Sleep, 20(7):505-511
© 1997 American Sleep Disorders Association and Sleep Research Society
Sleep Onset
Nighttime Drop in Body Temperature: A Physiological
Trigger for Sleep Onset?
Patricia J. Murphy and Scott S. Campbell
Laboratory of Human Chronobiology, Department of Psychiatry, Cornell University Medical College,
White Plains, New York, U,S,A,
Summary: Relationships between changes in the slope of the body temperature (BT) and the initiation of sleep
were examined in 44 subjects ranging from 19 to 82 years of age, Following an adaptation night, subjects remained
in the laboratory for a baseline night and 72 hours in temporal and social isolation, with strictly limited behavioral
options ("disentrainment") during which continuous electroencephalograph (EEG) and BT were recorded. Polysomnographic sleep variables (e.g. sleep onset, percentage of each sleep stage) were determined for nighttime sleep
periods at baseline and during the disentrainment period. Periods of the BT curve surrounding these sleep bouts
were examined for minute to minute changes, and the time at which the maximum rate of decline (MROD) in
temperature occurred was compared with the time of sleep onset (SO) and sleep quality parameters. On the baseline
night, the MROD occurred, on average, 60 minutes prior to SO. During disentrainment, the MROD occurred, on
average, 44 minutes prior to SO. The proximity of MROD to SO did not affect subsequent sleep quality on the
baseline night, but during disentrainment, there were significant correlations between the interval from MROD to
sleep onset and the amount of slow-wave sleep (SWS) obtained during the sleep bout. There were no significant
age differences on variables related to MROD on baseline night, but the timing of both MROD and SO were
significantly advanced in older, relative to younger, subjects during the disentrainment period. It is suggested that
a rapid decline in core body temperature increases the likelihood of sleep initiation and may facilitate an entry into
the deeper stages of sleep. Key Words: Core body temperature-Sleep-Disentrainment-Aging.
In healthy young adults, most nighttime sleep episodes occur between ~6 hours before and 2 hours after
the body temperature (BT) minimum (1,2). While the
association between the timing of major sleep episodes
and the circadian rhythm of body temperature is well
documented, less is known about the nature of the relationship between the initiation of sleep and body
temperature.
On one hand, it has long been maintained that the
changes in activity (e,g. posture) associated with going
to bed and the subsequent reduction in metabolism associated with falling asleep result in a decline in BT
(3,4). Indeed, this view is supported by considerable objective data (5-7). Yet, few studies have examined BT
changes in the interval prior to sleep onset in an effort
to determine whether temperature changes may provide
a physiological signal that the brain is ready for sleep.
Accepted for publication May 1997.
Address correspondence and requests for reprints to Patricia J.
Murphy, Ph.D., The New York Hospital-Cornell Medical Center, 21
Bloomingdale Rd., White Plains, NY 10605, U.S.A.
In an earlier study (8) we explored this possibility
using a group of elderly subjects whose bedtimes and
waketimes were not circumscribed, thus permitting an
assessment of the natural relationship between BT and
sleep onset (SO). In that study, we found that eight of
10 subjects showed a maximum rate of decline
(MROD) in BT prior to their decision to retire (i.e.
verbally informing staff that they were "going to
bed") and the subsequent initiation of sleep (i.e. polysomnographically defined SO), with an average interval between MROD and SO of 65 minutes.
Because those subjects were older and sleep-disturbed, there may be some question as to the generalizability of the findings. The current investigation
sought to replicate our previous study in a group of
young and old non-sleep-disturbed subjects. In addition, we wished to examine the consistency of the relationship between MROD and sleep initiation without
the influence of time cues or social factors in an environment with minimal levels of activity. Thus, the
current study was undertaken while subjects lived in
temporal and social isolation.
505
P. 1. MURPHY AND S. S. CAMPBELL
506
DISENTRAINMENT
EXPERIMENTAL DESIGN
I:
~;j ;;:;;;:ad<iP'Wion:;::j;
1: ,;1
/:: 1:i;1i4SeHnl(;H:+ls~;i~~ij.tr~!$.ro:erit: :{~:;:;:;':;:; ;;;,
In
M
••
~ .,~
".,
'
"~'
,~:
.",
,
.~,".
