International Journal of Psychophysiology 74 (2009) 14–18

Contents lists available at ScienceDirect

International Journal of Psychophysiology

j ourna l homepage: www.e lsev ie r.com/ locate / i jpsycho

Early-night serial awakenings as a new paradigm for studies on NREM dreaming

Valdas Noreika a,⁎, Katja Valli a,b, Hetti Lahtela a, Antti Revonsuo b,a

a Centre for Cognitive Neuroscience, Department of Psychology, University of Turku, Turku, Finlandb School of Humanities and Informatics, University of Skövde, Skövde, Sweden

⁎ Corresponding author. Centre for Cognitive Neuroscogy, University of Turku, Assistentinkatu 7, FI-20014, Fifax: +358 2 333 5060.

E-mail address: valnor@utu.fi (V. Noreika).

0167-8760/$ – see front matter © 2009 Elsevier B.V. Aldoi:10.1016/j.ijpsycho.2009.06.002

a b s t r a c t

a r t i c l e i n f o

Article history:Received 18 December 2008Received in revised form 15 June 2009Accepted 24 June 2009Available online 9 July 2009

Keywords:Serial awakeningsPsychophysiology of dreamingComplexity of dreamsDreamless sleepNREM sleepEEG

A new experimental paradigm called “Early-Night Serial Awakenings” (ENSA) was explored to find out itsstrengths and weaknesses for psychophysiological studies of NREM sleep dreaming. Five participants spent20 experimental nights in the sleep laboratory, and were serially awakened with approximately 24-minuteintervals during Stages 2 and 3 of NREM sleep. As a total, 164 awakenings were conducted during the sessionsthat lasted on average 193 min. Altogether, 30% of NREM sleep awakenings led to dream reports, 39% toreports of white dreaming, and 31% to reports of dreamless sleep. Results also show that sleep EEG spectralpower, dream recall frequency as well as dream complexity remained stable throughout the serial awakeningsessions. We conclude that, as ENSA dreams appeared to be static and very limited in content, the paradigmwe identified could be used in future studies to reveal the psychophysiological mechanisms of relativelysimple forms of early-night NREM sleep dreaming.

© 2009 Elsevier B.V. All rights reserved.

1. Introduction

Sleep-laboratory studies on dream content and neural correlates ofdreaming are time-consuming and expensive. Typically, the wholenight is spent trying to elicit one or at most a few reports, whereasstatistical analyses require at least some tens if not hundreds of dreamreports. This problem could be partially overcome by using serialawakenings. In traditional serial awakenings, the participants areawakened once during each cycle of sleep, and this way four or fivereports are collected during a single night (Bertolo et al., 2003;Wittmann et al., 2004). In the present study we explored whether it ispossible to extend the paradigm of serial awakenings further to collecta large number of NREM reports (up to about 10) in a single, relativelybrief session, and repeat this procedure over a period of several non-consecutive nights. We call this new paradigm “Early-Night SerialAwakenings” (ENSA). A somewhat similar paradigmwith 45 awaken-ings per night has been successfully used to study EEG correlates ofsleep onset hypnagogic imagery (Hori et al., 1994), but, to our bestknowledge, ENSA approach has never been applied to study Stage 2 or3 NREM sleep dreams.

Obviously, ENSA interferes with the usual sleep structure. Thus, it isquestionable whether the data obtained with this paradigm is valuablein psychophysiological studies of dreaming. It may be that serialawakenings which quickly follow each other will significantly changesleep physiology as reflected by EEG, or that it might significantly affect

ience, Department of Psychol-nland. Tel.: +358 2 333 6984;

l rights reserved.

dream recall in someway, or that it would be too uncomfortable for thesubjects because of accumulating tiredness and sleep pressure, orbecause of difficulties in falling quickly back to sleep and rapidly wakingup to give a report over and over again in the same session. Thus, weexplored whether ENSA is comfortable for the subjects and whetherENSAproduceshomogeneous samples of data, inparticular,whether theuse of numerous serial awakenings adversely affects dream recallfrequency and sleep EEG activity within one night and therefore wouldmake ENSA unusable for psychophysiological dream research.

