STUDIES IN THE
TOTAL SLEEP DEPRIVATION
HYMAN S. STERNTHAL
A DISSERTATION PRESENTED TO THE GRADUATE
COUNCIL OF THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR
THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
The author wishes to express his gratitude to
Dr. Wilse B. Webb for his patience and guidance in
supervising this research. The author also expresses
gratitude to the members of his supervisory committee,
Drs. Frederick A. King, Robert L. King, C. Michael Levy,
and Henry S. Pennypacker.
He also wishes to express his appreciation to
assistant Dennis Bailey, who developed a quicker technique
for scoring EEG records and greatly facilitated the work
of these studies.
TABLE OF CONTENTS
LIST OF TABLES..................................... iv
LIST OF FIGURES..................................... vi
ABSTRACT.......................... .. ............... vll
INTRODUCTION ......................................... 1
METHOD......................... .................... 2
EXPERIMENT I................... .......... ........ 7
METHOD.............. ....................... .... .... 9
RESULTS AND DISCUSSION............................. 11
SUMMARY.. ....................................... .. 22
EXPERIMENT II................. .......... ..... ..... 23
METHOD....... ................... ................... 27
RESULTS......................... ........... ........ 28
DISCUSSION.......... .. ................... .. ......... 35
SUMMARY........................ ......... ... ....... 41
EXPERIMENT III.................................... 43
METHOD ............... ....... ....... ............... 44
RESULTS AND DISCUSSION............................. 47
SUMMARY.................... ... ............. ...... .. 53
EXPERIMENT IV............................... ........ 56
METHOD.................. ....... .................... 63
RESULTS AND DISCUSSION. 0- .. ................... ... 64
SUMMARY... ........................... ......... 83
REFERENCES...... .................................. 86
BIOGRAPHICAL SKETCH....... .. ... ................. .. 90
LIST OF TABLES
1. Shocks per Hour per Individual S in Shock-S,
Shock-W, Group A and Total per S after 19
Hours of Deprivation......................... 13
2. Analysis of Variance Based on First 19 Hours
of Deprivation for all Three Groups for
Data in Table 1 ................... ........... 15
3. Hours of Deprivation by Procedure A for
Individual Ss till Override of Shock Whose
Intensity and Duration are Specified.......... 16
4. Frequency of Occurrence of Each Permutation
of Intershock Intervals for Group A.......... 21
5. Summary of the Literature on the Limit of
Sleep Deprivation............................ 24
6. Baseline and Recovery Sleep after Deprivation
Procedures A and P in Mean Percent per
24-Hour Day .................................. 48
7. Summary of Analyses of Variance for Data for
which Group-by-Day Means are in Table 6...... 49
8. Circadian Rhythm: Daytime and Nighttime TS
and Day/Night Ratio for Baseline and Days
after Deprivation Procedures A and P......... 50
9. Circadian Rhythm: Increase in Percent Sleep
over Baseline after Deprivation Procedures
A and P...................................... 51
10. Sleep Episodes in Terms of Frequency of
Occurrence of Specified Length of Episodes... 54
11. Overall Analyses of Variance for TS, SWS and
PS per 24-Hour Day for which Data are in
Tables 12, 13 and 14......................... 65
12. SWS Group Means in Percent Pre and Post
Deprivation (per 24-Hour Day) ................ 66
13. PS Group Means in Percent Pre and Post
Deprivation (per 24-Hour Day)................ 67
14. TS Group Means in Percent Pre and Post
Deprivation (per 24-Hour Day)................ 68
15. Percent of Lost SWS, PS and TS Recovered
During Days 1 to 9 per Individual Subject
and Group.... ............... ................ 73
16. Circadian Rhythm: Daytime and Nighttime
TS and Day/Night Ratios for Groups A and P.... 78
17. Circadian Rhythm: Increase in Percent Sleep
over Mean of Two Baseline Measures for
Daytime and Nighttime Hours after Deprivation
Procedures A and P................. .... ...... 79
18. Frequency of Occurrence of Minute Sleep
Episodes Grouped in Three Class Intervals..... 82
LIST OF FIGURES
1. Mean number of shocks/h of deprivation for
three levels of shock intensity............... 12
2. Intershock interval for S A-1.................. 19
3. Percent weight loss as a function of days
of deprivation................................ 30
4. Sleep percent despite deprivation
procedures as a function of days of
5. Sleep percent despite deprivation as a
function of time of day during deprivation.... 32
6. Midnight and noon sleep latencies for
two Ss as a function of days of deprivation... 34
7. Slow-wave sleep group means pre and post
deprivation per 24-hour day................... 70
8. Paradoxical sleep group means pre and post
deprivation per 24-hour day................... 71
9. Total sleep group means pre and post
deprivation per 24-hour day................... 72
Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
STUDIES IN THE TOTAL
SLEEP DEPRIVATION OF RATS
Hyman S. Sternthal
Chairman: Dr. Wilse B. Webb.
Major Department: Psychology.
Two deprivation procedures were used to keep rat Ss
awake. The first, Procedure A, consisted of delivering
shock to the feet contingent upon the appearance of SWS
in the EEG. The effect of shock intensity on the number
of shocks/h required to keep Ss awake was inconclusive.
The number of shocks/h was an-increasing function of the
time or duration of deprivation. After about 15 to 20 h
of deprivation, there was a relatively sharp increase in
the number of shocks/h reflecting sleep override of the
shock. That is, Ss obtained substantial amounts of sleep
by falling asleep about 20 s after each shock of 10 con-
The other deprivation method, Procedure P, consisted
of presenting a wide variety of changing stimuli, ex-
cluding shock, that were not contingent upon the appearance
of EEG SWS. When the assistants delivering the stimuli
and monitoring the EEG made a maximal effort, the procedure
resulted in Ss being awake up to 99% of the time. As dep-
rivation proceeded, it was found that the latency of sleep
onset, when Ss were placed in a cage and briefly left un-
disturbed, decreased during the first three days to about
20 s. Also, Ss' muscle tonus decreased greatly with time or
duration of deprivation.
With deprivation Procedure A, the limit or terminal
point was sleep override of shock. Ss fell asleep about
20 s after being shocked and were obtaining substantial
amounts of sleep after 15 to 20 h of deprivation. With
deprivation Procedure P, the terminal point seemed deter-
mined by lowered arousal which made responses necessary for
Ss' survival impossible. One S developed a pulmonary prob-
lem and one showed a disinterest in food and a sharp weight
loss. The deprivation procedure was still keeping both Ss
With deprivation Procedure A, it seemed possible to
keep Ss awake for a maximum of about 40 h. With Procedure
P, Ss could be kept awake for about 12 to 15 days.
It was tentatively concluded that Procedure A was far
less effective than Procedure P for two reasons. First,
intense pain is a stimulus for sleep even when Ss are not
sleep-deprived. Secondly, Ss may have habituated to shock
in Procedure A, but could not habituate to a wide variety
of stimuli which were changed when Ss seemed adapted to any
one, as in Procedure P.
When Ss were kept awake for a 24-h period, it was
found that Procedure A reduced the recovery of lost sleep
from about 74% with Procedure P, to about 31%.
After the 24-h deprivation period, more SWS was re-
covered than PS, 84.6 to 54.0% with Procedure P and 33.0
to 21.3% with Procedure A.
Recovery of the sleep lost with 24 h of deprivation
involved a shift toward more long sleep episodes and fewer
short and medium-length episodes. This shift was greater
with Procedure P.
Lost sleep was recovered by an increase in both the
daytime and nighttime sleep percentages. Generally, the
increases in daytime and nighttime sleep percent were not
proportional to the respective baseline amounts which re-
sulted in changes in the day/night ratio, which was a mea-
sure of circadian rhythm. The only consistency in the
changes in the day/night ratio seemed to be that groups that
were initially different in the baseline day/night ratio be-
came more alike after deprivation.
The purpose of this dissertation was to explore inten-
sively four .separate aspects of sleep deprivation. The se-
lection of the variables resulted from an intensive review
of the literature which revealed that these problems asso-
ciated with sleep deprivation had been quite neglected.
The selected variables were (a) the temporal course of
sleep demand resulting from sleep deprivation, i.e., the
intensity of the sleep need as a function of time of depri-
vation, (b) the temporal limit of total sleep deprivation
beyond which wakefulness cannot be maintained, (c) the con-
sequences of sleep deprivation on the subsequent sleep re-
sponse, and (d) the effect of methods of sleep deprivation,
particularly the effect of aversive stimulation,,used to
Each of these variables seemed critical in an under-
standing of sleep deprivation. The discussion of each ex-
periment will attempt to explicate this reasoning. The de-
cision was made to explore all four variables rather than
to intensively concentrate on any one. As a consequence,
the findings are recognizably incomplete, but do yield suf-
ficiently provocative findings to warrant attention and
Male Long-Evans hooded rats were used as Ss. Animals
were used after they had fully recovered from surgery
(about one month) and had regained all their lost weight
or more. Each S was about 150 days old at the beginning
of the experiment in which it participated.
Swisher (1961) studied the EEG sleep stages of the
rat. The recording technique used was adapted from his
work after a number of trials. The initial trials made it
clear that an EEG recording method was available in which
there was high discriminability between the EEG sleep stages
as commonly defined (e.g., see Swisher, 1961, or Levitt,
1965). Also, the EEG and sleep-wake behavior were well
correlated, as Swisher found.
The Ss were anaesthetized with Nembutal and given anti-
biotics to prevent infection. Using a stereotaxic instru-
ment, four points were marked on the skull. These were
(a) over the left motor cortex (1 mm to the left of bregma),
(b) over the right frontal cortex (1 mm to the right and 5
mm anterior to bregma), (c) over the left sensory cortex
(4 mm to the left and 5 mm posterior to bregma), and
(d) over the right sensory cortex (4 mm to the right and 5
mm posterior to bregma). A small stainless steel screw
was forced into a small hole drilled at each point on the
skull, so that the screw was in contact with the unpunctured
dura. One stainless steel wire, from a four-contact recep-
tacle to be implanted, was then wrapped around each of the
four screws. Dental acrylic cement was used to mold the
receptacle wires and screws into one unit. The receptacle
held four cylindrical female contacts in a square pattern.
A plug which held four wire-brush contacts in a square
pattern was joined to the receptacle for recording. A
four-wire shielded cable, easily flexed by Ss, connected
the plug to a Grass Instruments Model 3D or Model 4 elec-
The Ss were trained to avoid touching the cables so
that they would not chew through them. This was done by
arranging a 45 v difference between the cable's shield and
the bottom grid of the cages for a 24-h training period.
If an S chewed through a wire, this was repeated as nec-
EEG records and scoring
Two EEG channels were used for each S. The upper
channel was a recording of the voltage difference between
the dural screws on the left side of the head, one over
the motor and the other over the sensory cortex. The
lower channel recorded the voltage difference between the
right frontal and right sensory screws. During wakefulness
(W) both channels had a high frequency pattern. The excur-
sions of the pen, relative to the paper speed (10 cm/s),
were sufficiently frequent that the written lines over-
lapped and the individual movements of the pen could not
be discriminated. This defines W as used herein (Swisher,
1961). At the onset of slow-wave sleep (SWS) excursions
of the pen gradually became less frequent and the amplitude
of the excursions increased so that individual excursions
of the pen were discriminable. This is the definition of
SWS as used herein. Paradoxical sleep (PS) or activated
sleep onset occurred after Ss had obtained some SWS. It
was marked by the reversion of the upper channel from an
SWS pattern to a W pattern, in some Ss intermingled with a
theta pattern. The lower channel then changed from SWS to
a regular theta pattern. This is the definition of acti-
vated sleep or PS as used herein when scoring sleep in un-
disturbed Ss (Swisher, 1961).