,-,
iii' ~ ,~ :: .~' '.~
.< "'. ,<
:.i '.",,; " ..;:~ o,~' ~~'
.,vi,~
\:,";'
6
.,v ~
I
I
I
24
~'~;;~<,S;;I
,,<::' .;: ., ~:::.'
12
18
24
FIG. 1. Schematic of the "disentrainment" experimental protocol.
Subjects spent one adaptation night in the laboratory, followed immediately by one baseline night. An electrode montage was placed
on the subject by 2200 hours on the baseline night, after which time
he or she was alJowed to engage in leisure activities within the
laboratory until bedtime. From 0800 hours the following morning,
and for the next 72 hours, the subject was in temporal and social
isolation and was instructed as follows: "Do not structure your day
by eating and sleeping at regular intervals; rather, eat and sleep
whenever inclined to do so". Behavioral options were strictly limited in order to facilitate expression of sleep behavior.
METHODS
Procedures
Forty-four subjects (mean age, 41.5 ± 19.1 years,
range 19-82 years; 21 males, 23 females) participated
in the study. For some analyses (where noted) a subset
of young (n = 17; mean age = 22.9 ± 2.9 years, all
<30 years) and old (n = 13; mean age = 67.6 ± 6.1
years, all >60 years) subjects was used.
Potential subjects received a physical examination
and psychiatric screening [mini-mental status examination (9) and 24-item Hamilton depression rating
scale (10) for all subjects, plus geriatric depression
scale (11) for old subjects] to exclude from participation those with active, serious medical problems or a
current or past history of psychiatric disorder. Minor,
controlled medical problems such as mild hypertension
or mild arthritis were not grounds for exclusion from
the study. On the first night in the laboratory, all subjects over age 60 were screened for sleep pathology
including periodic limb movements (PLM) and sleep
apnea. A PLM index <5lhour and an apnea index
< 1Olhour were necessary for inclusion in the study. In
addition, all subjects completed a daily sleep log for
2 weeks prior to study participation. In this log they
noted bedtime and waketime each night, the approximate time and duration of aWakenings, as well as any
daytime naps, and provided a subjective estimate of
each night's sleep quality. Only subjects who reported
regular, normal sleep patterns (e.g. slept between 6 and
9 hours per night) were included in the study.
As shown in Fig. 1, the experimental period conSleep, Vol. 20, No.7, 1997
sisted of an adaptation night and a baseline night, followed immediately by three consecutive days and
nights in a "disentrained" environment (12,13). During this time, subjects lived individually in studio
apartments isolated from all cues to time of day.
Throughout disentrainment, subjects were encouraged
to eat and sleep whenever they felt inclined to do so;
they were specifically instructed not to try to overcome
bouts of sleepiness, but rather to respond to such feelings by taking a nap or major sleep episode. Compliance with these instructions was further encouraged by
severely restricting behavioral options during the disentrainment period. Subjects had only a deck of cards,
a jigsaw puzzle, and very limited reading material.
They were permitted only minimal exercise (e.g.
stretching) and were discouraged, both prior to participation and during their time in disentrainment, from
doing any physical activity that might keep them
awake.
This protocol was designed to maximally "unmask"
sleep propensity, specifically, to exlude social and temporal constraints that may exert a strong influence on
the natural expression of the endogenous sleep system.
One important requirement of the disentrainment protocol is that subjects are maintained in a relatively sedentary environment. Consequently, activity levels during disentrainment were consistently low, and most
subjects were sitting on the couch or laying in bed a
majority of the time. Showers were not permitted during the disentrainment period, and subjects usually
wore pajamalike, comfortable clothing throughout the
entire 72 hours. In addition, subjects were continuously monitored, thereby ensuring no activities other than
those explicitly permitted by the protocol (playing solitaire, working on a jigsaw puzzle). Illumination in the
isolation suite was provided by desk and floor lamps
and did not exceed 100 lux.
On the night following adaptation, subjects reported
to the laboratory between 2000 hours and 2100 hours
(baseline night). Within 30 minutes of their arrival,
they inserted indwelling rectal thermistors (Yellow
Springs Series 4400, VWR Scientific Products, Cleveland, OH) attached to ambulatory recording devices.
