Journal ol Abnormal Psychology1968, Vol. 73, No. 5, 403-415

EFFECTS OF "DREAM DEPRIVATION" ON DREAM CONTENT:

AN ATTEMPTED CROSS-NIGHT REPLICATION1

DAVID FOULKES, TERRY PIVIK," JOHN B. AHRENS, AND ETHEL M. SWANSON

University of Wyoming

The present investigation attempted to replicate, in a cross-night design, therecent finding that early-night REM-sleep deprivation leads to more dream-like late-night REM-sleep mentation. 8 5s were REM deprived on Night 1 andpseudodeprived on Night 3, while 8 5s were run in the reverse order. Reportsof mental content were elicited from NREM and REM sleep on Nights 2 and 4.5s and "watchers" completed Gough Adjective Check Lists (ACLs) on 5behavior during Days 2 and 4. Deprivation did not produce significant changesin waking behavior, NREM or REM mentation, or in the sleep cycle. Thesefindings were discussed in terms of dream theory and methodological problemsof deprivation research. Deprivation did lower daytime oral temperature, andACL responses did predict physiological compensation from deprivation.

A random low-voltage EEG pattern("ascending EEG Stage 1"), accompaniedphasically by bursts of conjugate rapid eyemovements (REMs) and tonically by amarked suppression of muscle tone in facialand neck musculature, is associated with vividnocturnal dreaming (Dement, 196Sa). At-tempts to assess the functions served bydreaming have, therefore, employed a methodof "dream deprivation" in which 5s are pre-vented, by nocturnal awakenings or drugs,from experiencing their usual quota of REMsleep.

In the earliest dream-deprivation experi-ments (Dement, 1960), effects observed in-cluded decreased latencies to REM-sleep onseton deprivation nights, increased amounts ofREM sleep on uninterrupted recovery nightsfollowing the deprivation manipulation, andincreased anxiety, irritability, and appetiteduring the waking hours between deprivationnights. These results were interpreted as a"buildup of a pressure to dream," and asindicating that dreaming plays, as Freud(19S6) imagined, an essential role in themaintenance of psychic stability.

However, more recent experimentation, in-dicating, for example, sleep-cycle effects ofdream deprivation in normal and pontile cats

1 This research was supported by National ScienceFoundation Grants GS-860 and GS-2123. The au-thors gratefully acknowledge the assistance of KentAnger and Thomas Croft as "watchers" and ofJames D. Larson with statistical analysis.

2 Now a student at Stanford University.

(Jouvet, 196S), has led many sleep researchers(e.g., Dement, 196Sb) to ascribe the previ-ously supposed effects of dream deprivationto the deprivation of the physiological state ofsleep in which dreaming occurs, rather than tothe deprivation of its correlated psychologicalexperiences. In this vein, research is now beingconducted in several laboratories in an at-tempt to discover essential physiological proc-esses that normally occur during REM sleepand whose inhibition might be responsiblefor the observed effects of the dream-depriva-tion manipulation.

That dreaming might nonetheless play somerole in the dream-deprivation phenomenonhas been suggested by the recent study ofPivik and Foulkes (1966). These authors de-prived 5s of REM sleep during the early por-tion of 1 night and pseudodeprived the same5s during the early portion of another night(awakenings made so as not to intereferewith REM sleep). They found that late-night dream reports on the experimental nightwere rated reliably more dreamlike on a scaleof dreamlike fantasy, the Df Scale (Foulkes,Spear, & Symonds, 1966), than were control-night reports. This effect was observed bothfor represser 5s and for sensitizer 5s—asdetermined by Byrne's Repression-Sensitiza-tion (R-S) scale of the MMPI (Byrne, Barry,& Nelson, 1963)—but was statistically sig-nificant only for repressors. Likewise, eye-movement frequency, an established correlateof dream intensity (Berger & Oswald, 1962b),

403

404 FOULKES, PIVIK, AHRENS, AND SWANSON

was greater in the REM period followingdream deprivation than in that followingpseudodeprivation both for repressers andsensitizers, but again the difference was sig-nificant only for repressors.

Pivik and Foulkes noted that their resultswere generally consistent with the Freudianconception of the dream as a "safety valve,"a conception that also figured prominentlyin the interpretation of the earliest dream-deprivation experiments. They suggested thatthe demonstration of specific experiential, aswell as physiological, effects during sleep ofexperimental interference with REM sleep in-dicates that the deprivation of the psycholog-ical experience of dreaming may be a factorin dream-deprivation experiments.

In the present study, the authors attemptedto replicate and extend the findings of thePivik and Foulkes (1966) study in a cross-night design. Rather than depriving orpseudodepriving and collecting posttreatmentREM-sleep reports all on the same night, itwas decided to either REM deprive orpseudodeprive over the course of 1 entirenight, collecting dream reports from the initialREM period of the immediately subsequentnight.

This design also allowed the present au-thors to assess the effects of REM-sleepdeprivation on non-REM (NREM) mentation(Foulkes, 1966, ch. 4) and on waking be-havior. Because, except in special circum-stances (e.g., narcolepsy, multinight REMdeprivation), NREM sleep precedes the initialREM period of a night's sleep, it was pos-sible to consistently collect reports from suchsleep in advance of the occurrence of REMsleep. The 5s were observed by "watchers"on the days following deprivation and pseudo-deprivation, making it possible to note anywaking behavioral effects of the deprivationvariable. Psychological observations duringwakefulness and NREM sleep were consideredessential, as any hypothetical accumulation ofa pressure to dream might be displaced tothese states before 5s could achieve their firstundisturbed postdeprivation REM period.

METHODSubjects

The 5s were recruited from among volunteers whoresponded to billboard posters placed on campus.

The posters stated that, in return for sleeping 4consecutive nights in the University sleep laboratoryas part of a study of the psychophysiology ofsleep, eligible male Ss would receive $40. To facilitatecomparison with the Pivik and Foulkes (1966)study, the Byrne R-S scale was administered as ascreening device. Only Ss scoring within the range ofthose authors' repressors (28 or below) were ac-cepted. The actual score range for the 16 5s finallychosen was 2-26, with a mean score of 13.2. All 5swere in their late teens or early 20s, and, as was alsotrue of Pivik and Foulkes' 5s, all were universitystudents.