The main purpose in scoring the sleep records of un-
disturbed Ss was to determine how much lost sleep was re-
covered. Thus, in scoring each minute of EEG record, it
was first decided whether there were 31 or more s of W
EEG, and if so, the minute was scored as W. If the epi-
sode had 29 s or less of W, the minute was scored as either
PS or SWS, depending upon which of the two sleep stages
About five estimates of E's rescoring reliability were
made using 100-min samples. It was found that E was highly
consistent with himself, as there was between 95 to 98%.
agreement between the original and the rescoring.
Deprivation Procedure A consisted of delivering shock
to Ss upon the appearance of slow waves in the EEG record.
Shock was delivered from a square wave or constant current
shock generator. Duration and intensity of the current
could be varied. Because of Ss' skin resistance, the most
intense constant current shock deliverable was 1.7 ma. The
shock was delivered to Ss through grids placed on the floors
of Ss' cages. Alternate bars of the grid were oppositely
charged during delivery.
Subjects were kept awake by students who could dis-
criminate between EEG SWS and EEG W. It was emphasized
that one cannot know whether Ss are awake or asleep by the
fact that their eyes are open. Assistants were told to
keep Ss awake by (a) using enforced movement, (b) placing
Ss in novel environments, and (c) anything they could in-
vent that would not harm Ss by endangering their health.
Assistants were instructed to try to keep Ss awake and pre-
vent them from falling asleep rather than to frequently re-
awaken Ss. After deprivation began, assistants learned to
anticipate when Ss would fall asleep. The main cue was that
Ss would slow down and stop moving first. Assistants were
also instructed to keep food always in the presence of Ss
and to offer them water frequently.
The final form that Procedure P took was highly
variable and depended largely upon the individual assis-
tant. Some assistants placed the Ss in a shallow pan and
tilted it frequently so that Ss were forced to move.
Others placed the Ss at the bottom of a bucket and shook
them frequently. Others sought incentives that Ss would
work for. One incentive was to place Ss in shallow water
which resulted in escape attempts. Other assistants used
a soft area such as a lab coat as an incentive. The Ss
attempted to sleep on the coat and returned many times
when removed from it. When an S began to fall asleep, its
environment was changed or a new means of maintaining
wakefulness was found.
Comparison of procedures
(a) While Procedure A used shock as its sole stimulus
source, Procedure P did not use shock at all.
(b) While Procedure A used only one stimulus, Procedure
P used a wide variety of stimuli.
(c) In Procedure A, shock delivery was contingent upon
the appearance of SWS in the EEG. In Procedure P, stimuli
were delivered so as to prevent Ss from falling asleep. Al-
though at times Ss were awakened, for the most part, stim-
uli were not contingent upon SWS.
(d) The two procedures were approximately equally ef-
fective for the first 24 h of deprivation in that, while Ss
were being successfully kept awake, both procedures kept Ss
awake about 97% of the time.
Webb (1957) studied the relationship between the
amount of time of sleep deprivation and latency of going
to sleep. Subjects were kept awake by being placed on a
slowly revolving wheel that was two-thirds submerged in
water. They were allowed to fall asleep in observation
cages. There was a distinct relationship between sleep
deprivation in hours and sleep latency. Webb found that
when Ss were taken off the wheel even after 30 h of dep-
rivation, before they fell asleep they went through groom-
ing and exploratory activity, which took an average of
over 10 min.
The purpose of this experiment was the same as Webb's
(1957) study, i.e., to investigate the temporal course of
the intensity of the sleep demand as a function of the
amount of time of sleep deprivation. An analog of Warden's
obstruction or barrier technique was used. In the classical
Columbia-obstruction-box studies, rats ran across an elec-
trified grid to get to food or some other goal object (see'
Kimble, 1961, p. 454). In this experiment, rats were
shocked whenever they were found to be asleep and shocks/h
was used as the measure of intensity of the sleep demand
rather than sleep latency. Exploration of a novel en-
vironment was not a factor, since Ss were always kept in
the same cages.
Webb (1957) visually monitored Ss and used cessation
of movement and relaxation of muscle tonus as the criterion
of sleep. The following evidence leads to the conclusion
that such monitoring Is inadequate. Boren (1960) noted
that monkeys could appear to be asleep with eyes closed,
hunched into typical sleep posture and still maintain a
steady rate of avoidance responding. Levitt (1967) found
that short bursts of sleep-like waves, which he called
microsleep, were found in the EEG of rats on a slowly re-
volving treadmill. The Ss learned to sleep as much as
14 mln/h within 32 h by moving to the front of the wheel
and remaining stationary while riding to the rear. Be-
cause Ss obtained slow-wave sleep on the treadmill, both
Levitt (1967) and Dement, Henry, Cohen and Ferguson (1967)
used treadmills to deprive Ss selectively of paradoxical
Preliminary experiments for this research study at-
tempted to keep Ss awake by observing them and moving
them when it seemed necessary. However, it appeared that
Ss may have been asleep with their eyes open, since at
times they seemed unresponsive to visual stimuli such as
E's finger near Ss' eyes. Recording of EEG during pre-
liminary attempts at sleep deprivation revealed that at
times, while Ss stood on all four legs and appeared to be
staring at E, EEG SWS was simultaneously present. Thus in
this experiment, Ss' state of sleep and wake was monitored
by means of the EEG during sleep deprivation.
In order to use the classical obstruction procedure
to study sleep in the rat, Ss had previously been prepared
for EEG recording and shock was delivered to the feet when-
ever EEG signs of sleep appeared.
The data resulting from this procedure were used to
answer the following questions: (a) What is the relation-
ship of amount of sleep deprivation measured to the num-
ber of shocks per unit time presented to maintain wakeful-
ness? (b) Does the pattern of the length of the intershock
interval or interval between instances of Ss' manifesting
EEG sleep signs reflect either (1) a tendency to learn to
stay awake or (2) a tendency to learn to sleep despite
shock or (3) a random or other pattern? (c) What is the
effect of shock intensity?
Five as participated in each of three groups which
differed in shock parameters and the number of hours for
which each group was successfully kept awake by means of
Procedure A. Deprivation was always started at 9:00 A.M.
which was hour i.
1) The shock intensity was 0.6 ma, and each shock
lasted 0.5 s.
2) Shock intensity was constant.
3) Deprivation proceeded until sleep "overrode" the
shock. Override was defined as having occurred when 10
consecutive shocks were delivered with less than 20 s
intervening between shocks, i.e., when Ss were obtaining
substantial amounts of sleep.
4) All Ss were effectively deprived for 19 h.
1) Initial shock intensity was 1.0 ma for 0.5 s.
This shock level made most Ss vocalize and jump off the
2) After 15 h, one S was sleeping despite the shock
and had achieved override. Since the goal was sleep dep-
rivation for 24 h, shock intensity and duration were
gradually increased to the limit of the generator which
was 1.0 s and 1.7 ma.
3) After 20 h of deprivation, shock was not adequate
in keeping this S awake. Thus, data for this group are
presented for 20 h of deprivation.
1) Group A was used in another experiment, where
sleep was recorded for two days prior to and 11 days after
deprivation by Procedure A. The Ss were to be kept awake
for 24 h as part of the design of that experiment.
2) The shock level for Group A was set at the
reliable maximum of the shock generator (1.7 ma), The
duration of shock delivery was 1.0 s. This shock resulted
in extremely vigorous escape attempts by Ss.
3) The shock level was not varied since the shock was
sufficient to keep Ss awake for 24 h.
Results and Discussion
As may be seen in Figure 1, the number of shocks/h is
an increasing function of the amount of time of deprivation.
Table 1 presents the ungrouped data upon which Figure 1 is
based. Examination of these data suggests that the form of
the curve in Figure 1 is not an artifact of grouping data
from individual Ss. Table 2 presents the analysis of vari-
ance based on the first 19 h of deprivation data for all
three groups, as in Table 1. As may be seen in Table 2,
amount of time of deprivation had a highly significant
effect upon the number of shocks/h.
Table 3 presents the override data. Five Ss were kept
awake for between 19 and 26 h, while two Ss were kept awake
for only 13 and 15 h respectively.
According to Murray (1965, p. 210), the effect of sleep
deprivation is to lower the level of arousal so that sleep-
deprived Ss are in a state between sleep and wakefulness.
The level of arousal was so lowered as a result of sleep
deprivation that eventually either (a) shock did not result
in enough arousal to awaken Ss or (b) shock resulted in
enough arousal to awaken Ss for only a few seconds.
It was found in Experiment II that Ss could be kept
awake for much longer periods of time with Procedure P.
54 -- Shock-S, 0.6 ma, 0.5 s
D.--------- Shock-W, 1.0 ma, 0.5 s
48 o-----o Group A, 1.7 ma, 1.0 s
12 .' ''d
-] I I I i I II I 1 I I I I I I I I i I I l ll l
2 4 6 8 10 12 14 16 18 20 22 24
HOURS OF DEPRIVATION
Figure 1. Mean number of shocks/h of deprivation for three
levels of shock intensity.
Shocks per Hour per Individual S
in.Shock-S, Shock-W, Group A
and Total per S after 19 Hours of Deprivation
S-1 S-2 S-3 S-9 S-10
W-2 W-7 W-9 W-11 W-12
A-i A-4 A-5 A-8 A-9
Total 110 470 306 248 671 190 74 648 75 53 329 70 204 213 58
Total 110 470 306 248 671
190 74 648 75
53 329 70 -204 213 58
TABLE 1 Continued
Hour Shock-S Shock-W Group A
S-1 S-2 S-3 S-9 S-10 W-2 W-7 W-9 W-ll W-12 A-1 A-4 A-5 A-8 A-9
20 31 97 9 24 3 4 58 7 14 33 6
21 27 36 9 19 44. 15
22 8 56 16 24 54 11
23 16 108 18 43 82 23
24 21 86 18 50 68 22
Analysis of Variance Based on
First 19 Hours of Deprivation for all Three Groups
for Data in Table 1
Source df F p
Between Subjects 14
Shock Level 2 1.21
error (b) 12
Within Subjects 270
Hours of Deprivation 18 7.83 <0001
Hours of Deprivation
by Shock Level 36 0.62
error (w) 216
Hours of Deprivation by Procedure A
for Individual Ss till Override of Shock
Whose Intensity and Duration are Specified
S-1 26 .6 ma, 0.5 s-
S-2 19 .6 ma, 0.5 s
S-3 19 .6 ma, 0.5 s
S-9 .6 ma, 0.5 s
S-10 13 .6 ma, 0.5 s
Note.--Where no number of hours is given,
the S had not overridden the shock when shock
delivery was stopped.
It was also found with one S that after 40 h of depriva-
tion by Procedure A, shock was ineffective, while strok-
ing its fur was very effective in maintaining wakefulness.
It was concluded that Procedure P was much more effective
as a means of long-term deprivation because of the use of
multiple types of stimulation and because intense shock may
have a de-arousal effect. Thus it is not possible to say
how much the data in Figure 1 reflect the effect of accumu-
lated hours of sleep deprivation versus the de-arousal ef-
fect of intense shock and habituation to shock.