Body core temperature was recorded continuously, in
1 minute epochs, throughout the baseline night and the
disentrainment period, except for during brief periods
for personal hygiene. Electrode placement for polysomnography (electroencephalogram (EEG) sites F3,
C3, P3, F4, C4, and P4 referenced to linked mastoids;
cross-referenced electrooculogram (EOG) sites; submental electromyogram (EMG) sites) was completed
by 2200 hours. As with temperature, EEG, EOG, and
EMG were recorded throughout the entire study. These
data were acquired using a Biosentry telemetry system
(Biosentry, Torrance, CA) that permitted complete free-
507
DROP IN BODY TEMPERATURE PRIOR TO SLEEP
dom of movement around the isolation suite. The continuous collection of telemetric data also allowed subjects to sleep anywhere in the isolation suite at any
time, without having to plug in a cable to start a sleep
recording.
Following electrode placement, subjects retired to
their isolation suites. That night's sleep, and all subsequent sleep episodes, were initiated and terminated
at the subjects' discretion. Electrographic data were
transmitted from the telemetry device to an Oxford
Instruments 847 Sleep Analyzing Computer for on-line
waveform display and storage to optical disk for subsequent scoring and analysis. Recordings from the
baseline night throughout the entire 72 hours of disentrainment were scored by trained scorers in 30-second
epochs according to standard criteria (14). Sleep onset
was defined as the first epoch of stages 2, 3,4, or rapid
eye movement (REM) sleep.
Data analysis
Temperature data sets for the baseline night were
generated for each subject. These temperature sets included data from the time of insertion of the rectal
thermistor until 0800 hours the following morning.
Additional data sets were also generated for the dis entrainment period. Each subject's 72-hour disentrainment period was examined for sleep bouts that met
two criteria: 1) they were initiated (i.e. SO occurred)
between 2200 and 0600 hours, and 2) they were at
least 4 hours in length, with < 1 hour of intervening
wake time. A temperature set was then generated that
included the 4 hours prior to initiation of each of these
sleep bouts, the duration of the sleep bout, and at least
2 hours following the sleep bout.
Although of interest (15), results from temperature
changes prior to daytime sleep episodes during disentrainment (i.e. initiated outside the 2200-0600 hour
window) were excluded for the following reasons: 1)
examination of all sleep bouts during disentrainment
revealed that the average length of sleep periods was 4:
31 (SD 3:09) and that 83% of all sleep bouts 2:4 hours
in length were initiated between 2200 and 0600 hours
(i.e. only 17% were initiated between 0600 and 2200
hours); 2) there were no baseline data collected during
the daytime prior to the disentrainment period that
would permit comparison with daytime data from the
time-free environment; and 3) altering the criteria to
include shorter daytime sleep episodes (i.e. naps) would
change the focus of the current paper, as the temperature
pattern during the daytime hours, prior to nap sleep, is
markedly different from the course of temperature during the nighttime hours prior to nocturnal sleep episodes. Application of the criteria stated above would
permit direct comparison with our earlier study (8).
37.60
37.40
....0
'"
37.20
~
;:l
f-<
~
'"::;:0..
'"f-<
37.00
0.00015
~
\\
0.00010
~E;
I
I
I
,
\
36.80
I
I
.... ,
'-
~
.-
, ,..,
36.60
36.40
I
I
~
\
\
-.."
*
•
2200 2400 0200
"'c
0...,
0.00000
~$
z:S:
-0.00005
",Z
_c
...,'"
\
I
I
l
""Z
0.00005
C')-
t1...,
-0.00010
-'"
-0.00015
0400
0600
0800
CLOCK TIME (hr)
FIG. 2. Sample temperature plot and rate of change curve. These
data depict a raw rectal temperature data set (solid line) across an
approximately 12-hour period. Temperature data sets for each subject comprised the period from 2 hours prior to sleep onset (SO)
through the entire sleep period and 4 hours following termination
of the sleep period. The raw data were fitted with a ninth-order
polynomial curve. The minute to minute rate of change of the polynomial curve fit was calculated and then superimposed (dotted line)
on the raw temperature data. The maximum rate of decline in the
rate of change curve (MROD) was determined (denoted by *) and
compared with the time at which SO (first epoch of stage 2) occurred
(denoted by 6.). This procedure was completed for each baseline
night sleep period and for sleep bouts during the disentrainment
period that were initiated between 2200 and 0600 hours and were
of at least 4 hours' duration with less than I hour of intervening
wakefulness.