OrientationWhen 5s volunteered for the experiment, they

were given the screening test and a face sheet con-taining a general overview of the experimentalprocedures to be followed. It was indicated to 5sthat the screening test was being administered toprocure 5s comparable to those of previous studiesthat had employed the same test.

The face sheet introduced the experiment as astudy designed to investigate some rather com-plicated intercorrelations among body temperatureduring sleep and wakefulness, brain-wave patternsduring sleep, and mental activity or dreaming duringsleep.

To achieve these ends, the following procedureswere specified:

Night 1: A night on which you will be awakenedfrom sleep at fairly frequent intervals, so thatwe might take your oral temperature. . . .Day 2: During the day you will be accompaniedby a "watcher" who has two main tasks: 1) to takeyour oral temperature every two hours and 2) toensure that you do not sleep during the day.Despite the best intentions of our subjects, theyhave sometimes found it impossible to resist a day-time nap, and this causes problems in the analysisof our results since we don't know how much orwhat kind of sleep they have had. . . . The watchersystem, although a potential bother to you, seemsthe best way around this problem.Night 2: A night on which you will be awakenedonly two or three times, for the purpose of de-termining if, and what, you were dreaming beforethe awakening.Day 3: No watcher this day.Night 3: Another night like Night 1.Day 4: Another day like Day 2 (watcher).Night 4: Another night like Night 2.

Design

The design of the study followed the outlinecommunicated to 5s. Nights 1 and 3 were depriva-tion and pseudodeprivation nights. The 5s wererun in pairs, with a coin toss responsible for theassignment of one member of each pair to the orderdeprivation-pseudodeprivation (Order A), the othermember of the pair being assigned to the remainingtreatment order (Order B). Nights 2 and 4 were

EFFECTS OF "DREAM DEPRIVATION" 405

for the retrieval of reports of mental activity fromsleep, the dependent variable of interest. On thesenights, only two awakenings were made: one inNREM sleep, 10 min. after sleep onset; and one inREM sleep, 5 min. after the onset of the initial REMperiod of the night. The watchers' tasks on Days2 and 4, in addition to those mentioned to 5, in-cluded observation and recording (in a pocket-sizednotebook ostensibly for the recording of temperaturereadings) of S's waking behavior.

Oral temperature recordings were taken on Nights1 and 3 to justify the enforced arousal (3 min.) onawakenings on these nights. It was also the authors'impression that reports of previous research did notalways indicate adequate controls for S expecta-tion as a factor mediating dream-deprivation effects,and it was hoped that the oral temperature measure-ments might disguise the true purpose (real orsham selective sleep deprivation) underlying awaken-ings on Nights 1 and 3. In the Pivik and Foulkes(1966) study, such disguise was apparently success-fully achieved by having S perform a strength-of-grip task on deprivation and pseudodeprivationawakenings (Tebbs & Foulkes, 1966).

Sleeping arrangements and electrophysiologicalrecordings were identical to those in the Pivik andFoulkes study. Continuous physiological recordingswere taken on an Offner Type R dynograph on allnights, with four recording channels per S: twofor monopolar electro-oculograms (EOGs) from theouter canthus of each eye, one for a bipolar electro-myogram (EMG) from the submental region, andone for a bipolar EEC tracing (parietal-occipital).

Deprivation awakenings were made immediatelyfollowing the first REM observed in conjunctioneither with an EMG tracing of markedly diminishedamplitude compared to that observed during wakeful-ness or NREM sleep and/or with a low-voltagedesynchronized EEG (ascending EEG Stage 1).Pseudodeprivation awakenings were made duringany stage of NREM sleep except ascending EEGStage 2, the stage immediately preceding REM sleep.For the eight 5s run in Order A, every attemptwas made to match awakening number and patternon Night 3 as closely as possible to that which ob-tained on Night 1. For the eight Ss run in Order B,pseudodeprivation awakenings, of necessity, had tobe scheduled using other less exact criteria than anactual match with an S's own deprivation night.The criteria employed were matching with the de-privation night of the paired S with whom thepseudodeprivation 5 was being run and, followingexperience with the first few Ss, an impressionisticmatching with the number and pattern of deprivationawakenings required for those Ss.

On deprivation and pseudodeprivation awakenings,a light was turned on beside S's bed, and, simul-taneously, he was called over an intercom con-necting the control room with the sleeping room.An E then entered the room and took a 3-min.oral temperature reading, keeping S awake duringthis period with a combination of informal con-versation and, when needed, physical prodding.Following the 3-min. interval, S was allowed to

return to sleep. Total sleep time (TST) on depriva-tion nights was limited by clock time (a usualtermination of the experiment by 6 A.M., whereSs generally went to bed from 11:15 P.M. to 11:45P.M.) or by the point where deprivation awakeningswere becoming so frequent as to be obviously dis-agreeable to S or damaging to the possibility ofdisguising their purpose.

Watchers were three male graduate students inpsychology: one observed eight Ss, one seven andone one. They arrived at the laboratory in themorning of Days 2 and 4 to leave with S as hestarted his daily rounds. The same watcher wasassigned to both days for any particular S. Eachwatcher was instructed to keep a sufficiently closecheck upon S so that he was always certain, eitherby direct observation or by the nature of S'sscheduled activity (e.g., working with others), thatS could not obtain any daytime sleep. The Ss werefollowed to class, to the library, on fishing trips, andin other passive social situations in which thereseemed to be any possibility of their obtainingsleep. To the best of the authors' knowledge, no Sin the study slept at any time during Days 2 and 4.The watchers were, of course, uninformed as to theorder of treatments for Ss they observed.

Unknown to Ss, watchers completed a Gough Ad-jective Check List (ACL) following each day oftheir observation, indicating their impressions of Son that day. The watcher who observed only oneS failed to follow instructions in completing hischecklists, and, therefore, watcher checklist datawere available for only 15 of the 16 Ss. The Ssthemselves also completed the checklist after theirarrival at the laboratory on the evenings of Days2 and 4, describing how they felt on these same(posttreatment) days.