Effect of shock intensity
As may be seen in Figure 1, the more intense the shock,
the fewer were delivered to each group. As may be seen in
Table 2, neither the differences in the total number of
shocks for 19 h between groups, nor the hours by group in-
teraction effect was significant.
As may be seen in Table 3, there is a great deal of
variance between Ss within groups in terms of the total
number of shocks required to keep an S awake. Since lack
of statistical significance may be due to error variance,
no conclusion can presently be reached regarding the effect
of shock intensity upon the number of shocks/h required to
keep Ss awake.
Pattern of intershock intervals
Figure 2 displays the intershock intervals of a single
S. An important question regarding these data is the
source of the variability between intershock intervals.
One possibility is that Ss may learn and relearn to stay
awake a number of times as deprivation proceeds. Thus if
each intershock interval is labeled '+' or '-' depending
upon whether the interval is longer or shorter than the
immediately preceding interval respectively, then one
would expect series of intervals labeled 1'+ to occur more
often than chance, if Ss learn and relearn to stay awake
as a function of shock. If there is a repeated extinction
process, i.e., if Ss extinguish the learned 'wakefulness'
response, or if the sleep demand consistently increases
from shock to shock, one might expect series of intershock
intervals labeled '-I to occur more often than by chance.
To see if such learning and/or extinction sequences oc-
curred more often than chance, sequences of four of the
signs were written out for the initial series of inter-
shock intervals. Thus the first sequence used the 2nd, 3rd,
4th and 5th signs, the second sequence used the 3rd, 4th,
5th and 6th, etc. The initial series of intershock inter-
vals where there was maximum variability (e.g., the first
44 shocks for S A-1, Figure 2) was used on the assumption
that the effect of the sleep drive would be less and learn-
ing or extinction effects would be more dominant. (After
about the first 50 shocks, the intershock intervals varied
little.) Group A data were used.
2 6 10 14 18 22 26 30 34 38 42 46 50 54 58 62
Figure 2. Intershock interval for S A-1.
In Table 4, all possible sequences of pluses and
minuses were written out. Any sequence of three or four
consecutive plus signs (++++, +++-, -+++) is considered
to be evidence of. learning of shock avoidance. Any se-
quence of three' or four minus signs (----, ---+, +---)
is considered to be evidence of extinction of learned
shock avoidance or of the effect of the increasing sleep
need. The sequences +-+- and -+-+ are considered to re-
flect an oscillating tendency, .i.e., for a longer inter-
shock interval to be followed by a shorter interval.
Other combinations of plus and minus signs are considered
to represent chance and uninterpretable events.
In Table 4, there are 16 permutations of plus and
minus signs and 261 events. Thus one would expect that on
the basis of chance only, each permutation should have had
16.3 (261 16) as the expected frequency. The data in
Table 4 suggest that an oscillation between shorter and
longer shocks was the predominant tendency and these (+-+-,
-+-+) occurred about twice as frequently as would have been
expected by chance. On the other hand, probability of in-
tervals becoming continuously longer (++++) or continuously
shorter (----, ---+, +---) for the part of the data show-
ing maximum variation, occurred about one fourth as often
as one would have expected on a chance basis. Thus, no
evidence of a tendency for Ss to learn, extinguish and re-
learn to stay awake was found.
Frequency of Occurrence
of Each Permutation
of Intershock Intervals for Group A
Sequence of Intershock
The intensity of the sleep demand was studied as a
function of time of deprivation using a procedure analo-
gous to Warden's obstruction or barrier technique. The
Ss were shocked whenever they fell asleep. Sleep was
operationally defined as occurring when SWS appeared on
the EEG, as a review of the literature indicated that other
methods such as visual monitoring of Ss were unreliable.
Using three groups of five Ss each, it was found that
the number of shocks/h required to maintain wakefulness was
an increasing function of the time of deprivation. After
13 to 26 h, the tendency to fall asleep became sufficiently
strong that Ss overrode the shock, i.e., fell asleep within
20 s after being shocked 10 times consecutively, and ob-
tained substantial amounts of sleep despite the deprivation
Since the differences between the three shock level
groups were not statistically significant, no conclusions
about shock intensity could be reached.
The pattern of intershock intervals was analyzed for
one group of Ss. No evidence of a tendency for Ss to
learn, extinguish and relearn to stay awake was found.
Many investigators have tried to keep organisms awake
to the point where it became impossible to keep them awake
any longer. Table 5 summarizes those studies in which Ss
were deprived of sleep to their limit. As may be seen in
Table 5, various animals, with the use of various tech-
niques of sleep deprivation, have been kept awake from one
up to 77 days. Some of the variability in the limit of
deprivation is probably the result of the inadequate means
of monitoring Ss to determine sleep or wakefulness, as was
discussed in Experiment I (see pp. 8-9 above). Presumably,
Ss were only partially sleep-deprived when the EEG was not
monitored. It is assumed that the more complete the dep-
rivation, the shorter the deprivation limit found.
Another factor which may have contributed to the vari-
ability in the limit of deprivation concerns the sleep dep-
rivation procedures. As may be seen in Table 5, most pro-
cedures are much like Procedure A in this study. Typically,
some presumably arousing stimulus situation is presented or
results when Ss fall asleep. In the revolving treadmill
studies, for example, Ss fall into water (Webb and Agnew,
1962). Pegram (1968) kept his monkey Ss awake by adminis-
Summary of the Literature on the
Limit of Sleep Deprivation
Organism Method Reason for
Bast et al. (1927)
and Senn (1927)
from all sides
of special cage
Death or com-
Death or com-
Death due to
9, 13 and
TABLE 5 Continued
Organism Method Reason for
Webb and Rats Treadmill Exhaustion or 1-27+ days
Agnew (1962)- termination of
Anderson and Humans -Set new record 102 days
Gulevich Humans -Set new record 11 days
et al. (1966)
tering sharp puffs of cool, compressed air to the back of
the neck and the side of the head when there were behavioral
or EEG signs of sleep. On the other hand, when the proce-
dures used to keep human Ss awake are specified, these ap-
pear to be much like Procedure P in which a variety of
non-contingent stimuli are presented to maintain wakeful-
ness. Thus Berger and Oswald (1962, p. 458) stimulated
their Ss to keep awake by playing a variety of games. They
considered the repeated introduction of variety into the
activity essential for the maintenance of wakefulness.
The original goal of this experiment was to determine
how long rat Ss could be kept awake before it became im-
possible to keep them awake any longer, using deprivation
Procedure A to maintain wakefulness and the EEG to monitor
it. One S was kept awake until it fell asleep a few seconds
after being shocked. It was then found that although the S
could no longer be kept awake by shock, it was easily kept
awake by gently stroking its fur. It was apparent that
Procedure A was not the method of choice, and it was discon-
tinued as a long-term method of deprivation. The data from
this one S and relevant data from Experiment I will be used
to compare the effectiveness of the two deprivation proce-
dures. It is believed that this is worthwhile because experi-
menters tend to use procedures much like Procedure A with
animal Ss when there is some evidence that a method like
Procedure P is more effective.
One male Long-Evans hooded rat, prepared for EEG re-
cording, was kept awake by Procedure A. Initial shock amp-
litude was 0.5 ma of 0.5 s duration. Amplitude and dura-
tion were gradually increased as necessary. In the last
hours of the deprivation period, a very intense shock,
which made S jump and squeal vigorously, was used.
Three male Long-Evans hooded rats were kept awake by
Procedure P. Continuous EEG recordings were taken for all
Ss. The last two Ss deprived by Procedure P were weighed
every 12 h. Immediately after each weighing, Ss were
placed in cages, and left undisturbed till asleep, allowing
a measure of sleep latency.
The EEG recordings of the last two Ss deprived by
Procedure P were used to determine how effective the dep-
rivation procedure had been. (The EEG data for the first S
could not be reliably scored and are not presented.) During
deprivation, sleep tended to appear in short episodes of
about 10 s every few minutes. The procedure for scoring re-
cords of undisturbed Ss does not consider sleep to have oc-
curred unless 31 or more s of a minute was SWS, and was
therefore inappropriate here. Instead, a 10-min sample was
taken from every hour of deprivation and the number of sec-
onds of sleep was counted and recorded. These data were
used to extrapolate an estimate of the total amount of sleep
obtained during deprivation. These data were examined also
to see if Ss were more.likely to fall asleep as a function
of days of deprivation and of the time of day.
Long-Term Deprivation by Procedure A
For the one S deprived to the limit by deprivation
Procedure A, it was found that after about 40 h, S's feet
were extremely reddened and bleeding, both from the shock
and because S attacked and chewed its own feet. Also, at
that time, even though the shock made S jump and squeal
vigorously, SWS was manifest on the EEG within about 10 to
15 s after shock delivery and S was receiving a substantial
amount of SWS. It was then found that stroking S's fur
gently was adequate to maintain wakefulness, while shock
Long-Term Deprivation by Procedure P
The attempt to keep the first S awake by Procedure P
succeeded in keeping S awake for six days. The experiment
was terminated because a loose stylus holder on the EEG
machine had given the appearance of persistent slow waves
in one channel.
The deprivation procedure for the last two Ss was
started at 8:40 A.M. The first S was terminated when death
due to pneumonia seemed imminent, since S was struggling to
breathe. Unfortunately, assistants had kept this S wet as
a method of maintaining wakefulness. This S seemed no
harder to keep awake in the last hours than a few days
before. This S was kept awake 97.3% of the time for 296 h
or 12 days and 8 h. It obtained 7.8 h of sleep during the
period of enforced wakefulness. Rats in this S's population
sleep 51.6% when undisturbed. Thus, S would have obtained
152.7 h of sleep in 296 h (the duration ofthe period of
enforced wakefulness) if it had been left undisturbed.
Since S slept only 7.8 h in that period, it had been de-
prived of, or had a sleep debt of, 144.9 h. This debt is
equivalent to the amount of sleep S would have obtained if
it had been undisturbed for 11 days and 17 h.
The second S was terminated when, refusing to eat and
drink, it suffered a sudden weight loss (see Figure 3).
This S was kept awake 94.6% of the time for 365 h or 15
days and 5 h. It slept 5.4% or 19.9 h during the period of
enforced wakefulness. It is assumed that this S'would have
obtained 188.5 h of sleep (51.6%) in that time if left un-
disturbed. Since S slept 19.9 h, it had a sleep debt of
168.6 h at the end of the deprivation period. This is the
amount of sleep that a rat S would obtain in 13 days and
14.4 h if undisturbed.
The amount of sleep per day per S was examined to see
if Ss slept more as a function of days of deprivation.
Figure 4 presents these data. One possibility suggested by
the data is that after about eight or nine days of depriva-
tion, Ss obtained much more sleep. However, these data are
confounded by an experimenter variable. As may be seen in
Figure 5, Ss obtained more sleep during the night hours
(which is opposite to the normal pattern for rats).
Correspondingly, E tended to supervise least from midnight
a 20 -
14 -- n----- 12-day S 0
So---o 15-day S
4 1 I I I I I I I I I I I I I I-
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
DAY OF DEPRIVATION
Figure 3. Percent weight loss as a function of days of
deprivation. (Midnight weighing have day numbers
o---o 12-day S
f- 15-day S
2 4 6 8 10 12 14
DAY OF DEPRIVATION
Figure 4. Sleep percent despite deprivation procedures as a
function of days of deprivation.
o---o 12-day S
0---o 15-day S
8 AM-12 PM 4 PM-8 PM 12 AM-4 AM
TIME OF DAY
Figure 5. Sleep percent despite deprivation as a function
of time of day during deprivation.
to 8:00 A.M. Also, for the last three days of deprivation
when only one S was yet to be terminated, E, who was aware
of the amount of sleep Ss were obtaining during deprivation,
increased supervision of assistants and exhorted them to
monitor the EEG and provide continuous stimulation to Ss.