Both the baseline and dis entrainment temperature
data sets were smoothed using a ninth-order polynomial curve fit. This type of curve fit was chosen because such a procedure does not involve restrictive assumptions about the nature of the data, because polynomial regression reliably accounts for more of the
variance in a given data set, and because using this
procedure allowed for a closer fit to transient changes
in temperature (16). A rate-of-change curve (degrees/
minute) was then generated for each data set to determine the magnitude by which body temperature declined (or increased) from minute to minute. A sample
temperature plot, with the corresponding rate-ofchange curve, is shown in Fig. 2.
From each rate-of-change curve, the time at which
BT declined at its maximum rate could be ascertained.
The time at which this MROD occurred was then compared with the time of SO for the corresponding sleep
period to determine the MROD to sleep onset interval
(Fig. 2).
A series of correlation coefficients was generated to
examine relationships between MROD, SO, various
sleep parameters, and the temperature minimum
(Tmin) in closest proximity to each sleep bout. To determine if there were differences between the young
and old subjects in sleep quality, or in the relationship
between BT and sleep timing, one-way analyses of
variance (ANOVAs) for age were performed for the
Sleep, Vol. 20, No.7, 1997
P. J. MURPHY AND S. S. CAMPBELL
508
SLEEP ONSET
wave sleep (SWS)] throughout the sleep period, SWS
in the first hour after SO, and the number of awakenings in the sleep period. For the group as a whole,
the MROD to sleep interval was not significantly correlated with any of these measures of sleep quality on
the baseline night. Further, there was not a reiationship
between the timing of MROD and the BT minimum.
T
...
en
u
~
fO
::J
en
~f
I
20
15
....
0
I"i
~
10
'"~
Baseline night, young versus old subjects
Z
o
o
~~~~~~~~~~~~~L-~
-5
-3
-4
-2
-1
o
+1
+2
MROD RELATIVE TO SLEEP ONSET
(HOURLY INTERVALS)
FIG. 3. Distribution, in hourly intervals, of timing of the maximum rate of decline (MROD) in body temperature (BT) relative to
sleep onset (SO) for sleep baseline night sleep periods.
baseline night and disentrainment sleep periods. In addition, to investigate potential differences between the
baseline night sleep periods and sleep bouts during
disentrainment, paired t tests were performed.
RESULTS
Baseline night all subjects
A verage SO time for all subjects on the baseline
night was 2430 hours (SD = 54 minutes). Average
MROD time was 2330 hours (SD = 72 minutes).
Thus, the average difference between MROD and SO
was 60 minutes (SD = 78 minutes). The average nadir
of temperature (obtained from the ninth-order polynomial fit) was 0354 hours (SD = 126 minutes) and
occurred, on average, 264 minutes (SD = 138 minutes) after MROD.
Figure 3 shows the distribution of the timing of
MROD relative to SO for each hourly interval for the
baseline night sleep periods. The MROD occurred prior to SO in 89% (39/44) of subjects (X2(l) = 13.68, P
< 0.001); for 77% (30/39) of those subjects, MROD
occurred within the 2 hours prior to sleep onset (X2(2)
= lO.74, P < 0.01). If only those subjects whose
MROD occurred prior to sleep are considered (i.e. excluding the five subjects whose MROD occurred after
SO), the average interval from MROD to SO was 75
minutes (SD = 73 minutes).
To examine the relationship between the timing of
MROD relative to SO and subsequent sleep quality,
the MROD to sleep onset interval for each subject was
correlated with the following measures of sleep quality: percentage of wakefulness after sleep onset
(WASO) throughout the sleep period, WASO in the
first hour, percentage of stages 3 and 4 sleep [slowSleep, Vol. 20, No.7, 1997
There were 17 young subjects «30 years) and 13
old subjects (>60 years) from whom baseline night
data were analyzed. Average SO time was 2430 hours
(SD = 48 minutes) for young versus 2412 hours (SD
= 60 minutes) for old subjects (ns). The MROD occurred at 2330 hours (SD = 72 minutes) for the young
group and at 2324 hours (SD = 66 minutes) for the
old group (ns). Thus, MROD occurred 60 minutes (SD
= 67 minutes) prior to SO in young subjects and 47
minutes (SD = 79 minutes) prior to SO in old subjects
(ns). The average Tmin occurred at 0400 hours (SD =
114 minutes) for the young group and at 0254 hours
(SD = 108 minutes) for the old group (ns). The time
from MROD to Tmin was 270 minutes (SD = 132
minutes) and 210 minutes (SD = 120 minutes) for the
young and old groups, respectively (ns). Of the five
subjects whose MROD occurred after SO, two were
from the young and three were from the old group
(ns). When these subjects were not considered, the interval from MROD to SO was 64 minutes (SD = 52
minutes) and 52 minutes (SD = 67 minutes) for the
young and old subjects, respectively (ns).