Nocturnal interviews on Nights 2 and 4 wereconducted from a room adjacent to the main con-trol room by an E who had not been present onNights 1 and 3 and who did not have access tothe physiological recordings on Nights 2 and 4,that is, by someone unaware of the treatment con-ditions for any particular S preceding any particularcontent retrieval night and of the awakening cate-gories on any such night. The interviews were con-ducted following a format identical to that em-ployed in the Pivik and Foulkes (1966) study.

Two of the authors, neither aware of the treat-ment variable associated with any content retrievalnight, scored the typescripts of the tape-recordednocturnal interviews on the same Df Scale employedin the Pivik and Foulkes study. Pearson product-moment reliabilities for these ratings were: post-deprivation-night REM reports, .85; postdepriva-tion-night NREM reports, .77; postpseudodepriva-tion-night REM reports, .99; and postpseudodepriva-tion-night NREM reports, .97. The numbers of Sreports on which total agreement was reached were,respectively, 12, 12, 13, and 14. Two-rater averageswere employed in all subsequent analyses of the Dfvariable.

Electrophysiological recording writeouts were scoredon the spot by one of the authors and were later

406 FOULKES, PIVIK, AHEENS, AND SWANSON

checked and rescored by another author for TSTand, following Dement's unpublished scoring manual,for REM time (the proportion of TST spent inREM sleep). One of the report raters, still un-informed as to the order of treatments administeredto any S, counted the proportion of 2i-sec. intervalscontaining one or more REMs (EM%) during theS min. of REM sleep preceding the content retrievalawakening from the first REM period (Ri) onNights 2 and 4 and also scored, in like fashion,EM% for the first 5 min. of the second, undisturbed,REM period (Ra) on these nights.

ACLs were scored for the number of adjectiveschecked; the proportion of the total number ofadjectives checked (rather than the absolute numberof items checked, to control for widely differing totalnumbers of adjectives checked—a range of 17-130)that fell upon the Defensiveness, Favorable, Un-favorable, Self-Confidence, Self-Control, Lability, andPersonal Adjustment Scales (Gough & Heilbrun,196S); and the ratio of unfavorable to favorableadjectives checked.

RESULTS

Sleep-Cycle Observations

Table 1 contains data from deprivationnights and pseudodeprivation nights on TST,REM time, and number of experimentalawakenings. Table 2 contains data from thepostdeprivation and postpseudodeprivation

nights on TST, REM time, EM.% for RIand R2, and latency to RI (minutes of NREMsleep preceding Rj).

It will be noted in Table 1 that the matchbetween treatment conditions is fairly goodfor TST and awakening number, particularlyfor Order A, where judgments did not haveto be made prior to the deprivation night asto S's probable reaction to REM-sleep de-privation. It is clear from Table 1 that therewere fairly considerable individual differencesin immediate sleep-cycle response to thedeprivation manipulation. (Sis required onlyeight REM-supressing awakenings for 307min. of TST, while 14 such awakeningswere required to suppress REM sleep duringonly 179 min. of TST for 56.) Such differ-ences introduced inevitable errors into at-tempts to run matched control nights in ad-vance of the deprivation night. Overall, how-ever, there was an average discrepancy of onlyIS min. of TST and of only one awakeningbetween the two treatments.

Deprivation-night REM-time values arereported in Table 1 according to two differentsets of criteria: one in which the onset of a

TABLE 1SLEEP-CYCLE OBSERVATIONS: TREATMENT NIGHTS

Subject

123456789

10111213141516

X

Order

BABABABABAABBABA

Deprivation night

'po'pi ib 1

309222266254238179247249202262227224256271196307

244

REM time

Stage 1

1.32.81.3.9

1.12.84.33.39.51.92.65.12.92.94.31.1

3.0

REM

1.0.7

1.1.5.5

2.4.7.7

1.9.6

2.32.5

.5

.52.3.5

1.2

No. experi-mental

awakenings

111110101014128

209

15148

10168

12

PspiiHnH^nr'vafinn n* ht

TST

298238282266258179288251250262231248250271260307

259

REM time

12.027.528.216.018.227.115.120.419.929.130.617.421.218.817.921.4

21.3

No. experi-mental

awakenings

111110101014108

109

14121210108

11

Note.—Abbreviations: TSTREM sleep.

> total sleep time; REM = rapid eye movement; REM time = proportion of TST spent in

EFFECTS OF "DREAM DEPRIVATION" 407

TABLE 2

SLEEP-CYCLE OBSERVATIONS: POSTTKEATMENT NIGHTS

Subject

123456789

10111213141516

X

Order

BABABABABAABBABA

Postdeprivation night

TST

367379309368399381379385349332341382306392381384

365

REM time

29.422.419.819.121.826.023.425.723.924.014.626.722.015.724.926.4

22.9

EM%

Ri

16.73.3

33.315.831.712,525.810.828.328.331.715.89.2

25.06.74.2

18.7

R>

30.019.256.716.716.755.020.824.216.715.833.328.315.825.814.240.8

26.9

LatencytoRi

5912920061

136118125855953

16667673

7488

93

Postpseudodeprlvatlon night

TST

367379309368399381379385349332341382306392381384

365

REM time

20.713.921.819.419.927.123.624.123.018.823.823.18.1

20.316.929.0

20.8

EM%

R!

16.718.335.010.86.7

37.514.213.313.313.329.229.29.2

12,518.37.5

17.8

R!

29.234.227.535.815.015.07.5

13.315.025.860.038.322.510.023.38.3

23.8

LatencytoRi

80153656466

1051165149696772

15414718678

95

Note.—Abbreviations: TST = total aleep time; REM — rapid eye movement; REM time = proportion of TST spent inREM sleep; EM%_ = interval containing one or more REMs; Ri — first REM period; Rz « second REM period; Latency toRi = min. of NREM sleep preceding Ri.

Postdeprivation versus postpseudodeprivatlon: REM time, ( = 1.45, p < .10, one-tailed; EM%, Ri, t =.28; EM%,Ri, * ~ .65; Latency to Ri, I - .11.

Pseudodeprivation REM time versus Postdeprivation REM time, t = .80; postpseudodeprlvation REM time, < = .27.