Thus, that S which had 14% sleep on deprivation day 12 had
only 2% sleep on day 15 (see Figure 4).
Figure 6 summarizes the latency data, i.e., the num-
ber of seconds between Ss being placed in cages and the
first SWS episode. There was no systematic difference be-
tween the midnight and noon latencies.
Two factors appear to account for the latency measure
changes as seen in Figure 6. The first factor appears to
be a systematic tendency for the latencies to decrease
over the first three days to a minimum range which is be-
tween 5 to 20 s. In addition, there seems to be a random
tendency to have sleep latencies much longer than about
20 s. One contributor to this random factor was the fact
that Ss were often wet when put into the cages and Ss would
groom, or partially dry themselves, or eat and drink food
and water that was in their cages before falling asleep.
The following effects of the long-term deprivation by
Procedure P were apparent with non-quantifiable observa-
tions. Asdeprivation proceeded, Ss' muscular tonus de-
creased greatly so that when an S was lifted by the shoul-
ders, it tended to hang limply as if anaesthetized. The Ss
tended to move more slowly and by the end of deprivation
---o 12-day S
0---o 15-day S
o11 1 1 1 11 1 I L I
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
DAY OF DEPRIVATION
Figure 6. Midnight and noon sleep latencies for two Ss as a function of
days of deprivation. (Midnight latencies have day numbers
tended not to walk at all. They jumped off table tops, did
not right themselves and hit the ground snout first, cutting
their lips. (Since Ss were kept walking far more than usual,
the bottoms of their feet began to bleed; Bandaids were put
on their feet, making small boots.) Also, Ss showed no
apparent signs of increased aggressiveness.
The EEG had not been calibrated for testing the amp-
litude and form of SWS during deprivation. However, in-
spection of a number of samples of SWS episodes that oc-
curred during deprivation revealed no EEG differences as a
function of time of deprivation.
The Limit of Deprivation
A deprivation procedure, Procedure P, was found which
kept two Ss awake about 95 to 97% of the time during dep-
rivation for about 12 to 15 days. A sleep debt was calcu-
lated, taking into account both the,amount of sleep that
rats obtain when undisturbed and the amount of sleep that
Ss obtained despite deprivation. It was found that rat Ss
can be kept awake till they accumulated sleep debts, one of
144.9 h, the other 168.6 h. These debts are equivalent to
the amount of sleep that rat Ss would obtain in about 12 to
14 days of undisturbed sleep. Comparable data with a state-
ment of sleep debt are not available elsewhere for compar-
ison. Human Ss, however, have been kept awake for up to
11 days and presumably incurred a sleep debt of about 88 h.
Effect of Long-Term Deprivation
The attempt to keep an S awake indefinitely by depriva-
tion Procedure A resulted in the S sleeping despite being
shocked. In contrast, the results from the two Ss subjected
to deprivation Procedure P suggested that the terminal point
is likely to be determined by physiological variables other
than sleep. One S developed a pulmonary condition; the
other showed a sharp weight loss. The weight loss is in-
terpreted as having been due to the lowering of arousal,
which was a consequence of sleep deprivation. For any motor
performance to occur, a certain level of arousal is essen-
tial. Murray (1965, p. 210) has concluded that sleep depri-
vation leads to a lowered state of arousal. This was ob-
served by E in this experiment. Ss' muscular tonus dropped
greatly and Ss became quite inactive. Eventually, the
arousal level was so low that eating and drinking were not
One seems forced to conclude, based on the above data,
that while powerful mechanisms exist to lower Sst level of
arousal with time of deprivation, sleep deprivation per se
did not result in death. Rather, death would have resulted
secondarily from the lowered level of arousal and resulting
starvation or other physiological concomitants.
Effectiveness of Procedure A versus Procedure P
In Experiment I, it was found that deprivation Pro-
cedure A succeeded in preventing override (falling asleep
20 s after being shocked for 10 consecutive shocks) up to
26 h. The most intense shock used was 1.7 ma for 1.0 s. In
this experiment, with only one S, shock intensity was mini-
mal at first and increased only as needed. Also, S was kept
awake beyond the override criterion. (The Ss frequently a-
chieved override and afterward began responding to the shock
by staying awake for periods longer than 20 s.) This S was
shocked when SWS appeared, until S was asleep within about
7 to 10 s after being shocked most of the time. Since, it
took E a few seconds to determine EEG SWS, S was obtaining
about 10 or more s/min of sleep. (The EEG was obliterated
for about 5 s after a shock was delivered. With this S, SWS
appeared as the EEG record returned to normal at the time of
termination.) This S initially achieved override at about
30 h. At this point, it may be tentatively concluded that
increasing shock intensity to a level where S attacks its own
feet merely forestalls, not only override as defined herein,
but also S's falling asleep within about 5 s after being
shocked. In sharp contrast to this is the fact that one S
was kept awake for six days and another two Ss were kept a-
wake for about 12 and 15 days using deprivation Procedure P.
The most important finding of this study was the wide
difference between the two limits of deprivation found using
the two different procedures. Interpretation of this differ-
ence, however, must be tentative because the two procedures
differed in a number of ways as was outlined above (see the
comparison of procedures on p. 6).
Pavlov (1927) believed that "the fundamental condi-
tion of the appearance and development of internal in-
hibition and sleep is exactly the same" (p. 251).
One means by which the state of inhibition
could be produced was by exposing the dog to
certain stimuli which would have evoked a re-
sponse had they not been excessively intense--
"such conditioned stimuli too strong to give
the maximal conditioned reflex, Pavlov termed
transmarginal or supramaximal." (W. H. Gantt in
his introduction to his translation of Pavlov,
1941, p. 14.) Oswald, 1960, p. 1450.
In the sense that shock (used in this experiment, however,
as a UCS) is a very intense stimulus, it may be considered
to be transmarginal or supramaximal, and thus a producer
of inhibition or sleep.
Oswald (1962) has concluded that sleep or some state
having many characteristics of sleep appears in response
to five situations, only two of which are relevant here and
are presented below.
First, Oswald has concluded:
Monotonous repetition of stimulation
brings about decreased responsiveness or
habituation.... We can think merely in terms
of stimuli entering the reticular formation
via collateral afferents over and over again,
with consequent habituation of the usual re-
sponse of arousal. As reticular formation
responsiveness declined, sleep would tend to
appear. P. 160.
In this experiment, animals kept awake by.Procedure A
could be described as receiving monotonous stimulation.
However, deprivation Procedure P was deliberately designed
to avoid monotony. Sources of stimulation included
(a) cooling, (b) vestibular stimulation, (c) all forms
of tactile stimulation (excluding pain), (d) various new
odors from different foods, assistants and rooms, (e) a
wide variety of sounds, and (f) a wide variety of visual
stimulation. Thus, the mechanism of habituation to any
one source of stimulation was circumvented, since the
sources of stimulus input were modified as frequently as
was found to be necessary.
Secondly, Oswald has concluded that sleep appears in
response to "a continued situation which may be interpreted
as overwhelmingly terrifying" (p. 160). The following data,
which Oswald apparently used to infer his conclusions, will
make it clear that deprivation Procedure A is very similar
to situations which Oswald considered to be overwhelmingly
terrifying, particularly since many involved shock.
An inert, stuporose state is found in some
infants immediately following, for instance,
the sudden pain of the dislocation of a joint.
Burton and Derbyshire (1958) reported also the
case of a one-year-old boy with acute secondary
After an hour of frenzied screaming at
the onset, he stopped abruptly and fell asleep
and on examination was found to be unrousable.
He remained so until, after seven days, the
eye was enucleated, whereupon he became alert
and responsive within a few hours. P. 154.
Oswald also cites Jouvet and Hernandez-Peon's (1957)
finding, with one cat, that
if stimulus combinations of a tone and an elec-
tric shock were given "at a high rate" (three
or four per minute), and if the shock was made
strong, EEG signs of sleep were produced.
Pavlov (1955) described the appearance of
sleep in dogs subjected to regular electric
Liddell (1956) has described experiments in which
most profound states of inert unresponsiveness
occurred in lambs and kids subjected to repeated
mild electric shocks (provided they were deprived
of their mothers' company), and these had full
freedom of movement within a room. P. 149.
In experiments with six human volunteers
(Oswald, 1959c) who were not sleep-deprived and
who received electric shocks, sleep appeared in
four subjects. Powerful shocks were given by
the regular discharge, half a dozen times a
minute, of a 0.1 microfarad capacitor charged
to 320 volts, and discharged through the wrist
or ankle of the subject. Sleep tended to appear
between each shock. In one case it became con-
tinuous and deepened to the C stage (Fig. 30)
and ended following a dishabituating stimulus,
viz. a clap. P. 151.
The above data indicate that sleep may occur in re-
sponse to intense shock. Thus, deprivation Procedure A may
have self-defeating features, which Procedure P does not
have. This and the counter-habituation measure may have
made Procedure P more effective.
As may be seen in Figure 6, sleep latencies decreased
as a function of time of deprivation. Webb (1957) had a
similar result. However, his shortest latencies were in
the range of 11 to 12 min, even after 30 h of deprivation.
In this experiment, the latencies were much shorter, rang-
ing from 5 s to about 3 min despite the fact that in both
experiments, latency measures were taken by placing Ss in
an environment different from the one in which they were
being kept awake. Dement et al. (1967) deprived cats of
REM (PS) sleep and reported that Ss did not show evidence
of excessive sleepiness. They stated that the animals were
restless, prowled a great deal and in general were quite
active. Levitt (1967) reported that the revolving-water-
wheel method of deprivation differentially deprived his
rat Ss of PS. The Ss were able to learn to obtain as much
as 14 min/h of SWS within 32 h of deprivation by walking to
the front of the wheel and remaining stationary for 3 to
4 s while riding to the rear. It is thus possible that Webb
deprived Ss mostly of PS, and that his Ss were more active
and took longer to fall asleep than the PS- and SWS-deprived
Ss in this experiment.
Rat Ss were kept awake by means of two deprivation
procedures. One, Procedure A, used shock contingent upon
the appearance of EEG SWS. The other, Procedure P, used
a wide variety of changing stimuli that were not contingent
upon the appearance of EEG SWS.
Using two Ss, the EEG was monitored in order to eval-
uate the effectiveness of deprivation Procedure P, which
was measured as about 94% for one 3, and about 97% for the
other. However, when the individuals keeping the Ss awake
made a maximal effort, the procedure resulted in Ss being
awake up to about 99% of the time.
Using Procedure P, three Ss were kept awake, one for
6, one for 12 and one for 15 days.
During the first three days of deprivation by
Procedure P, the latency of sleep onset when Ss were un-
disturbed decreased to about 20 a.
Subjects' muscle tonus decreased greatly with dep-
rivation. The terminal point seemed determined by vari-
ables other than sleep. One S developed a pulmonary con-
dition and one showed disinterest in food with a resulting
sharp weight loss. It was concluded that Ss' level of
arousal fell to a point where essential responses neces-
sary for Ss' survival became impossible.