To determine whether the timing of MROD differentially influenced sleep quality and to assess whether
the age groups differed in sleep quality, one-way
ANOVAs were performed to compare young and old
subjects on the following variables: WASO and SWS
in the first hour after SO, amount of WASO and SWS
throughout the night, and number of awakenings
throughout the night. Although group differences were
not significant for MROD time, SO time, Tmin time,
MROD to SO interval, or MROD to Tmin interval,
significantly lower sleep quality was observed in old
subjects. Old subjects had more WASO in the first
hour after SO [F(l,28) = 8.85, p < 0.01], which resulted in less SWS in the first hour of sleep [F(l,28)
= 8.98, P < 0.01]. In addition, old subjects had more
WASO throughout the night [F(1,28) = 23.51, p <
0.001]. The number of awakenings did not differ between young and old subjects.
Disentrainment period, all subjects
For the disentrainment period, 65 sleep bouts that
met the criteria stated above were obtained. Average
DROP IN BODY TEMPERATURE PRIOR TO SLEEP
SLEEP ONSET
T
25
...
'";:J=>
<Jl
-
20
U
15
<Jl
'"
0
~
'"::E=>
10
;:J
Z
5
0
0
-5
-4
-3
-2
-1
0
+1
+2
MROD RELATIVE TO SLEEP ONSET
(HOURLY INTERVALS)
FIG_ 4_ Distribution, in hourly intervals, of timing of the maximum rate of decline (MROD) in body temperature (BT) relative to
sleep onset (SO) for sleep periods during the 72-hour disentrainment
period.
onset time for nighttime sleep episodes recorded during disentrainment was 2455 hours (SD = 148 minutes), which was not significantly different from the
mean SO time on the baseline night. Corresponding
MRODs occurred, on average, at 2411 hours (SD =
148 minutes). Thus, the average difference between
MROD and SO was 44 minutes (SD = 64 minutes)
when subjects were in a disentrained environment. For
the group as a whole, the MROD time was significantly later during disentrainment than at baseline (t
= -2.7, P < 0.05), and the interval from MROD to
SO was significantly shorter than on the baseline night
(t = 7.77, P < 0.001). When only subjects whose
MROD occurred prior to SO were considered, the average time from MROD to SO was 76 minutes (SD =
45 minutes). The average time of the temperature nadir
associated with these sleep bouts was 0518 hours (SD
= 130 minutes), which was 307 minutes (SD = 128
minutes) after MROD. The time of Tmin was significantly later during disentrainment than at baseline (t
= -2.4, P < 0.05), but because MROD was also later,
the proximity of MROD to Tmin was not significantly
different from baseline.
Figure 4 shows the distribution of MRODs relative
to SO for each hourly interval for disentrainment period sleep bouts. The MROD occurred prior to SO in
74% (48/65) of the sleep bouts (X2(1) = 6.89, P <
0.01) and within the 2 hours prior to SO in 79% (38/
48) of those subjects (X2(1) = 7.70, P < 0.01). Figure
4 also demonstrates the consolidation of the MROD to
SO intervals during disentrainment, as well as the decrease in variability of this interval. That is, of subjects
whose MROD occurred prior to SO, all fell asleep
within the 3 hours after their MROD.
As with baseline sleep periods, the relationship be-
509
tween the MROD to SO interval and measures of sleep
quality were examined for these sleep episodes. During disentrainment, there was a significant negative
correlation (r = -0.24) between the proximity of
MROD to SO and the amount of SWS obtained
throughout the night. When only those subjects whose
MROD occurred prior to SO were considered, the relationship between the proximity of MROD to SO and
SWS throughout the sleep period was stronger (r =
-0.34). In addition, significant correlations between
the proximity of MROD to SO and SWS in the first
hour of sleep (r = -0.30), as well as the amount of
WASO throughout the sleep period (r = 0.28), were
observed. These results indicate that subjects with a
shorter interval between MROD and SO spent less
time awake within the sleep bout and obtained more
SWS in the first hour of sleep and throughout the
night.