REM period is calculated, according to therules in Dement's unpublished manual, at theonset of EEG Stage 1, regardless of theabsence or presence of REMs, the other inwhich the onset of the REM period is calcu-lated at the first REM. The former method isemployed elsewhere in Tables 1 and 2 and isnow conventional within the field of sleep re-search. Values are reported for the latterprocedure since deprivation awakenings fol-lowed the initial REM criterion and sincethey help to account for the occasionally rela-tively large amounts of REM time (i.e., Stage1 time) achieved by Ss on deprivation nights(e.g., S9's 9.5%).

By either criterion, substantial deprivationwas achieved for all Ss. A comparison ofdeprivation-night and pseudodeprivation-nightREM times showed a REM-sleep deprivationnever less than 50% and averaging 86% bythe Stage 1 criterion. Further validation ofthe deprivation manipulation is seen in themean number of awakenings (12) required tosuppress REM sleep during an average TSTof 244 min. In the absence of deprivation,

only two, or possibly three, REM periodswould be expected during such a sleep period(or were observed during a comparable periodon the pseudodeprivation night).

Posttreatment nights were perfectlymatched for the number of content retrievalawakenings (two). By eliminating excess sleeptime on the longer of the two posttreatmentnights, it was also possible to match themperfectly for TST.8 Posttreatment-night ob-servations did not, however, show the expectedvalidation of the deprivation manipulation interms of greater REM time, increased EM%in RI or Rg, or decreased latency to RIfollowing true deprivation. In each case the

8 The procedure of eliminating excess sleep timemeans that not all laboratory sleep is reflected inTST totals in Table 2, a fact of some significance inevaluating the cumulative sleep-loss hypothesis dis-cussed below. It should also be noted that Sa's309 min. of TST on posttreatment nights does notreflect his true TST on those nights; due to equip-ment failure, recordings were not possible after 4:32A.M. on one of the posttreatment nights, and com-putations for both such nights are accordinglyabbreviated.

408 FOULKES, PIVIK, AHRENS, AND SWANSON

data were in the expected direction, but onlyin the case of REM time did they approachstatistical significance. In absolute terms, how-ever, mean REM times on pseudodeprivationnights, postdeprivation nights, and post-pseudodeprivation nights were quite similar,and posttreatment REM times did not differsignificantly from pseudodeprivation-nightREM times.

REM time was greater on Night 4 than onNight 2, both when postpseudodeprivationnights are compared (Night 2, 19.6% versusNight 4, 22.0%) and when postdeprivationnights are compared (Night 2, 21.7% versusNight 4, 24.0%). In both cases, however,two-tailed probabilities exceeded .20. Theslight trend to order effects, however, sug-gests another kind of comparison to validateREM deprivation: a cross-S comparison ofREM times of the eight 5s in Order A withthose of the eight 5s in Order B on bothNights 2 and 4. The data once again are inthe expected direction, but neither in theNight 2 comparison nor in the Night 4 com-parison did t exceed 1, one-tailed probabilitiesexceeding .15. Thus, with order effects heldconstant, no significant tendency for REMcompensation following deprivation was ob-served; that is, for example, 5s in Order A,with an average of 6 min. of REM time onNight 1, did not have significantly more REMtime on Night 2 than 5s in Order B, who hadan average of SO min. of REM time onNight 1.

Mean latency to RI was practically iden-tical following deprivation and pseudodepriva-tion. As brief bursts of REMs occasionallypreceded the first full-blown REM period ofposttreatment nights, the physiological record-ings were checked to see whether theseabortive REM episodes might have occurredsignificantly more often following deprivationthan following pseudodeprivation. They didnot (six times on postdeprivation nights,seven times on postpseudodeprivation nights).

Effects on Mental Content

Reports from REM sleep following depriva-tion were given an average rating on the DfScale of 5.41, those following pseudodepriva-tion, an average Df rating of 4.78. Nine Sshad higher Df ratings for REM reports fol-

lowing deprivation, six for those followingpseudodeprivation, with one 5 showing nodifference between the two conditions. Re-ports from NREM sleep following deprivationwere given an average rating on the Df Scaleof 3.34, those following pseudodeprivation,an average Df rating of 3.66. Six 5s hadhigher Df ratings for NREM reports follow-ing deprivation, eight for those followingpseudodeprivation, with two 5s showing nodifferences between the two conditions.Neither the REM nor the NREM mean dif-ference between deprivation and pseudo-deprivation reports was statistically significant(Wilcoxon matched-pairs signed-rank test).

REM compensating 5s (i.e., those withhigher REM times following deprivation thanfollowing pseudodeprivation) did not differsignificantly from REM noncompensating 5sin their distribution of relatively higher Dfratings to deprivation or pseudodeprivationnights. Neither the order of treatments (Aversus B) nor any checklist variable relatingto differences in daytime behavior precedingthe posttreatment nights was related towhether deprivation or pseudodeprivationREM and NREM reports were judged moredreamlike.

Effects on Waking Behavior

No ACL variable, either as scored from5's own checklist or from that completedby his watcher, successfully discriminated thedeprivation treatment from the pseudodepriva-tion treatment. When 5s were grouped asREM-time "compensators" or "noncompen-sators," however, several checklist variableswere found to be associated with the com-pensation-noncompensation variable.

Noncompensators tended to have a greaterproportion (in relation either to the totalnumber of adjectives checked or to the num-ber of favorable adjectives checked) of un-favorable terms checked following pseudo-deprivation, compensators, following truedeprivation (x

2 = 5.82, df=l, p< .02).Noncompensators tended to have a higherproportion of their checkmarks fall on thePersonal Adjustment Scale following truedeprivation, compensators, following pseudo-deprivation (x2 = 2.85, df=l, p< .10).Compensators, then, seemed to feel relatively

EFFECTS OF "DREAM DEPRIVATION" 409

more uncomfortable on the day followingdeprivation, noncompensators on the day fol-lowing pseudodeprivation. Watchers' check-lists tended to discriminate compensators fromnoncompensators on the same scales and inthe same directions as did Ss' own checklists(unfavorable ratio, x2 = 3.90, df = 1, p < .05;Personal Adjustment Scale, x2 = 2-80, # = 1,p < .10). The differential relative discomfortfelt by compensators and noncompensatorsfollowing deprivation apparently was alsoevident in the behavior they manifested inthe presence of their watchers.