It was concluded that Procedure A was ineffective in
keeping Ss from obtaining substantial amounts of sleep
after about 15 to 20 h because sleep occurs as a response
to intense shock. Also, Procedure A relied on only one
stimulus (shock) which resulted in habituation of the
arousal response to shock. With Procedure P, habituation
was circumvented by changing the stimulus when Ss fell
In order to deprive an organism of sleep, active in-
tervention is required to maintain wakefulness. In much
of the animal research, this intervention was much like
Procedure A. Pegram (1968), for example, kept his monkey
Ss awake by administering sharp puffs of cool, compressed
air to the back of the neck and the side of the head when
there were behavioral or EEG signs of sleep. The air
puffs' being made.contingent upon sleep may well have been
an unintentional contaminating variable. Only one reviewer
has considered the difference in the effects of different
methods of deprivation. Murray (1965) reviewed the depri-
vation literature relative to the degree of frustration/
aggression in sleep deprivation studies with human Ss. He
concluded that a major factor contributing to the degree of
aggression observed in subjects was whether gentle or co-
ercive and frustrating techniques were used.
The primary goal of this experiment was to see if
making shock contingent upon the sleep response, i.e.,
Procedure A, versus non-contingent, multiple stimulation,
Procedure P, changes the sleep response after deprivation.
This was done in order to investigate the possibility that
sleep might decrease after deprivation, since Procedure A
resembled a learning paradigm in which a negatively re-
inforcing event follows a response whose frequency is to
In addition to the amount of post-deprivation sleep,
a number of relevant subsidiary questions or dependent
variables were studied in this experiment: Is the circa-
dian rhythm affected by the deprivation techniques? Does
recovery of lost sleep occur during daytime hours, night-
time hours or both? Is the length of sleep episodes af-
facted by the deprivation procedures?
Six male Long-Evans hooded, 264-day-old rats were
deprived of sleep for 24 h by deprivation Procedure P.
One day of baseline was recorded before deprivation and
sleep records were obtained for the two days immediately
after deprivation and for the eighth day after deprivation.
Thirty-three days later, the same Ss were deprived of
sleep uding deprivation Procedure A. A shock of 1.0 ma of
0.5 s duration was adequate to keep Ss awake for 24 h.
Analysis of Data
TS SWS and PS.--The data for each S, for each day,
were divided into four components, and converted to percent
as follows: (a) The number of minutes of SWS per daytime
TS is total sleep, which is SWS and PS.
hour (9:00 A.M. to 9:00 P.M.) was divided by the number of
minutes of record with lights on, and then multiplied by
100. (b) The number of minutes of PS per daytime hour
(9:00 A.M. to 9:00 P.M.) was divided by the number of
minutes of record with lights on, and multiplied by 100.
(c) The same as (a) but for nighttime hours or lights-off.
(d) The same as (b) but for nighttime hours or lights-off.
TS percent per day was obtained by adding all four compo-
nents and dividing by 2. Slow-wave sleep percent per 24-h
day was obtained by adding components (a) and (c) and di-
viding the sum by 2. Paradoxical sleep percent per 24-h
day was obtained by adding components (b) and (d) and di-
viding the sum by 2. In the various tables in the results
section below, TS percent per day as derived above were
summed across Ss for individual days separately, divided
by the number of Ss and presented as group means for days
as specified within the various tables. Paradoxical sleep
and SWS were treated and presented in the same way as TS.
Where the overall main days or treatment effects were sig-
nificant according to analysis of variance, paired compar-
isons of days post deprivation with baseline were done
with Dunnett's test (Kirk, 1968) using the same error mean
square'as for the F test. To contrast the effects of Pro-
cedure P and Procedure A, paired comparisons were made be-
tween corresponding days after deprivation by Procedure P
and Procedure A using t tests with the same error term as
for Dunnett's test. All probabilities presented are based
upon two-tailed tests.
Circadian rhythm.--The day/night ratio was obtained
for individual Ss for individual days by dividing TS day
by TS night. A decrease in this ratio would reflect a
decrease in circadian rhythm.
Length of sleep episodes.--To best explicate the
analysis of the data for length of sleep episodes, a
sample of raw data is given as illustration. Numbers
refer to the number of consecutive minute episodes scored
as SWS, PS and W.
W2 [SWS5. W, SWS3, PS2, wl, sws11
w3 [SWS4, PS1, SWSi] W5.
The square brackets isolate episodes of sleep. Paradoxi-
cal sleep and SWS were not differentiated. Single minute
episodes of W which were both preceded by and succeeded
by minutes of SWS or PS were counted as sleep. Wake epi-
sodes of 2 or more min divide the record into sleep epi-
sodes. Accordingly, the illustration above shows two
sleep episodes of 23 and 6 min. Chi square was used as a
descriptive statistic since the data did not meet the re-
quirements of independence (since data from Ss were pooled).
Results and Discussion
TS, SWS and PS
As may be seen in Tables 6 and 7, sleep amounts in-
creased after deprivation. For day 1 post deprivation for
both deprivation Procedures P and A, TS, SWS and PS were
all statistically significantly increased relative to the
Significantly more PS was obtained on day 1 after
Procedure P when compared to day 1 after Procedure A.
There is also a trend in the data (p( .1) suggesting that
less TS was recovered on day 1 after Procedure A (see
Table 8 presents the data reflecting the changes in
daytime and nighttime sleep and in the day/night ratio
after 24 h of both deprivation procedures. The same data
are presented again in Table 9 where the increase over
baseline percent sleep post deprivation is presented rather
than the actual percentage of sleep as in Table 8.
As may be seen in Tables 8 and 9, both daytime and
nighttime sleep increased on days 1 and 2 after both dep-
rivation procedures. Thus, sleep lost during 24 h of
deprivation is recovered during both daytime and nighttime
The data in Tables 8 and 9 suggest that the increases
in daytime and nighttime percent sleep are not proportional
to the respective baseline amounts. For example, on the
Baseline and Recovery Sleep after Deprivation
Procedures A and P in Mean Percent per 24-Hour Day
Baseline 1 2 8
1 2 8
vs. Procedure P
Procedure A b
vs. Procedure P
Procedure A b
vs. Procedure P
64.5 51.4 45.4
60.3 51.4 49.3
49.1 41.9 36.0
15.4 9.5 9.4
47.1 42.1 40.8
13.2 9.3 8.6
bContrasting corresponding days after deprivation, using t test.
Summary of Analyses of Variance for. Data
for which Group-by-Day Means are in Table 6
Days x Ss
Circadian Rhythm: Daytime and Nighttime TS
and Day/Night Ratio for Baseline and Days after
Deprivation Procedures A and P
1 2 8
1 2 8
Daytime TS 61.5 77.0 63.9 56.2 75.2 65.7 60.8
Nighttime TS 31.1 52.0 38.8 34.5 45.4 37.1 37.8
Day/Night Ratio 2.11 1.50 1.67 1.68 1.70 1.93 1.69
Circadian Rhythm: Increase in Percent
Sleep over Baseline after Deprivation
Procedures A and P
Daytime TS 61.5 15.5 2.4 -5.3 13.7 4.3 -.7
Nighttime TS 31.1 20.9 7.8 3.5 14.3 6.0 6.8
first night after deprivation Procedure P, sleep percent
was about one and two-thirds of the baseline level,
whereas the daytime sleep on the same day was only about
one and one-quarter of the baseline level. Rather the in-
crease in percent sleep seems to be additive in that during
the first 12 daytime and the first 12 nighttime h, sleep
percent increased by an absolute amount of about 14 to 21%.
Because the recovery increase was approximately additive,
the circadian rhythm, as measured by the day/night ratio in
Table 8, decreased on day 1 after deprivation from 2.11 to
between 1.68 and 1.93.
Levitt reported a decrease in circadian rhythm post
deprivation by dextroamphetamine, and that sleep recovery
occurred during nighttime only (1965, P. 37). However,
after deprivation by treadmill, recovery occurred during
both daytime and nighttime, with the circadian rhythm being
significantly increased for days 3 to 5 post deprivation
(Levitt, 1965, pp. 65-66). The results of this experiment
support the findings that when behavioral or non-chemical
deprivation techniques are used, sleep is recovered during
both daytime and nighttime hours. It is possible that a
side-effect of deprivation by dextroamphetamine was to
somehow modify the post-deprivation circadian rhythm.
Sleep episode length
Examination of the sleep episode data suggested that
the episodes be grouped into three intervals: up to 15 min
(short episodes), 16-40 min (medium episodes) and 41 min or
more (long episodes). The number of episodes between 16
and 40 min changed little from baseline as a function of
deprivation. Chi square was calculated in fourfold tables
contrasting (a) the proportion of long (41+) and short epi-
sodes (up to 15 min) for the individual days post depriva-
tion with baseline, and (b) corresponding days after the two
deprivation methods. The frequency data and Chi square ana-
lyses are presented in Table 10. After deprivation, the
proportion of sleep episodes shifted so that the proportion
of short episodes decreased and the proportion of long epi-
sodes increased. This change was observed on day 1 after
both deprivation Procedures P and A. This shift was greater
after Procedure P than after Procedure A. Thus, (a) when Ss
were recovering lost sleep, sleep episodes tended to lengthen
and (b) deprivation Procedure A tended to reduce this effect.
One group of six rat Ss was kept awake for two 24-h
periods, 33 days apart, and after a baseline day had been
recorded. Two techniques of deprivation were used, both
of which used the EEG to monitor the sleep-wake state.
The first, Procedure P, involved many stimuli, not contin-
gent upon the sleep response, while the second, Procedure A,
Since the data did not meet the requirements of
independence (since data from Ss were pooled), Chi square
was used only as a descriptive statistic.
Sleep Episodes in Terms of Frequency of Occurrence
of Specified Length of Episodes
8 1 2
1-15 min 224 92 180 169 153 170 201
16-40 min 74 62 86 79 74 72 76
41+ min 18 49 25 20 35 29 22
Deprivation vs. 46.05 1.596 3.6798 12.24 5.838 .8698
Procedure A vs.
Procedure P 11.03 .493 .0574
involved shock to the feet contingent upon EEG SWS.
Total sleep, SWS and PS were all significantly in-
creased for the first 24 h after both methods of depriva-
tion. Significantly more PS was obtained on day 1 after
Procedure P and there was a trend in the data suggesting
that more TS was obtained after this procedure.
Sleep (TS) was recovered after both methods of dep-
rivation during both daytime and nighttime hours. The
sleep increase was approximately additive and not propor-
tional to the baseline daytime and nighttime sleep. This
resulted in a decrease in the circadian rhythm as measured
by the day/night sleep ratio.
Examination of the sleep episode data indicated that
after deprivation there were fewer short sleep episodes
(15 min or less) and more long episodes (41 min or more).
This change was greater after Procedure P than after
A major question in the area of sleep remains essen-
tially unanswered: When an organism loses sleep, what pro-
portion of the lost sleep is later recovered? In the lit-
erature on human sleep post deprivation, the question which
has received most attention has been the percent sleep time
of various sleep stages post deprivation.