Disentrainment period, young versus old subjects
When the sample was subdivided into young and
old groups, there were 26 sleep bouts obtained from
13 young subjects and 14 sleep bouts obtained from
nine old subjects. Four young and four old subjects
had no sleep bouts that met the criteria. In every case,
these eight subjects contributed no sleep bouts during
disentrainment because the sleep bouts that occurred
between 2200 and 0600 hours were <4 hours in
length, rather than because sleep bouts that met the
duration criterion were initiated outside of the 22000600 hour window. Average SO time for the young
group was 0122 hours (SD = 128 minutes) versus
2322 hours (SD = 166 minutes) for the old group
[F(1,38) = 6.44, p < 0.05]. Average MROD time was
2446 hours (SD = 142 minutes) for the young group
and 2247 hours (SD = 156 minutes) for the old group
[F(1,38) = 5.06, p < 0.05]. The average interval between MROD and SO was 36 minutes (SD = 62 minutes) in young subjects and 35 minutes (SD = 66 minutes) in old subjects (ns). The temperature nadir occurred at 0525 hours (SD = 118 minutes) for the
young group and at 0506 hours (SD = 140 minutes)
for the old group (ns). The amount of time from
MROD to the temperature nadir averaged 279 minutes
(SD = 117 minutes) and 379 minutes (SD = 143 minutes) for the young and old groups, respectively. Although the time of Tmin did not differ between age
groups, the MROD was earlier in the old relative to
the young group; thus, the interval from MROD to
Tmin was significantly different between age groups
[F(1,38) = 8.06, p < 0.01]. Of the 13 sleep bouts that
were initiated prior to MROD, nine were contributed
by young subjects and four were contributed by old
subjects (ns). When these 13 sleep bouts were not conSleep, Vol. 20, No.7, 1997
P. J. MURPHY AND S. S. CAMPBELL
510
sidered, the average time from MROD to SO was 54
minutes (SD = 65 minutes) and 50 minutes (SD = 62
minutes) for the young and old subjects, respectively
(ns).
As with the baseline night, one-way ANOVAs
across age groups were performed for variables related
to sleep quality. In contrast to the baseline night, during these disentrainment sleep bouts, there were no
significant differences between age groups on any
sleep quality measures with the exception that the
amount of W ASO in the 1st hour of the sleep bout
was greater for old subjects [F(l,38) = 4.57, P <
0.05]. However, the old subjects had significantly less
WASO throughout the night during the disentrainment
sleep periods compared to their baseline night sleep (t
= 2.88, P < 0.05).
While the time of SO for sleep periods initiated between 2200 and 0600 hours was advanced during disentrainment for old subjects, it was delayed for young
subjects relative to baseline SO times. The withingroup changes in SO time from baseline to disentrainment were not significant for either age group, even
though the disentrainment MROD and SO times were
significantly later for the young group compared to the
old group, as stated above.
DISCUSSION
These data indicate that BT shows a maximum rate
of decline, on average, 60 minutes before the initiation
of major nighttime sleep episodes. This relationship
held even when subjects were isolated from time cues
and had no behavioral, social, or experimental constraints restricting sleep initiation (i.e. sleep onset).
These results, using both young and old non sleepdisturbed subjects, support and extend those previously reported for a group of sleep-disturbed elderly individuals, and they strongly suggest that a rapid decline in BT may act as the brain's signal that one is
physiologically ready for sleep. Nearly 90% of all subjects studied exhibited their maximum rate of decline
in BT within a few hours prior to falling asleep.