The watchers also seemed to find the be-behavior of the compensators more varied orcomplex following deprivation than followingpseudodeprivation. They tended to check moreterms for compensators following deprivationthan following pseudodeprivation and moreterms for noncompensators following pseudo-deprivation than following deprivation (x2 =7.78, d}= 1, p< .01).

Several effects of the experimental pro-cedures upon waking behavior were clearboth to Bs and to the watchers, but werenonspecific (i.e., as likely to occur followingpseudodeprivation as following deprivation).First, there was resentment at being"watched." A definite effect was seen in onlya few 5s, but others were undoubtedly nottotally immune. It was seen in extreme formwith 52, who, from the very outset of theexperiment, impressed the authors as havingsociopathic tendencies. Following his pseudo-deprivation night, he commented that hecould put up with the electrode attachments,the frequent awakenings, etc., but he couldnot stand another day with his watcher. Thewatcher commented wistfully that never be-fore had he been so thoroughly hated on thebasis of so little social interaction. Observa-tions such as this suggest that, for certainpersonality types, the experience of beingunder constant surveillance is quite threaten-ing and may be a contributing factor to thedramatic changes in personality that havebeen observed subsequent to REM-sleepdeprivation with surveillance (Dement,196Sb). It would seem, therefore, that con-trols for the effects of surveillance may benecessary in experiments on the deprivation ofREM sleep.

A second nonspecific effect of the experi-mental procedures was that most Ss arrivedat the laboratory following Days 2 and 4 in afatigued state that far surpassed the ordinaryfatigue experienced toward bedtime. This ef-fect seemed particularly pronounced followingDay 4, irrespective of the immediately preced-ing treatment. Checklist data confirmed thisimpression: both watchers and 5s themselveschecked a greater proportion of unfavorableadjectives on Day 4 than on Day 2. Com-bining order and treatment effects, the ob-served ranking of conditions in which 5s in-cluded a greater proportion of unfavorableterms among all those checked was post-pseudodeprivation, Night 4 (« = 6); post-deprivation, Night 4 (n = 4); postpseudo-deprivation, Night 2 (n = 3); and post-deprivation, Night 2 (n = 1). The appearanceof order effects in the checklist results arguesstrongly for both control nights and a counter-balancing of such nights with deprivationnights—controls that have not been employedconsistently in previous dream-deprivationresearch.

Effects on Oral Temperature

Oral temperature readings were taken at2-hr, intervals during the days followingdeprivation and pseudodeprivation nights.Oral temperature was not a variable of majorinterest in the design of the study, but wasused, rather, as a means of disguising themanipulations and behavioral observations.Readings were taken on standard clinicalthermometers, with 5 having engaged in nostrenuous physical activity for the 5 min.preceding and being seated during the 3-min.interval. The authors have no guarantee thatthe conditions of observation were preciselycomparable from day to day and from 5 to 5,and the thermometers had not been pretestedfor reliability or comparability (although thesame one was employed, unless, as sometimeshappened, it was broken, on both days for thesame 5). Observations on oral temperatureshould be considered merely suggestive, then,rather than conclusive.

Mean oral daytime temperature was lowerfollowing deprivation than following pseudo-deprivation, a difference of .23° F (97.74° Fversus 97.97° F) that was significant

410 FOULKES, PIVIK, AHEENS, AND SWANSON

statistically (t = 2.56, p < .05, two-tailed).Considered individually, 12 of IS 5s for whomcomplete data were available had a lower oraltemperature following deprivation than follow-ing pseudodeprivation; 6 were REM-timecompensators and 6, REM-time noncompen-sators. Neither the magnitude of the tem-perature discrepancy between treatment nightsnor the absolute temperature values obtainedover both nights significantly discriminatedREM-time compensators from REM-timenoncompensators.

DISCUSSION

Sleep-Cycle Observations

The failure of REM deprivation to produceREM-compensation phenomena at statisticallysignificant levels on the postdeprivation nightmight be attributed to two factors: cumula-tive sleep loss over the course of the 4 experi-mental nights, and individual differences inreaction to REM deprivation, which, underminimal deprivation, may have generated in-ter-S variability encompassing the point ofzero compensation.

Cumulative sleep loss. William Dement hassuggested to the authors that their failure toconsistently observe REM compensation fol-lowing deprivation may be attributed tocumulative sleep loss over the 4 nights ofthe experiment. On this hypothesis, Night 1for 5s in both orders was, due to abbreviatedTSTs, a night of both REM- and NREM-sleep deprivation. The 5s in Order A failed tocompensate on Night 2 due to the greaterrelative potency of the generalized sleepdeprivation than of the experimental REMdeprivation. (Berger and Oswald, 1962a, haveshown that total sleep deprivation leads toincreased amounts of NREM sleep on a firstpostdeprivation night, REM compensationoccurring only on the subsequent night.)Night 2 is viewed, for both groups of Ss, asadding slightly to cumulative generalized sleepdeprivation, arid Night 3, much as is true ofNight 1, is again a night of both NREM- andREM-sleep deprivation. By Night 4, it isassumed that REM deprivation was morepotent than NREM deprivation, and that 5s"compensated," leading to results in the pre-dicted direction for 5s in Order B and to

results in the nonexpected direction for Ss inOrder A.

This hypothesis is consistent with severaltrends in the present data:

1. More Order B than Order A 5s showedpostdeprivation REM compensation. This dif-ference, however, was not statistically sig-nificant.

2. There was a tendency to greater REMtime on Night 4 as compared to Night 2,treatments held constant. This difference,however, did not approach statistical signifi-cance.

3. There was also a tendency (t = 2.02,p < .10, two-tailed) for pseudodeprivation-night REM time to be higher (£ = 23.9%)on Night 3 than on Night 1 (£ = 18.7%),as would be predicted by the hypothesis thatREM compensation induced by the firsttreatment was delayed in its execution untilthe second half of the experiment.

The sleep-loss and delayed-compensationhypotheses, however, cannot totally explainthe failure to consistently observe REM-timecompensation.