Berger and Oswald (1962) recorded Ss' baseline sleep
for four nights, deprived Ss of sleep for four nights and
then recorded the recovery of sleep for four nights. Sub-
jects were allowed 12 h of sleep on the first recovery
night. On the baseline and other recovery nights Ss were
allowed 7 to 8 h of sleep. This procedure prevented answers
to the question of amount of recovery of lost sleep. De-
tailed comparisons of sleep stages were made between the re-
covery nights and the baseline nights. Percent of time
taken by Stages 1-REM and 2 were significantly decreased
when the first 7j h of the first recovery night was compared
to the baseline nights which were 7j h. Based on the first
7* h of sleep on the first recovery night, percent Stage
1-REM time was 7.4 as compared to 22.5% on baseline. For
the full 12 h, Stage 1-REM time was 18.2%. On the second
recovery night, Stage 1-REM time was 27.5%, 26.9% on the
third and 23.8% on the fourth. No other details were
given by the authors. Berger and Oswald's basic finding
of increased Stage 4 and decreased Stage 1-REM on the
first night after deprivation and then increased Stage
1-REM on the second recovery night has been.supported by
Naitoh, Kales, Kollar and Jacobson (1968), after 204 h of
wakefulness, by Kollar, Slater, Palmer, Docter and Mandell
(1966), after 120 h, by Williams, Hammack, Daly, Dement
and Lubin (1964), after 64 h, and by Williams and Williams
(1966), after 64 h of wakefulness. Gulevich, Dement and
Johnson (1966) studied one S that had stayed awake for 264
h. This S, unlike all of those in the studies above, was
allowed to sleep until he awoke spontaneously. For this S,
both Stages 1-REM and 4 were increased for the first as well
as the second and third recovery nights. This suggested
that the reports of decreased Stage 1-REM on the first night
after deprivation were a result of terminating the first
night's sleep arbitrarily, after some predetermined number
of hours. The Berger and Oswald study above, with Stage
1-REM sleep percentages of 7.4 and 18.2 for the first 72
and 12 h respectively, would support this notion. The data
of Gulevich et al. (1966) also suggested that recovery was
not complete after three days. One week after deprivation,
S slept 424 min as compared to 388 and 391 min, 6 and 10
weeks later respectively.
A number of additional phenomena in the recovery sleep
of humans after deprivation have been described. Gulevich
et al. mentioned that Stage 1-REM periods were unstable and
were often interrupted by bursts of Stage 2 activity, i.e.,
spindles. Naitoh et al. (1968) noted a similar phenomenon.
In both studies, Ss were deprived for over 200 h. A de-
crease in body movements in the sleep after deprivation was
reported by Williams and Williams (1966) after 64 h of wake-
fulness, by Cooperman, Mullin and Kleitman (1934) after 60
h of wakefulness, and by Marbach and Schaff (1960) with one
night's sleep loss. Williams et al. (1964) reported a de-
crease in Ss' responsiveness to stimuli as measured by
changes in EEG, vasoconstrlction and the response of press-
ing a microswitch after deprivation.
The data on the amount of total sleep humans make up
after sleep deprivation are very sparse, inadequate and
variable. Edwards (1941) reported that Ss required be-
tween 86 and 299 of lost sleep time for recovery. Appar-
ently, Edwards meant that Ss recovered between 86 and 29%
of the sleep lost during deprivation in the weeks after
deprivation. Katz and Landis (1935) summarized the stud-
ies done prior to this report, and concluded that a very
small percentage of the lost sleep was made up. Except
for Edwards, authors since Katz and Landis have not attemp-
ted to estimate the proportion of lost sleep made up by
A number of authors have studied various indices
which may have reflected the effects of sleep deprivation,
after some recovery sleep. Essentially, one can conclude
that one night of sleep after deprivation is not adequate
to restore Ss to the pre-deprivation state. Ax and Luby
(1961) reported that autonomic measures were not at baseline
after three days. Nine h of sleep, taken on the day.after
98 h of sleep deprivation, were not enough to restore EEG
alpha to normal. An additional night's sleep was required
for EEG alpha to be essentially normal (Armington and'
Mitnick, 1959). Based on subjective reports, the Ss stud-
ied by Edwards (1941) had not returned to normal until they
had lived and slept normally for a week or even more.
Wilkinson (1963) studied the after-effect of one night's
sleep.loss on performance after one recovery night of sleep.
His Ss had about an extra 2 h of sleep beyond the night's
sleep by napping in the evening before going to bed for the
night. Performance on a vigilance and on serial reaction
tasks were below the control levels. Williams, Granda,
Jones, Lubin and Armington (1962) found a similar result
with a vigilance task, when Ss slept for 9 h following 64 h
Based on the review above, one could conclude that
after sleep deprivation, human Ss manifested their in-
creased need for sleep by sleeping more than before depri-
vation. However, the question of the absolute amount of
sleep recovered relative to that lost is not answerable.
The data on sleep stage recovery are not easily interpret-
able. Studies need to be done in which Ss are allowed to
sleep as much as they like both before and after depriva-
Since laboratory animals can be allowed to sleep
without the interference of social conditioning of sleep,
they offer a much better opportunity to study the question
of recovery of lost sleep. In this area, Levitt (1965, 1966
and 1967) performed the most extensive experiments avail-
able on sleep after deprivation. These experiments were
done in the same laboratory and using the same strain of Ss
used by this E. In one experiment (Levitt, 1965), three
groups of rat Ss were kept awake by injections of dextro-
amphetamine sulphate for 24, 72 and 120 h respectively.
The dependent variable measure of sleep was activity, re-
corded by an ultrasonic device. Levitt found that follow-
ing sleep deprivation, there was a decrease in circadian
rhythm. He found no difference in total amount of com-
pensatory sleep.due to amount of deprivation. He esti-
mated the average amount of compensatory sleep. Compen-
sation was arbitrarily considered to have ended on the
first day that the mean sleep time for a deprivation group
was equal to the baseline mean. The 24-h-deprivation group
recovered 829 min of sleep whereas it lost only 784 min of
sleep. The 72-h-deprivation group and the 120-h-deprivation
group lost,2,517 and 3,970 min of sleep respectively; they
regained 450 and 568 min or 18 and 14% of the lost sleep
respectively. Unfortunately, the multiple effects of the
drug and sleep loss are confounded in this study.
Pegram (1968) has provided data which allow an impor-
tant contrast to Levitt's findings. Unfortunately, his
deprivation procedure was similar to Procedure A. Pegram
kept rhesus monkeys awake for eight days and recorded their
EEG for five baseline and eight recovery days. He found
that PS percent returned to its baseline level after about
two days. Stage 4 recovery was similar in showing a rapid
recovery occurring mostly in the first 4 h of recovery
sleep. However, Stage 2 was elevated for all eight re-
covery days and together with total sleep time was still
substantially above baseline on the eighth recovery day.
Although Pegram did not analyze his data in terms of the
amount of sleep recovered, calculations based on his raw
data suggested that Ss slept enough over baseline in the
eight recovery days to make up for the sleep lost in about
two to three days of normal sleep. (Pegram presented raw
data for only five of the eight recovery days. The data for
the three days not presented were interpolated by averaging
the day before and the day after the omitted days.) These
data were in sharp contrast to Levitt's data where the rats
deprived for three and five days recovered less than one
day's lost sleep. Also, Pegram's data were in sharp
contrast to the human Stage 1-REM recovery data. Pegram's
Ss did not recover the Stage 1-REM sleep lost and in fact
recovered more Stage 2 than Stage 1-REM.
Levitt (1965) also deprived two groups of rat Ss of
sleep by means of a treadmill for 24 and 72 h respectively.
Again, sleep was measured by activity recordings. In this
experiment, (a) there was no circadian rhythm reduction,
(b) deprivation level did not have a significant effect,
and (c) the 24-h group lost 810 min of sleep and recovered
168 min or 21%, while the 72-h group lost 2,430 min and made
up 313 min or 13% of the lost sleep. However, in another
experiment, Levitt (1967) noticed that bursts of slow-wave
sleep, which he referred to as microsleep, appeared in the
EEG of Ss on the treadmill. In the same study, sleep was
recorded using an EEG, before and after total sleep depriva-
tion by means of dexedrine for three Ss and by means of the
treadmill for three other Ss. Slow-wave sleep increased
following deprivation by dexedrine only, whereas PS in-
creased after both deprivation procedures. The finding of
an increase in PS only, following deprivation, was reported
by Kiyono, Kawamoto, Sakakura and Iwama (1965) using cats.
They used visual monitoring to determine whether Ss were
awake or asleep, and their results, like Levitt's,were not
surprising considering the fact that their deprivation pro-
cedures deprived animals almost completely of PS and only
partially of SWS.
In this experiment, two groups of Ss (Group A and
Group P*) were deprived of sleep for 24 h by Procedures
A and P respectively. This eliminated any interaction
effects between the two methods since each group was de-
prived once and by one method only. Subjects were se-
lected so that their baseline sleep levels were equiv-
alent, and a very intense shock was used to keep Group A
Five male Long-Evans hooded rats were selected at ran-
dom from a pool of 11 rats, all of which had been prepared
for EEG recording. They were kept awake by deprivation
Procedure P, and are thus referred to as Group P. These
Ss were about 160 days old on the first baseline recording
A second group of eight Ss was prepared for EEG re-
cording and five were selected to form a group whose base-
line sleep percent matched those of Group P as closely
as possible. This second group was deprived of sleep by
means of Procedure A and is therefore called Group A.
Two baseline days were recorded a few days prior to
deprivation. Recordings were begun immediately after dep-
rivation and continued until the end of the ninth day
These groups are defined in greater detail below.
Analysis of Data
The data were converted to sleep percentages and ana-
lyzed as described in Experiment III. The statistical de-
sign for the analysis of variance used is described by
Lindquist (1953, pp. 266-273). The difference between the
mean percent of sleep for any one group for any one day
and the mean amount of sleep for that day for any other
group was tested by a t test as recommended by Lindquist.
The difference between the baseline mean and the means of
days post deprivation for any one group were tested by
Dunnett's test for multiple comparisons using a control
Results and Discussion
The data were analyzed in a manner which allowed
measuring the effect of the deprivation procedures on SWS
and PS as well as on TS, as in Experiment III. Table 11
presents the overall analyses of variance for SWS, PS and
TS. Tables 12, 13 and 14 present group means in percent
per 24-h day before and after deprivation for SWS, PS and
TS respectively. Tables 12, 13 and 14 also present pro-
babilities for Dunnett's test, which contrast each day
after deprivation with the mean of the two baseline days,
and the probabilities for the t tests as suggested by
The t test used the square root of twice the mean
square for error within cells divided by the N for either
group since Ns were always the same. A cell in this ana-
lysis included the individual scores of Ss in a particu-
lar treatment group aversivee or positive deprivation) for
any one particular.day.
Overall Analyses of Variance for TS, SWs and PS
per 24-Hour Day for which Data are in
Tables 12, 13 and 14
TS SWS PS
Source df F p F p F p
Between Ss 9
Deprivation Groups 1 1.02 1.03 .25
error (b) 8
Within Ss 100
Days 10 27.63 (.001 13.35 (.001 23.89 <.001
Days x Deprivation 10 1.45 1.80 (.10 1.97 -=.05
error (w) 80
SWS Group Means in Percent
Pre and Post Deprivation
(per 24-Hour Day)
Day Post Deprivation
Mean 1 2 3 4 5 6 7 8 9
Group P 42.2 54.3 49.7 48.0 44.8 43.5 43.9 44.6 43.8 42.2
pa .01 (.01 ( .01
Group A 42.4 49.4 44.1 44.0 43.7 42.1 42.8 43.1 43.5 42.4 ON
Group p b =5 .05
vs. Group A P =.05 .