Nevertheless, it is clear that the putative signal provided by a precipitous drop in BT may be overridden
by motivation or homeostatic factors since there were
sleep episodes that were initiated before the occurrence
of MROD. The results from this study merely note that
the occurrence of MROD greatly increases the likelihood of the occurrence of SO within approximately 60
minutes, without the implication that SO must occur
within a few hours of MROD. In this regard, there
were no apparent or systematic differences between
sleep episodes that were initiated before the occurrence
of MROD and the majority that occurred shortly after
MROD. No variables that were examined, including
Sleep, Vol. 20, No.7, 1997
gender, age, time of Tmin, proxImity of MROD to
Tmin, or sleep quality parameters distinguished between sleep episodes that were initiated before versus
after MROD. There was a trend for sleep episodes that
were initiated prior to the occurrence of MROD to start
at an earlier clock time than the group average and for
MROD to occur at a later clock time than the group
average. In several of these cases, a secondary drop in
BT that was not at the maximum rate, but nonetheless
a rapid and steep temperature decline, preceded SO.
lt may be argued that any such relationship between
BT and sleep observed under entrained conditions is
confounded by social and temporal cues. However, in
the current experiment, when time and social cues
were eliminated, the amount of time between MROD
and SO decreased. We interpret this finding as demonstration of a "tighter" relationship between MROD
and sleep initiation, suggesting that subjects were more
likely to "respond" to their MROD by going to sleep
when other behavioral responses were limited.
Given that the electrode montage was placed on the
subjects 2-3 hours prior to SO on the baseline night,
the drop in BT that produced the MROD, as well as
the relationship between MROD and SO, could have
been as artefact of the behavior of sitting in a chair
and the general period of relaxation that accompanied
the hookup procedure. Yet, the fact that the relationship between the drop in BT and subsequent SO held
while subjects were in the disentrained environment,
when a hookup prior to SO was not necessary, argues
against this notion. Additionally, the strict limitations
on behavior and activity during the disentrainment period and the minimal activity required prior to sleeping
(i.e. subjects did not need to change into pajamas or
brush their teeth and were generally already laying in
their beds for long periods prior to SO) decrease the
likelihood that the MROD prior to sleep initiation was
the result of a change in activity levels. Nevertheless,
the effects of activity on BT are considerable, and the
present results cannot rule out the possibility that activity changes in the hours prior to SO were associated
with the MROD. Experiments that analyze activity
levels are needed to confirm the present findings.
Although there was not a strong relationship between how soon after the MROD subjects fell asleep
and subsequent sleep quality in this study, there were
statistically significant correlations between proximity
of MROD to SO and 1) the amount of SWS (i.e. more)
during both the first hour after SO and throughout the
night, and 2) the amount of WASO (i.e. less) throughout the sleep episode for the disentrainment sleep periods. As stated above, the average time from MROD
to SO was significantly shorter in the disentrained condition compared to the baseline, entrained conditions.
Taken together, these results suggest that responding
511
DROP IN BODY TEMPERATURE PRIOR TO SLEEP
to one's MROD signal may facilitate a more rapid entry into the deeper stages of sleep. However, it should
be emphasized that these correlations were not robust,
explaining only 8-12% of the variance in the data.
Further evidence that responding to one's MROD
may provide some benefit for subsequent sleep relates
to the advance of old subjects' MROD and SO times
during disentrainment. Old subjects' MRODs occurred
nearly an hour earlier when in temporal isolation than
in entrained conditions, and these subjects fell asleep
earlier than on their baseline nights. Interestingly, there
was a significant decrease in W ASO throughout the
sleep period for the old subjects when in disentrainment, and the differences in sleep quality between
young and old subjects that were observed on the
baseline night all but disappeared. These results suggest that although the advance in MROD and SO times
from baseline to disentrainment was not statistically
significant for these subjects, the shorter interval between MROD and SO may have contributed to a reduction in WASO in the disentrainment sleep periods.
In practical terms, this may mean that if older individuals were to initiate nighttime sleep at an earlier clock
time, the quality of their nighttime sleep might improve.
One might expect that old subjects' SO times would
be earlier than 2200 hours, and the time window criterion used for these analyses (i.e. 2200-0600 hours)
might constitute a source of bias against inclusion of
old subjects' nighttime sleep episodes. However, even
though the SO time was at 2230 hours and the MROD
time was at 2247 hours for the old group, there were
not a significant number of sleep bouts "missed" by
the use of the time window criterion. In fact, during
disentrainment, only five of 177 sleep periods of >4
hours in length were initiated between 1800 and 2200
hours, and only three of those were in old subjects. Of
all sleep periods, regardless of duration, only 17 began
in the interval from 1800 to 2200 hours. The potential
bias that might have been a problem if old subjects
were going to bed outside the time window used for
these analyses was, therefore, not a problem.
