stanford University, 1978.
Czeisler CA, Richardson GS, Coleman R, Dement W, and Weitzman ED. Successful non-drug treatment of delayed sleep phase syndrome with chronotherapy: Resetting a biological clock in man.
Sleep Res 8: 179, 1979.
Czeisle"r CA, Zimmerman lc, Ronda JM, Moore-Ede MC, and Weitzman ED. Timing of REM sleep is
coupled to the circadian rhythm of body temperature in man. Sleep 2:329-346,1980.
Sleep, Vol. 2, No.4, 1980
406
E. D. WEITZMAN ET AL.
Daan Sand Pittendrigh CS. A functional analysis of circadian pacemakers in nocturnal rodents: II. The
variability of phase response curves. J Comp Physiol 106:253-266, 1976.
Dement WC. Some Must Watch While Some Must Sleep. Stanford Alumni Association, Stanford,
1973.
Dement W, Rechtschaffen A, and Gulevich G. The nature of the narcoleptic sleep attack. Neurology
16:18-33,1966.
Feinberg 1. Changes in sleep cycle patterns with age. J Psychiatry Res 10:283-306, 1974.
Globus GG. Rapid eye movement cycle in real time. Arch Gen Psychiatry 15:654-659, 1966.
Jouvet M, Mouret J, Chou vet G, and Siffre M. Toward a 48-hour day: Experimental bicircadian
rhythm in man. In: F Schmitt and F Worden (Eds), The Neurosciences: Third Study Program. MIT
Press, Cambridge, Mass, 1974, pp 491-497.
Kupfer DJ. REM latency: A psychobiologic marker for primary depressive disease. Bioi Psychiatry
11:159-174,1976.
Mills IN. Human circadian rhythms. Physiol Rev 46:128-171,1966.
Mills IN. Phase relations between components of human circadian rhythms. In: LE Scheving, F
Halberg, and JE Pauly (Eds), Chronobiology. Igaku-Shoin, Tokyo, 1974, pp 560-563.
Mills IN, Minors DS, and Waterhouse JM. The circadian rhythms of human subjects without
timepieces or indication of the alternation of day and night. J Physiol 240:567-594, 1974.
Moses J, Lubin A, Johnson LC, and Naitoh P. Rapid eye movement cycle is a sleep-dependent
rhythm. Nature 265:360- 361, 1977.
Pittendrigh CS. Circadian rhythms and the circadian organization of living systems. Cold Spring
Harbor Symp Quant Bioi 25: 159-184, 1961.
Pittendrigh CS and Daan S. Circadian oscillations in rodents: A systematic increase of their frequency
with age. Science 186:548-550, 1974.
Pittendrigh CS and Daan S. A functional analysis of circadian pacemakers in nocturnal rodents. 1. The
stability and lability of spontaneous frequency. J Comp Physiol 106:223-252, 1976a.
Pittendrigh CS and Daan S. A functional analysis of circadian pacemakers in nocturnal rodents. IV.
Entrainment: Pacemaker as clock. J Comp Physiol 106:291- 331, 1976b.
Rechtschaffen A and Kales A (Eds). A Manual oj Standardized Terminology, Techniques and Scoring
System Jor Sleep Stages oj Human Subjects. UCLA Brain Information Service/Brain Research
Institute, Los Angeles, 1968.
Rechtschaffen A, Wolpert E, Dement W, Mitchell S, and Fisher C. Nocturnal sleep of narcoleptics.
Electroencephalogr Clin Neurophysiol 15:599-609, 1963.
Richardson G, Carskadon M, Flagg W, van den Hoed J, Dement W, and Miller M. Excessive daytime
sleepiness in man: Multiple sleep latency measurement in narcoleptic and control subjects. Electroencephalogr Clin Neurophysiol 45:621-627, 1978.
Schulz H, Dirlich G, and Zulley J. Phase shift in the REM sleep rhythm. Pfleugers Arch 358:203 - 212,
1975.
Siffre M. Beyond Time. Chatto and Windus, London, 1965.
Siffre M, Reinberg A, Halberg F, Chata J, Perdriel G, and Slind R. L'isolement souterrain prolonge.
Presse Med 74:915-920, 1966.
Webb WB and Agnew HW Jr. Sleep and waking in a time-free environment. Aerospace Med
45:617-622, 1974.
Weitzman ED, Nogeire C, Perlow M, Fukushima D, Sassin J, McGregor P, Gallagher TF, and
Hellman L. Effects of a prolonged 3-hour sleep-wake cycle on sleep stages, plasma cortisol,
growth hormone and body temperature in man. J Clin Endocrinol Metab 38: 1018-1030, 1974.
Weitzman ED, Czeisler CA, and Moore-Ede MC. Sleep-wake neuroendocrine and body temperature
circadian rhythms under entrained and non-entrained (free-running) conditions in man. In: M Suda,
o Hayaishi, and H Nakagawa (Eds), Biological Rhythms and Their Central Mechanism.
Elsevier/North Holland, Amsterdam, 19790, pp. 199-227.
Weitzman ED, Czeisler CA, Coleman R, Dement W, Richardson G, and Pollak C. Delayed sleep phase
syndrome: A biological rhythm sleep disorder. Sleep Res 8:221, 1979b.
Wever R. The Circadian System oj Man. Springer-Verlag, New York, 1979.
Zimmerman JC, Czeisler CA, Laxminarayan S, Knauer R, and Weitzman ED. REM density is dissociated from REM sleep timing during free-running sleep episodes. Sleep 2:409-415, 1980.
DISCUSSION
Dr. Dement asked what the differences were between the subjects with long and short
sleep episodes. Dr. Kripke mentioned that Lund (1974) reported significant differences on
Sleep. Vol. 2. No.4. 1980
TIMING OF REM AND STAGES 3 + 4 SLEEP IN MAN
407
psychological tests between subjects with "internal de synchronization" and the others.
Dr. Weitzman, after objecting to the term "internal desynchronization," emphasized that
all subjects had been psychologically screened and were without significant psychological
abnormalities, yet some developed these alternating long and short sleep episodes. Dr.
Czeisler added that every normal subject reported in the literature living in temporal isolation for more than 2 months always developed patterns of long and short sleep episodes,
suggesting that the difference between these subjects may relate to the rate at which those
patterns develop.
Dr. Dement asked if the early concerns with "total isolation" from all cues were necessary in the design of these studies. Dr. Weitzman felt that the absence of scheduling, rather
than isolation from subtle environmental cues, was the key factor.
References
Lund R. Personality factors and the de synchronization of circadian rhythms. Psychosom Med
36:224-228, 1974.
Sleep. Vol. 2. No.4. 1980