PS Group Means in Percent
Pre and Post Deprivation
(per 24-Hour Day)
Day Post Deprivation
1 2 3 4 5 6
7 8 9
vs. Group A
9.15 12.80 9.18 9.62 9.72 8.66 8.62 9.70
9.17 13.92 8.18 9.40 8.90 8.34 8.54 9.24 10.04 7.76
Means in Percent
Day Post Deprivation
1 2 3 4 5 6 7 8 9
67.1 57.2 57.6 54.4 52.2 52.5 54.3 52.9 51.3
<.01 ( .01
63.3 52.3 53.4 52.6 50.4 51.3 52.4 53.5 50.1
vs. Group A
Lindquist (1953, p. 272), which contrast the effect of the
two deprivation techniques on corresponding days after
deprivation. Figures 7, 8 and 9 are graphic plots of the
data presented in Tables 12, 13 and 14 respectively.
Table 15 presents the percentage (or proportion) of lost
SWS, PS and TS respectively, recovered during days 1 to 9
after deprivation per individual .subject and per group.
Increase in SWS, PS and TS post deprivation
As may be seen in Table 11, the treatment days effect
was statistically significant for SWS, PS and TS. Only
for PS was the interaction of days by treatment groups
statistically significant. As may be seen in Tables 12,
13 and 14, there is a significant increase in SWS, PS and
TS relative to the baseline after both techniques of dep-
rivation, for at least the first day after deprivation.
This is the same as was found in Experiment III, Table 6.
In the literature review, it was suggested that the
reports, in the human sleep literature, of decreased
Stage 1-REM percent on the first night after deprivation,
were a result of terminating the first night's sleep ar-
bitrarily after some predetermined number of hours. Al-
though this suggestion ultimately can only be validated
or rejected by research with human Ss, the results of this
experiment offer some limited support to the notion. The
Ss in this experiment and in Experiment III showed a sig-
nificant increase for the first 24-h recovery day, after
both techniques of deprivation, in the percentage of the
0------0 Group A
o----o Group P
S46 2- \ "
1 2 1 2 3 4 5 6 7 8 9
BASELINE DAY POST DEPRIVATION
Figure 7. Slow-wave sleep group means pre and post
per 24-hour day.
o---o Group P
o -----o Group A
I I I I I I I I 1 I
1 2 3
4 5 6 7 8 9
,sleep group means pre and post
per 24-hour day.
r i I I I li I I I l I
1 2 1 2 3 4 5 6 7 8 9
BASELINE DAY POST DEPRIVATION
Figure 9, Total sleep group means pre and post deprivation
per 24-hour day.
---o Group P
S-----.o Group A
Percent of Lost SWS, PS and TS Recovereda
During Days 1 to 9 per Individual Subject and Groupb
Percent of Lost Sleep
Group P 1
Group A 1
aEach 1% of sleep may be thought of as a unit of
sleep; thus Ss obtained about 50 units of sleep per day.
Recovery sleep percent is the number of units of extra
sleep over baseline divided by the number of units of
sleep on baseline.
bMinus signs indicate a net loss of sleep relative
to baseline during recovery days 1 to 9.
24-h day devoted to PS, which is the rat's equivalent to
Stage 1-REM (see Table 6, Table 13 and Figure 8).
Pegram (1968) found that after eight days of depriva-
tion with rhesus monkeys, PS returned to its baseline
level after about two recovery days while Stage 2 was el-
evated for all eight recovery days. In the rat, SWS is
probably equivalent to primate Stages 2 and 4 combined,
since the rat has only PS and SWS. In this experiment,
as may be seen in Table 12, it was found that SWS (or
non-PS) was noticeably elevated (more than 1%) for eight
of the eight days after deprivation Procedure P and for
five of the eight days after deprivation Procedure A. On
the ninth day, both groups had returned to the baseline
level. As may be seen in Tables 6 and 13, PS was elevated
for only day 1 post deprivation, after both techniques of
deprivation, in this experiment and in Experiment III.
This suggests that whatever PS is recovered after depriva-
tion by subhumans, or possibly by all Ss allowed to sleep
ad lib after deprivation, is recovered in less time than
it takes for recovery of non-PS or SWS. Pegram's data
strongly suggest that his Ss recovered proportionately
more non-PS than PS. Similarly, in this study, it was
found that proportionately more SWS was recovered than PS.
Proportion of lost sleep recovered
Levitt (1965) kept a group of rat Ss awake for 24 h
by using dextroamphetamine sulphate. This group recovered
more sleep than it lost. As may be seen in Table 15, only
one of six Ss kept awake by Procedure P recovered more
sleep than it lost and no Ss kept awake by Procedure A
recovered as much as was lost. This suggests that Levitt's
chemical means of deprivation may have had some unusual ef-
fects on recovery sleep, not found when behavioral means
are used to keep Ss awake. (This is also suggested by the
fact that Pegram's Ss recovered about two to three days'
worth of the eight days' sleep lost while Levitt's five-day-
deprived groups recovered less than one day's worth of
As may be seen in Table 15, the group deprived of sleep
by Procedure A recovered less than one-half as much PS, SWS
and TS as the group deprived by Procedure P. Definitive
evidence that these differences are not a function of chance
is not available. This is because, as may be seen in
Table 11, the overall differences between the total amounts
of SWS, PS and TS obtained by the two deprivation groups
for the sum of the two baseline days plus the nine recov-
ery days are not statistically significant despite the fact
that the baseline sleep levels were almost identical. How-
ever, there is some evidence which suggests that the differ-
ences, particularly for SWS, are not a function of chance.
This evidence comes from the t tests which reflect the sig-
nificance of the differences between the mean percent of
sleep for any one group for any one day and the mean amount
of sleep for that day for any other group. As may be seen
in Table 12, Ss in Group P obtained significantly more SWS
on days 1 and 2 post deprivation than did Group A. The Ss
in Experiment III obtained more (but not significantly
more) SWS on day 1 after deprivation by Procedure P than
on day 1 after Procedure A. However, this difference was
non-existent on day 2 post deprivation and was reversed on
day 8 post deprivation.
As may be seen in Table 13, there were non-significant
differences for PS between the two groups, except for re-
covery day 9, when Group P obtained significantly more PS
than Group A. In Experiment III. however, as may be seen in
Table 6, Ss obtained significantly more PS on day 1 post
deprivation by Procedure P.
As may be seen in Table 14, there were trends in the
data (p<.l) suggesting that more TS was obtained on days
2 and 3 after Procedure P than after Procedure A for the
same day. (The critical t at the .1 level was 1.68 and the
t for day 1 was 1.63 which was thus just short of signifi-
cance at the .1 level. Also, the critical t at the .05
level was 2.18 and the t for day 2 was 2.11.) There was
also a similar trend in Experiment III as more TS was ob-
tained on day 1 after Procedure P, the difference being sig-
nificant at about the .1 level. In the light of all this
evidence, it may be tentatively concluded that making de-
livery of shock contingent upon the sleep response, as a
means of deprivation for 24 h, results in a large decrease
in the amount of sleep recovered after deprivation. Ss may
have learned to sleep less.
As may be seen in Figure 7, after Procedure P, SWS is
recovered in a pattern describable as a decelerating curve
of the percent of sleep as a function of time so that the
bulk of the recovery is complete after three days and the
level declines slowly back to the baseline level for another
five days. The pattern for recovery after Procedure A is
similar, the only apparent difference being that there is a
consistently smaller increase after deprivation. As may
be seen in Figure 8, the recovery pattern of PS is irreg-
ular except for the large increase in PS on day 1 after dep-
rivation. As may be seen in Figure 9, the pattern for TS
does not differ notably from the pattern for SWS.
Table 16 presents the data reflecting the changes in
daytime and nighttime sleep, and in the day/night ratio
for Groups A and P. The same data are presented again in
Table 17, where the increase over baseline percent sleep
after deprivation is presented rather than actual percent
sleep as in Table 16.
As may be seen in Tables 16 and 17, both daytime and
nighttime sleep increased substantially on day 1 after
deprivation for both groups. The conclusion in Experiment
III that sleep lost in 24 h of deprivation is recovered
during both daytime and nighttime hours is thus supported.
This conclusion is very different from Levitt's (1965,
pp. 41-42). After deprivation by dextroamphetamine,
Circadian Rhythm: Daytime and Nighttime TS
and Day/Night Ratios for Groups A and P
Day Post Deprivation
Mean, 1 2 3 4 5 6
7 8 9
Daytime TS 58.7
Nighttime TS 43.9
Daytime TS 67.4
Nighttime TS 35.7
77.7 63.8 67.1 64.6 63.3 64.9
48.9 40.7 39.6 40.6 37.6 37.8
1.61 1.58 1.74 1.61 1.70 1.73
66.4 65.7 64.0
38.3 41.3 36.3
1.76 1.62 1.81
Circadian Rhythm: Increase in Percent Sleep over
Mean of Two Baseline Measures for Daytime and Nighttime
Hours after Deprivation Procedures A and P
Increase over Baseline
Day after Deprivation
1 2 3 4 5 6 7 8 9
Daytime 58.7 17.8 9.5 12.8 10.4 9.2 8.2 7.9 9.8 6.4
Nighttime 43.9 13.8 2.4 .-.2 -4.1 -7.4 -5.8 -1.8 -6.6 -6.6
Daytime 67.4 10.3 -3.6 -.3 -2.8 -4.1 -2.5 -1.0 -1.7 -3.4
Nighttime 35.7 13.2 5.0 3.9 4.9 1.9 2.1 2.6 5.6 .6
a slight decrease in sleeping time during
the,day. The entire compensation is ac-
counted for by an increased sleep time
during the dark or night phase of the light
cycle. An explanation of this phenomenon
may be that the rat, being a nocturnal
creature, sleeps maximally during the day,
awakening only to satisfy other need states
which require attention. The rat, according
to this hypothesis, cannot constantly re-
main asleep for 8-12 hours because of other
need requirements. Thus, only at night is
there time available to compensate for the
heightened sleep need.
In this experiment, Ss on day 1 after deprivation slept
more than 75% of the time, which is equivalent to 8 h of
sleep in the 12 daytime h.
The data in Tables 16 and 17 suggest that the in-
creases in daytime and nighttime sleep percent for day 1
after deprivation are not proportional to the respective
baseline amounts. For example, for Group A, on the first
night after deprivation, sleep percent was about one and
one-third of the baseline level, whereas the daytime sleep
was about one and one-seventh the baseline level. Because
of the disproportionality, the day/night ratio for Group A
decreased from the baseline level of 1.92 to 1.61 on day 1
As may be seen in Table 16, for Group P on day 1 post
deprivation, more sleep was recovered during nighttime
hours. For the other eight days after deprivation by Pro-
cedure A, there was a shift toward more nighttime TS than
on baseline, and a decrease in TS relative to baseline during
daytime hours. This resulted in an increase in circadian
rhythm. For Group A, as may be seen in Table 16, this
shift was reversed with a resulting decrease in circadian
rhythm. These changes seem to represent a tendency.for
the two groups to become more alike in terms of circadian
rhythm. Group P during baseline had a smaller day/night
ratio than Group A, and after deprivation Group P increased
in circadian rhythm while Group A decreased. After the
second recovery day, the ratio was between 1.6 and 1.9 for
both groups. This range also includes the post-depriva-
tion data in Experiment III.
Sleep episode length
The data were analyzed to reflect length of sleep
episodes as described in Experiment III. In Experiment
III, the frequency of episodes between 16 and 40 min did
not change after deprivation. In this experiment, the
frequency of episodes between 16 and 40 min increased by
a factor of almost one-third and Chi squares were cal-
culated using the three class intervals as presented in
Table-18. For Group P, the baseline days differed from
each other-very little (X = .67) and the mean of the two
baseline days was used in the baseline and the post-dep-
rivation contrasts. For Group P, for days 1 to 9 post
deprivation, sleep episodes between 1 and 15 min (short
Since these data did not meet the exact criteria of
independence (because of pooling of data from Ss) Chi
square was used only as a descriptive statistic.