1. The 5s were encouraged to take morningor afternoon naps on Day 3, the day on whichthey were not under surveillance, to minimizeany carry-over from one treatment to theother. Most 5s' daytime schedules permittedsuch naps. The 5i4 reported that he sleptfor more than 5 hr. on the morning of Day 3.This S hr. plus his 11 hr. of sleep in thelaboratory on Nights 1 and 2 should havegiven him his full share of sleep, both NREMand REM, before entering the second phaseof the experiment; yet he still showed moreREM time following pseudodeprivation thanfollowing deprivation.

2. Six months after the completion of their4 nights of laboratory service, 56 and 5iereturned to the laboratory for 2 additionalnights: pseudodeprivation and postpseudo-deprivation. The new Night 1 pseudodepriva-tion nights matched the original Night 3pseudodeprivation nights for TST and numberof awakenings, the new Night 2 postpseudo-deprivation nights, the original Night 4pseudodeprivation nights for TST. Both 5s'Night 2 postpseudodeprivation REM timeswere slightly lower than their Night 4 values(24.6% versus 27.1% for 5e, 27.1% versus

EFFECTS OF "DREAM DEPRIVATION" 411

29.0% for Sie), but both were still quitehigh, suggesting that their originally highpostpseudodeprivation values were more stableS characteristics than products of the orderin which treatments were administered. TheSe, originally a hairline noncompensator, didbecome an equally tenuous compensator; Sie,on the other hand, remained a noncompen-sator. These results cast doubt on the hy-pothesis that the placement of pseudodepriva-tion nights in the experimental sequence (orcumulative sleep loss) can explain the origi-nally observed failures to compensate.

3, There was no significant difference inREM compensation on Night 2 betweenOrder A and Order B Ss, who, with equivalentTSTs, had widely differing REM times onNight 1 (.£ = 6 min. and SO min., respec-tively). The cumulative sleep-loss hypothesissupposes that Ss with an average of 240 min.of NREM-sleep time and only 6 min. ofREM-sleep time (mean figures for the depri-vation night for eight Order A 5s) will have aconsiderably greater "need" on a recoverynight for NREM sleep than for REM sleepand hence will show no compensation. On a7-hr. TST base line, however, with the 24%REM-time base line appropriate for youngadult males (Williams, Agnew, & Webb,1964), the deprivation night gave these 5s75% of their nocturnal quota of NREMsleep, but only 6% of their quota of REMsleep. Even in terms of actual minutes of TSTto be "made up," the figures would be: REMsleep, 95 min.; NREM sleep, 79 min. To as-sume a complete priority for NREM-sleepcompensation in such circumstances is toassume an extremely low importance forREM sleep.

4. The cumulative sleep-loss hypothesisdoes not explain why three Order A 5s didcompensate on Night 2. Their average TSTon Night 1 was 244 min., while that of OrderA noncompensators was 248 min. The au-thors take this evidence as indicating that, inaddition to the effects of cumulative sleep loss,there are also important individual differencesin reaction to REM deprivation, differenceswhich help to account for the failure to ob-serve consistent REM compensation followingshort-term REM deprivation.

Individual differences. Evidence for the

existence of individual differences may befound in previous studies of dream depriva-tion. The 5 N.W. in the first deprivationexperiment (Dement, 1960) showed no sig-nificant increase in REM time on any of 5recovery nights following S consecutive nightsof REM deprivation. In a more recent in-vestigation (Dement, Greenberg, & Klein,1966), a first recovery night REM-time eleva-tion of 54% was observed in one 5, while thecomparable figure for another 5, run underidentical conditions, was only 13%. Theseresults, after 19 nights of partial REM de-privation, suggest that under considerablylesser deprivation (e.g., 1 night) compensa-tion might or might not be observed, depend-ing upon 5. Cartwright, Monroe, and Palmer(1967) have recently failed, with conventionalscoring of sleep stages, to find a significantREM-time compensation following 3 con-secutive nights of REM deprivation. Theysuggested, as do the data of the present study,that there may be significant individual dif-ferences in reaction to REM-sleep loss.

Strictly speaking, discussion of such in-dividual differences may only be justified byfindings of intraindividual consistency in thetendency to greater or lesser compensationfollowing deprivation. It should also be bornein mind that compensation-noncompensationdifferences in the present study may merely re-flect situational factors, such as sleep ex-periences on the few nights preceding 5'sexperimental service. The present authorshave no reason to believe this is so, butneither they nor Cartwright et al. (1967) ranadaptation nights immediately preceding thoseof laboratory service.*

In spite of these qualifications, the data ofthe present study do suggest the desirabilityof more attention in future deprivation re-search to factors underlying differences inreaction to REM-sleep deprivation. Amongfactors seemingly ruled out in the presentstudy is the repression-sensitization variable,since all 5s were repressers. Among factorsdeserving further scrutiny on the basis of the

* Neither Night 1 REM time for Ss in Order Bnor latency to first REM period in Order A showedthe "first night effect," a REM-time depression thatis generally considered to justify such adaptationnights.

412 FOULKES, PIVIK, AHRENS, AND SWANSON

present data is relative psychological dis-comfort experienced on postdeprivation days.

There is every reason to believe that in-dividual differences will be as significant indeprivation research as elsewhere in psy-chology and psychiatry and that their in-vestigation may be of crucial significance inunderstanding the nature of the REM-depriva-tion manipulation. The discovery, for instance,of just one consistently noncompensating S,one who behaves in the fashion of Dement's5 N.W. in replicated long-term deprivationstudies, would immediately call into questionalmost all of the speculations heretofore madeon the functional significance of REM sleep.

Methodological problems of the deprivationexperiment. If the cumulative sleep-loss ex-planation is meant to imply that a factorforeign to REM deprivation itself has in-advertently been injected into the depriva-tion experiment, the present authors must re-ject this implication. In their experience, gen-eralized sleep deprivation must be an integralpart of any attempt to effectively deprive Ssof REM sleep, if Ss are to observe their nor-mal bedtime routine. Awakenings soon beginto accumulate at such a rate that very littlesleep of any type is possible, such is the"pressure" to experience REM sleep. TheSis, for instance, accumulated only 18 min.of TST between his seventh deprivationawakening and his sixteenth.