Our aim was to answer the following questions concerning theutility of ENSA: (1) How many awakenings can be conducted duringone experimental session without exhausting the participant andrendering dream recall unreliable? How long does it take to conductup to 10 awakenings during a single session? (2) Do serial awakeningsconducted within a session provide homogeneous sleep EEG, i.e. canwe avoid spectral power changes during the ENSA session? (3) Doserial awakenings within and across laboratory nights affect dreamrecall frequency? With respect to the mental contents reported, thefollowing questions were explored: (4) How complex are the dreamsobtained with ENSA from different stages of NREM sleep, as measuredby the modified version of Orlinsky's (1962) scale? (5) What is theoptimal stage of early-night NREM sleep for obtaining similarproportions of dream reports and reports of dreamless sleep?

2. Methods

2.1. Participants

Five volunteer unpaid students (4 females) gave their informedconsent to participate in the study (age: M=23.7, SD=1.51). Each

15V. Noreika et al. / International Journal of Psychophysiology 74 (2009) 14–18

participant spent one adaptation night and four non-consecutiveexperimental nights in the sleep laboratory, with at least two nights ofhome sleep in-between experimental nights, so that possible sleepdeprivation effects would not accumulate from night to night. Theparticipants were awakened until they expressed the wish to quit thesession, but not more than 12 times per night, so that the ENSAwouldnot extend to the late-night awakenings.

2.2. Procedure

Polysomnography recording included two EOG and four EEGelectrodes (C3, C4, O1, O2), which were positioned using a standard10-20 system and referenced to the right ear mastoid. The EEG signalwas amplified (SynAmps Model 5083), recorded and sleep-scoredusing NeuroScan equipment and software. Awakenings were per-formed during the first NREM Stages 2 and 3 after falling asleep(Rechtschaffen and Kales, 1968), and during a transitional period thatwas defined by the first appearance of a slow wave, when anotherslowwave or waves were seen in at least one of the two following 20 sepochs. An EEG wave was accepted as a slow wave only if it had atleast 75 µV amplitude from peak to peak, lasted for at least 500 ms,and did not include a positive component typical for K-complex.Participants were instructed in advance to give a free report of‘everything that was going through their mind before awakening’immediately after being awakened by a sound signal. If a reportcontaining perceptual experiences was given, the participant wasconsequently presented with a set of 18 questions by a recorded malevoice, including inquiries about objects, feelings, self, and the durationof the dream. Negative recall reports were followed by 3 questionsconcerning the subjective certainty of dreamless sleep. Finally, a freeinterview was conducted, if necessary, to clarify ambiguous parts of areport. Participant and experimenter stayed in separate rooms andcommunicated through a digital audio system.

2.3. Variables

Analysis was carried out with three groups of independentvariables that were based on the timing of awakenings: 1) stages ofsleep (Stage 2, transition, Stage 3); 2) nights of experiments (first twonights, last two nights), and 3) awakenings (median split into earlyENSA: 1–4, N=80, late ENSA: 5–12, N=84). Each post-awakeningreport was scored according to the following recall categories:dreamless sleep, if no perceptual experience was recalled andreported after awakening; white dream, if the participant remem-bered having a dream, but did not recall any specific content; ordream, if the participant remembered and reported at least oneperceptual experience. Furthermore, the content of dream reports wasclassified according to the perceptual complexity categories ofOrlinsky's Modified Scale (see Table 1). Subjective reports were finallygrouped according to: 1) recall categories (dreamless, white dream,

Table 1Orlinsky's modified scale for perceptual complexity of dreams.

Categories of perceptual complexity

1. Participant remembers a specific topic but in isolation: a fragmentary percept,unrelated to anything else.

2. Participant remembers several unrelated perceptual experiences.3. Participant remembers several interconnected perceptual experiences.4. Participant remembers a short but coherent dream, the parts of which seem to beimmersed within a unified scene.

5. Participant remembers a coherent dream, in which something happens within aunified scene, i.e., one perceptual experience is replaced by another.

6. Participant remembers a long, detailed dream in which the whole scene is replacedby other scenery.

7. Participant remembers a long, detailed dream inwhich thewhole scene is replaced byother scenery more than once.

Adapted from Orlinsky (1962).

dream) and 2) complexity of dream reports (simple: Orlinsky'scategories 1–3, complex: Orlinsky's categories 4–7).