In summary, SO during the nighttime hours is facilitated by a precipitous drop in BT. In a manner similar to the increased likelihood of sleep termination on
the rising portion of the temperature curve, sleep initiation is more likely to occur shortly after temperature
declines most rapidly. The relevance of this drop in
BT for the entirety of the subsequent sleep period is
not clear, but the amount of SWS obtained immediately following SO, as well as the amount of WASO
throughout the sleep period, may be influenced by the
response to MROD.
One corollary to the current findings is that introducing a rapid and steep decline in BT might facilitate
SO. For example, inducing a rapid decline in BT in
an individual with SO insomnia, with, for example, an
appropriately timed hot bath or administration of exogenous melatonin, might decrease SO latency.
Whether there is a corresponding change in temperature, or perhaps another type of change in temperature
prior to daytime naps, is also of interest. We are currently examining the possibility that changes in the
slope of the BT rhythm during the daytime nap phase
are related to the initiation of naps, as well as the quality of the nap sleep.
Acknowledgements: This research was supported by
grants ROl AGl2112, ROl MH45067, K02 MH01099, P02
MH49762, and a grant from the Tolly Vinik Trust, CUMC.
The authors would like to thank Catharine Boothroyd and
Tom Stauble for assistance with data processing and the
technicial staff of the Laboratory of Human Chronobiology.
REFERENCES
1. Czeisler CA, Weitzman ED, Moore-Ede M, Zimmerman J,
Knauer R. Human sleep: its duration and organization depend
on its circadian phase. Science 1980;210: 1264-7.
2. Zulley J, Wever R, Aschoff J. The dependence of onset and
duration of sleep on the circadian rhythm of rectal temperature.
Pfiugers Arch 1981;391:314-8.
3. Fraser G, Trinder J, Colrain 1M, Montgomery I. Effect of sleep
and circadian cycle on sleep period energy expenditure. J Appl
Physiol 1989;66(2):830-6.
4. Kleitman N, Doktorsky A. The effect of the position of the body
and of sleep on rectal temperature in man. Am J Physiol 1933;
104:340-3.
5. Barrett J, Lack L, Morris M. The sleep-evoked decrease of body
temperature. Sleep 1993;16(2):93-9.
6. Wever RA. Internal interactions within the human circadian system: the masking effect. Experientia 1985;41 (3):332-42.
7. Gillberg M, Akerstedt T. Body temperature and sleep at different
times of day. Sleep 1982;5(4):378-88.
8. Campbell SS, Broughton RE. Rapid decline in temperature prior
to sleep onset: fluffing the physiological pillow? Sleep 1994; 17:
126-31.
9. Folstein MP, Folstein SE, McHugh PRo Mini-mental state: a
practical method for grading the cognitive state of patients for
the clinician. J Psychiatr Res 1975; 12: 189-98.
10. Hamilton M. A rating scale for depression. J Neural Neurosurg
1960;23:56-62.
II. Yesavage JA, Brink TL, Rose TL, et al. Development and validation of a geriatric depression screening scale: a preliminary
report. J Psychiatr Res 1982;17(1):37-49.
12. Campbell SS, Zulley J. Ultradian components of human sleep/
wake patterns during disentrainment. In: Schulz H, Lavie P, eds.
Ultradian rhythms in physiology and behavior. Berlin: SpringerVerlag, 1985:234-55.
13. Zulley J, Campbell SS. Napping behavior during "spontaneous
internal de synchronization ": sleep remains in synchrony with
body temperature. Hum Neurobiol 1985;4(2):123-6.
14. Rechtschaffen A, Kales A. A manual of standardized terminology. techniques and scoring system for sleep stages of human
subjects, vol. 204. Washington, DC: National Institutes of
Health, 1968.
15. Campbell SS. The timing and structure of spontaneous naps. In:
Stampi C, ed. Why we nap: evolution, chronobiology, andfunctions of polyphasic and ultrashort sleep. Boston: Birkhauser,
1992:71-81.
16. Broughton R, Campbell S, Dunham W. Curve fitting of chronobiological data: some issues and techniques. Sleep Res 1993;22:
395.
Sleep, Vol. 20, No.7, 1997