The influence of generalized sleep depriva-tion in previous dream-deprivation experi-ments is difficult to assess. In Dement's(1960) first study, deprivation-night TSTswere not reported. The same is true of the ex-perimental reappraisal of Dement's work byKales, Hoedemaker, Jacobson, and Lichten-stein (1964). TSTs are reported for two ofthe three Ss in Dement's (196Sb) more recentand extended deprivation studies: for thefirst S, deprivation-night TST averaged only252 min.; for the second, mean deprivation-night TST was only 320 min. For the thirdS, it was reported that his NREM-sleep timeon 11 nights when deprivation was achievedby dexedrine (which suppresses REM sleep)in conjunction with awakenings "was equalto, or above, the total sleep time on the base-line nights [p. 593]." Presumably this wasnot true on the S nights when deprivation was

achieved solely by the methodologicallycleaner method of experimental awakenings,for the number of interruptions per hour ofsleep was markedly greater under the lattercondition.

It must be acknowledged, then, that theincreasing pressure toward REM sleep duringdeprivation nights—a consistent finding indeprivation studies—does serve to promoteboth generalized sleep deprivation andNREM-sleep loss. As a consequence, it be-comes impossible to achieve the goal of suchexperiments, the abolition of REM sleep inan otherwise intact S. The authors are in-clined to agree with the remark of Ephronand Carrington (in press), made in a some-what different context: "It does not seemplausible to us that REM deprivation, whichinvolves disruption of a profound physiolog-ical rhythm, can result in other than a trau-matized subject [p. 216]." Part of suchtrauma undoubtedly comes from the concomi-tant disruption of NREM sleep which issufficiently interconnected with REM sleep tomake it impossible to manipulate either inisolation of the other.

The mounting number of awakenings in thenight-long deprivation experiment also posesa further problem: it is no longer possible (asit apparently was in the early-night depriva-tion study of Pivik & Foulkes, 1966) todisguise the purpose of the awakenings. Evenin the relatively unsophisticated locale inwhich this experiment was conducted, manySs were familiar with the now well-publicizedexperiments on dream deprivation. Indeed,one of the comments now encountered fromprospective Ss, in the absence of any localpublicity of the authors' own limited researchin this area, is, "Is this going to be a dream-deprivation experiment?" Furthermore, amongSs apparently not acquainted with prior re-search on dream deprivation, suspicion aroserapidly toward the end of treatment nightsthat awakenings were being made for purposesother than oral temperature readings. The S7,for instance, commented, after the ninthawakening of his pseudodeprivation night,"If you do this often enough, a fellow couldlearn how not to dream."

That this last comment came on a nightof frequent NREM awakenings suggests

EFFECTS OF "DREAM DEPRIVATION" 413

several interesting conclusions. First, NREMawakenings may also, in view of the existenceof moderately dreamlike mentation duringNREM sleep, be a form of dream interrup-tion. Second, the series of awakenings onpseudodeprivation nights may provide anadequate control for the factor of S expecta-tion. Third, frequent awakenings from sleep,in and of themselves, may produce S sets thatmight account for some of the observedsequelae of dream deprivation in experimentslacking adequate controls for the number ofinterruptions from sleep.

In these comments, and earlier ones on thedesirability of controls for surveillance andof counterbalanced experimental and controltreatments, the present authors do not wishto seem unappreciative of the magnificentlysuggestive research of Dement and others. Asrelative strangers to the field, however, theauthors were impressed by the extent to whichthe deprivation experiment is fraught withperils of both design and interpretation. Nothaving been prepared for this by reading inthe area, it was felt to be important that theseimpressions be recorded here.

Effects on Mental Content

The failure to observe any significant effectof 1-night REM deprivation upon REM con-tent of the subsequent night might be at-tributed to the choice of the first REM periodfor the elicitation of a postdeprivation REMreport. The first REM period of a night gen-erally is associated with rather brief andvague dream imagery (Foulkes, 1966). Butthese relatively low base-line qualities of thecontent typically reported from the initialREM period of the night would make thisREM period an especially good setting inwhich to observe any hypothesized intensifi-cation of mental imagery during sleep and inwhich to judge whether REM mental contentappeared to be more nearly related to theprevious experience of dream deprivation orto the type of content typically reportedfrom comparable REM periods in the absenceof deprivation.

There is, moreover, inferential evidence thatthe second REM period of postdeprivationnights did not produce REM content sig-nificantly more intense than that of the second

REM period of postpseudodeprivation nights.Eye-movement rate was not significantlygreater during the second REM periodfollowing deprivation than in that follow-ing pseudodeprivation. In the Pivik andFoulkes (1966) study, significant intensifica-tion of eye-movement activity accompaniedsignificant intensification of dream content.

There was one important difference in inter-view procedure between this study and thatof Pivik and Foulkes. In the latter, the inter-viewer knew whether the content retrievalawakening that he was making was precededby the deprivation or the pseudodeprivationcondition. In the present study, the inter-viewer was naive as to the condition as-sociated with the awakenings that he wasmaking.

The authors do not feel, however, that thisdifference in procedure explains the differingresults of the two studies. In the Pivik andFoulkes study, preawakening eye-movementintensity discriminated postdeprivation frompostpseudodeprivation REM reports; that is,there was an index showing intensificationof dream content that was independent of,and prior to, the interview. Failure to observeeye-movement intensification here is taken asindicating that the difference in results be-tween the two studies relates to a difference insleep phenomena rather than in interviewbehavior.

There was no evidence in the present resultsof a compensatory increase in NREM dream-ing that might help to discharge any drive todream in advance of the content retrievalREM period nor were there indications ofchanges in daytime functioning that mightindicate such discharge in advance of therecovery night. There is the possibility that,on deprivation nights themselves, the hy-pothetically accumulating pressure to dreammight have found release during NREM sleep.This, however, does not seem to be probable.Toward the end of deprivation nights so littleNREM sleep accumulated that such sleepwould seem incapable of psychic dischargethat normally requires more than an hourof the highly activated REM phase.

Pivik and Foulkes (1966) noted that theirresults were consistent with Freudian theory,in particular with the concept of a drive

414 FOULKES, PIVIK, AHRENS, AND SWANSON

toward dream expression. The present resultsare clearly quite difficult to reconcile withsuch a formulation. They seem more nearlycompatible with the position that there is nopsychological necessity underlying the periodicrecurrence of periods of dreaming (REM)sleep.