Frequency of Occurrence of Minute Sleep
Episodes Grouped in Three Class Intervals
Day Post Deprivation
Group Lengths Baseline
(min) 1 2 1 2 3 4 5 6. 7 8 9
Group P 1-15
Group A 1-15
Group P 2
187 199 121 157 141 172 159 157 163 177 146
65 64 86 91 94 76 67 85 55 66 71
24 31 45 30 32 33 32 27 35 32 33
.7a 21,.9 8.2 13.9 2.7 2.3 6.2 2.3 .9 5.1
199 149 114 198 160 194 174 206 151 134 156
54 68 67 69 67 72 61 51 55 55 51
20 20 34 17 22 16 20 20 24 27 24
6.4a 13.1 .7 .8 1.1
0 2.6 1.0 3.4 1.1
1.7 3.6 1.2 30.7 7.0 7.3 3.9 16.0 1.4 3.0 4.3
aContrast of baseline days 1 and 2.
episodes) decreased in frequency while episodes of 16 to
40 min-(medium episodes) increased notably for days 1 to 4.
Episodes of 41 min or more (long episodes) increased no-
tably for only day 1 post deprivation. For Group A, short
episodes decreased and long episodes increased notably in
frequency for only day 1 post deprivation.
Thus the conclusions in Experiment III are supported:
(a) During recovery of lost sleep, length of minute episodes
increases so that there are fewer short episodes and more
medium and long episodes. (b) Procedure A reduces this
Two groups of five rat Ss were kept awake, each for
one 24-h period, after two baseline days of sleep had been
recorded. A different deprivation technique was used with
each group, but with both groups, the EEG was used to mon-
itor the sleep-wake state. Group P was kept awake by
means of Procedure P, which involved a variety of stimuli,
not contingent upon the sleep-wake state. Group A was kept
awake by means of the deprivation Procedure A, which in-
volved shock to the feet contingent upon EEG SWS.
As in Experiment III, TS, SWS and PS were all sig-
nificantly increased for at least day 1 post deprivation
for both groups. As in Experiment III, PS was increased
for day 1 post deprivation only and SWS was increased
for more than day 1. Thus the data for both experiments
suggest that a larger proportion of the lost SWS is re-
covered than PS. In Experiment IV, the group deprived by
Procedure P recovered 84.6 and 54.0% of the SWS and PS
respectively, lost during deprivation, while the group
deprived by Procedure A recovered correspondingly 33.0
The data from both experiments suggest that Pro-
cedure A results in decreased recovery of both PS and SWS
relative to the amount recovered after Procedure P.
Possibly, Procedure A resulted in Ss' learning to sleep
Sleep was recovered, in both experiments and after
both deprivation techniques, during both daytime and night-
time hours. These sleep increases were generally not pro-
portional to the baseline daytime and nighttime sleep.
This resulted in changes in the circadian rhythm, as mea-
sured by the day/night sleep ratios. These changes seemed
to be in the direction of making all groups alike after the
second day post deprivation.
SWS was recovered in a pattern describable as a de-
celerating curve of the percentage of sleep as a function
of time so that the bulk of the recovery is complete after
three days and the percent sleep level declines slowly back
to the baseline level for another five days. All the lost
PS that was recovered was obtained on day 1 post deprivation
The data reflecting the length of sleep episodes
indicated that there was a shift in both experiments toward
longer sleep intervals after deprivation. Procedure A
appeared to have decreased this shift in both experiments.
Anderson, L. M., and Gorfein, D. S. A case of prolonged
sleep deprivation. J. gen. Psychol., 1964, 71,
Armington, J. C., and Mitnick, L. L. Electroencephalogram
and sleep deprivation. J. appl. Physiol., 1959,
Ax, A., and Luby, E. D. Autonomic responses to sleep
deprivation. Arch. gen. Psychiat., 1961, 4,
Bast, T. H., and Loevenhart, A. S. Studies in exhaustion
due to lack of sleep. I. Introduction and
methods. Amer. J. Physiol., 1927, 82, 121-126.
Berger, R. J., and Oswald, I. Effects of sleep deprivation
on behavior, subsequent sleep, and dreaming.
J. ment. Sci., 1962, 108, 457-465.
Boren, J. J. Decrement in performance during prolonged
avoidance sessions. J. exp. Anal. Behav.,
1960, ., 201-206.
Cooperman, N. R., Mullin, F. J., and Kleitman, N. Studies
on the physiology of sleep. XI. Further obser-
vations on the effects of prolonged sleeplessness.
Amer. J. Physiol., 1934, 107, 589-593.
Crile, G. W. Studies in exhaustion. Arch. Surg., 1921,
Dement, W., Henry, P., Cohen, H., and Ferguson,.J. Studies
on the effect of REM deprivation in humans and in
animals. In Research publications: association for
research in nervous and mental disease. XLV. Sleep
and altered states of consciousness. Baltimore:
Williams and Wilkins, 1967. Pp. 456-467.
Edwards, A. S. Effects of the loss of one hundred hours of
sleep. Amer. J. Psychol., 1941, 14, 80-91.
Gulevich, G., Dement, W., and Johnson, L. Psychiatric and
EEG observations on a case of prolonged (264
hours) wakefulness. Arch. gen. Psychiat., 1966,
Katz, S. E., and Landis, C. Psychologic and physiologic
phenomena during a prolonged vigil. Arch.
Neurol. Psychiat., 1935, 24, 307-317.
Kimble, G. A. Hilgard and Marquis' conditioning and
learning. -2nd ed.) New York: Appleton-
Kirk, R. E. Experimental design; procedures for the
behavioral sciences. Belmont, Calif:
Kiyono, S., Kawamoto, T., Sakakura, H., and Iwama, K.
Effects of sleep deprivation upon the para-
doxical phase of sleep in cats. Electroenceph.
Clin. Neurophysiol., 1965, 19, 34-40.
Kleitman, N. Studies on the physiology of sleep. I. The
effects of prolonged sleeplessness on man.
Amer. J. Physiol., 1923, 66, 67-92.
Kleitman, N. Studies on the physiology of sleep. V.
Some experiments on puppies. Amer. J. Physiol.,
1928, 84, 386-395.
Kollar, E. J., Slater, G. R., Palmer, J. 0., Docter, R. F.,
and Mandell, A. J. Stress in subjects undergoing
sleep deprivation. Psychosom. Med., 1966, 28(2),
Leake, C., Grab, J. A., and Senn, M. J. Studies in
exhaustion due to lack of sleep. II. Symptom-
atology in rabbits. Amer. J. Physiol., 1927,
Legendre, R., and Pieron, H. Recherches sur le besoin de
sommeil cons6cutif a une veille prolong6e. Z.
all Physiol., 1912, 14, 235-262. Cited by N.
Kleitman, Sleep and wakefulness. (2nd ed.)
Chicago: University of Chicago Press, 1963.
Levitt, R. A. The sleep need: sleep deprivation in the rat.
Doctoral dissertation, University of Florida,
Levitt, R. A. Sleep deprivation in the rat. Science, 1966,
Levitt, R. A. Paradoxical sleep: activation by sleep
deprivation. J. comp. physiol. Psychol., 1967,
Licklider, J. C. R., and Bunch, M. E. Effects of enforced
wakefulness upon the growth and the maze-
learning performance of white rats. J. comp.
Psychol., 1946, 2, 339-350.
Lindquist, E. F. Design and analysis of experiments in
psychology and education. Boston: Houghton
Manac6ine, M. de. Quelques observations expdrimentales
sur l'influence de l'insomnie absolue. Arch.
Ital. Biol., 1894, 21, 322-325. Cited by N.
Kletman, Sleep and wakefulness. (2nd ed.)
Chicago: University of Chicago Press, 1963.
Marbach, G., and Schaff, G. Recherche d'une m6thode
devaluation de la quantity de sommeil.
II. Variations horaires de diff6rentes
expressions quantitatives de la motility
spontande au course du sommeil. Compt. Rend.
Soc. Biol., 1960, 154(2), 408-412.
Murray, E. J. Sleep, dreams, and arousal. New York:
Naitoh, P., Kales, A., Kollar, E. J., and Jacobson, A.
Interpretation of non-sleep EEG and sleep EEG
pattern in recovery nights after 204 hours of
prolonged wakefulness. Psychophysiology,
1968, 4(3), 392.
Okazaki, S. An experimental study of the lack of sleep.
Shinkei Gaku Zatshi, 1925, 21(2), 55-100.
(Psychological Abstracts, 1928, 2, No. 2773)
Oswald, I. Falling asleep open-eyed during intense
rhythmic stimulation. Brit. med. J., 1960,
Oswald, I. Sleeping and waking: physiology and psychology.
New York: Elsevier, 1962.
Pavlov, I. P. Conditioned reflexes: an investigation of
the physiological activity of the cerebral
cortex. New York: Oxford University Press, 1927.
Pegram, G. V. Changes in EEG, temperature, and behavior
as a function of prolonged sleep deprivation.
Doctoral dissertation, University of New Mexico,
Swisher, J. E. "Activity, electroencephalogram and ob-
served behavior of the sleeping rat." Master's
thesis, University of Florida, 1961.
Tarozzi, G. Sull'influenza dell'insonnio sperimentale sul
ricambio material. Riv. Pat. Nerv. Ment.,
1899, 4, 1-23. Cited by N. Kleitman, Sleep and
wakefulness. (2nd ed.) Chicago: University of
Chicago Press, 1963. P. 215.
Webb, W. B. Antecedents of sleep. J. exp. Psychol.,
1957, 5(3), 162-166.
Webb, W. B., and Agnew, H. W., Jr. Sleep deprivation,
age, and exhaustion time in the rat. Science,
1962, 136, 1122.
Wilkinson, R. T. Aftereffect of sleep deprivation.
J. exp. Psychol., 1963, 66(5), 439-442.
Williams, H. L., Granda, A. M., Jones, R. C., Lubin, A.,
and Armington, J. C. EEG frequency and finger
pulse volume as predictors of reaction time
during sleep loss. Electroenceph. Clin.
Neurophysiol., 1962, 14, 64-70.
Williams, H. L., Hammack, J. T., Daly, R. L., Dement, W.,
and Lubin, A. Responses to auditory stimulation,
sleep loss and the EEG stages of sleep.
Electroenceph. Clin. Neurophysiol., 1964, 16,
Williams, H. L., and Williams, C. L. Nocturnal EEG
profiles and performance. Psychophysiology,
1966, 1(2), 164-175.
Hyman.Solomon Sternthal was born in Montreal, Canada
on September 19, 1939. He attended Strathcona Academy and
graduated from high school in June, 1956. In 1960, he
received the B.Sc. degree from McGill University in
Montreal. In 1963, he moved to Gainesville to attend the
University of Florida and there received the M.A. degree
in 1965. From 1965 to the present time, he pursued his
work toward the degree of Doctor of Philosophy in Experi-
mental Psychology. He also completed an internship in
Clinical Psychology at the University of Kansas Medical
Center in 1971. Presently, he is employed at Florida State
Prison. He has established permanent resident status in
the United States and, with his wife and daughter, now
makes his home in Florida.