If psychological necessity does not underliethe periodic recurrence of dreaming, how isone to explain dream intensification followingdeprivation within a single night? There areclearly other explanations of this finding thanthat suggested by Freudian theory. One might,for example, hypothesize that the nature ofthe dream is determined by the nature of theREM period. Where deprivation producessignificant physiological compensation, as forPivik and Foulkes' (1966) repressers, thedream consequently is also more "activated";where deprivation fails to effect physiologicalcompensation at statistically significant levels,as for Pivik and Foulkes' sensitizers and for5s in the present study, the dream con-sequently also shows no change.

The Pivik and Foulkes experiment was con-ducted to determine if REM-sleep deprivationcould also justifiably be considered "dreamdeprivation," and it provided a tentative posi-tive answer to this question. On the basis ofthe findings of this study, the present au-thors are now inclined to answer this question,still tentatively, in the negative.

This answer may appear equivocal, in viewof the failure to observe significant physiolog-ical compensation. However, if there is such athing as dream deprivation over and above,and partially independent of, REM-sleep de-privation, the failure of physiological com-pensation provides an ideal situation for itsdemonstration. Sixteen Ss failed, over thecourse of a full night, to achieve a single fullyelaborated dream of the type normally ac-companying REM sleep, of the type mostdiscussed by psychoanalysts and other dreamtheorists. When, on the next night, theyentered a period of dreaming sleep, thesedreams should have been more intense than,or in some way different from, those follow-ing a night of sleep in which several fullyelaborated dreams were experienced. Theywere not. They were the kinds of dreams wellpredicted by the nature of the physiological

phenomena characterizing that REM period,but not well predicted by the nature of Ss'recent experiences of dreaming or inhibiteddreaming. In this respect, it appears thatdream deprivation is a follower rather thana leader in the chain of events set in motionby interference with REM sleep.

Effects on Oral Temperature

It is both ironic and significant that, in astudy designed to examine psychological ef-fects of "dream deprivation" and in which asole physiological variable was deployed todisguise Es' purpose, this one variable alonedifferentiated deprivation from pseudodepriva-tion. As has been stated above, the authorsview the finding of lowered oral temperaturefollowing deprivation as suggestive, ratherthan conclusive, but it is felt that it invitesindependent attempts at replication.

Murray, Williams, and Lubin (1958) havereported lowered body temperature followinggeneralized sleep deprivation. Should thepresent findings of a greater decrease in bodytemperature following REM-sleep depriva-tion prove replicable, one would have furtherevidence of physiological difference betweenthe two types of sleep (and, perhaps, somesuggestions as to their different metabolicfunctions). It would also further indicate thepredominantly physiological, and only in-directly behavioral, nature of REM-sleepdeprivation.

REFERENCES

BEROER, JR., & OSWALD, I. Effects of sleep deprivationon behaviour, subsequent sleep and dreaming.Journal of Mental Science, 1962, 108, 457-465. (a)

BEEOER, R., & OSWALD, I. Eye movements duringactive and passive dreams. Science, 1962, 137,601. (b)

BYRNE, D., BARRY, J., & NELSON, D. Relation ofthe revised Repression-Sensitization scale to mea-sures of self-description. Psychological Reports,1963, 13, 323-334.

CARTWRIOHT, R., MONROE, L. J., & PALMER, C.Individual differences in response to dream depriva-tion. Archives of General Psychiatry, 1967, 16,297-303.

DEMENT, W. The effect of dream deprivation.Science, 1960, 131, 1705-1707.

DEMENT, W. An essay on dreams. In T. M. New-comb (Ed.), New directions in psychology. Vol.2. New York: Holt, Rinehart & Winston, 1965. (a)

EFFECTS OP "DREAM DEPRIVATION" 415

DEMENT, W. Studies on the function of rapid eyemovement (paradoxical) sleep in human subjects.In, Aspects anatomo-fonctionnels de la physi-ologie du sommeil, Paris: Centre National de laRecherche Scientifique, 1965. (b)

DEMENT, W., GREENBERO, S., & KLEIN, R. Theeffect of partial REM sleep deprivation and de-layed recovery. Journal of Psychiatric Research,1966, 4, 141-152.

EPHRON, H., & CARRINOTON, P. S. Ego functioningin rapid eye movement sleep: Implications fordream theory. In J. Masserman (Ed.), Scienceand psychoanalysis. Vol. 9. New York: Grune &Stratton, in press.

FOULKES, D. The psychology of sleep. New York:Scribner, 1966.

FOULKES, D., SPEAR, P. S., & SYMONDS, J. D. In-dividual differences in mental activity at sleeponset. Journal of Abnormal Psychology, 1966, 71,280-286.

FREUD, S. The interpretation of dreams. New York:Basic Books, 1956.

GOUGH, H. G., & HEILBRUN, A. B. The AdjectiveCheck List manual. Palo Alto, Calif,: ConsultingPsychologists Press, 1965.

JOUVET, M. Etude de la dualite des ctats de sommeilet des mecanismcs de la phase paradoxale. In,Aspects anatomo-fonctionnels de la physiologic dusommeil. Paris: Centre National de la RechercheScientifique, 1965.

KALES, A., HOEDEMAKER, F. S., JACOBSON, A., &LICHTENSTEIN, E. L. Dream deprivation: Anexperimental reappraisal. Nature, 1964, 204, 1337-1338.

MURRAY, E. J., WILLIAMS, H. L., & LUBIN, A. Bodytemperature and psychological ratings during sleepdeprivation. Journal of Experimental Psychology,1958, 56, 271-273.

PIVIK, T., & FOULKES, D. "Dream deprivation": Ef-fects on dream content. Science, 1966, 1S3, 1282-1284.

TEBBS, R. B., & FOULKES, D. Strength of gripfollowing different stages of sleep. Perceptualand Motor Skills, 1966, 23, 827-834.

WILLIAMS, R. L., AGNEW, H. W., & WEBB, W. B.Sleep patterns in young adults: An EEG study.Electroencephalography and Clinical Neurophysi-ology, 1964, 17, 376-381.

(Received February 27, 1967)