2.4. Statistical analyses

Inter-rater agreement for sleep scoring and dream report char-acteristics was evaluated with Cohen's kappa coefficient test, andwhen applicable, aweighted quadratic kappa test was used. To explorethe effects of independent variables (stages of sleep, nights ofexperiments, awakenings) on the recall and complexity categories,within-subject percentages of the respective dependent categorieswere first calculated for each of the independent condition. Forinstance, participant 4 provided 19 NREM sleep reports during the firsttwo nights (dreamless: 7, white dream: 8, dream: 4), which wereconverted to the respective percentages (dreamless: 36.84%, whitedream: 42.11%, dream: 21.05%). Then, due to the small sample size andthe likelihood of Type II error, several one-way repeated measureanalyses of variance (ANOVA) tests were carried out. Friedman'sANOVA test was used as a non-parametric alternative, when Shapiro–Wilk test indicated significant deviation from normal distribution ofdata. Paired samples t tests were carried out for comparisonswith onlytwo conditions of independent variable. Spearman's rank ordercorrelation was used to evaluate possible relationship between thefirst eight serial awakenings and standard sleep parameters, such assleep latency and duration of sleep Stages 1 and 2. Sleep latency wasdefined as EEG recoding duration until the first uninterrupted sleepperiod, and, for later awakenings, as an interval between an awakeningand a next sleep period. Analysis of duration of NREM sleep Stages 3and 4 was not possible because ENSA does not permit genuine sleepcycles, i.e. participants were always awakened latest in Stage 3.

2.5. EEG analyses

All pre-awakening EEG recordings of 60 s length were visuallysleep-scored in three separate 20 s epochs. Recordings with the samesleep stage in all epochs were processed for a quantitative EEGanalysis (N=118), which was carried out using the EEGLab (v6.01b)and the Brain Vision Analyzer (v1.05.0002) programs. All recordingswere grouped together according to the sleep stage and theawakening variables (Stage 2: N=37, transition: N=39, Stage 3:N=42; early ENSA: N=58, late ENSA: N=60). A low cutoff of 0.5 Hzand a high cutoff of 32 Hz were applied to continuous EEG data, whichwas later downsampled to 64 Hz, and segmented in equal sizedsegments of 2 s. Automatic artifact rejection function was used toremove segments that did not fulfill the requirements of at least onecriterion as follows: 1) gradient criterion (maximal allowed voltagestep/sampling point: 35.00 µV); 2) difference criterion (maximalallowed absolute differenceof twovalues in the segment: 200.00µV); 3)amplitudecriterion (minimal allowedamplitude:−150.00µV,maximalallowed amplitude: 150.00 µV); and 4) low activity criterion (lowestallowed activity: 1.50 µV, interval length: 100.00 ms). 2402 artifactfree segments were Fast Fourier Transformed (Hanning window, 10%),and the mean spectral power (µV2) was calculated for five frequencybands (delta: 0.5–4.0 Hz, theta: 4.0–8.0 Hz, alpha: 8.0–12.0 Hz, beta1:12.0–16.0Hz, beta2: 16.0–32.0Hz). Separate repeatedmeasures ANOVAtestswere carried out for each stage of sleep (Stage 2, transition, Stage 3)with three variables per test: 1) electrodes (C3, C4, O1, O2); 2) bands(delta, theta, alpha, beta1, beta2); and 3) awakenings (early ENSA,late ENSA).

3. Results

3.1. Awakenings

A total of 164 awakenings were performed during 20 experimentalnights with an average of 8.2 awakenings (SD=2.6) per single session

Fig. 1. Sleep latency and duration of NREM sleep Stages 1 and 2 as a function of serialawakenings.

16 V. Noreika et al. / International Journal of Psychophysiology 74 (2009) 14–18

(total recording time: M=193 min, SD=57.4). The mean intervalbetween two adjacent awakenings was 23.5 min (SD=14.7).Awakenings took place from 3 to 4 min after entering one of thethree sleep stages (M=208 s, SD=122): Stage 2 (M=241 s,SD=128), transition (M=205 s, SD=133), or Stage 3 (M=182 s,SD=96). There was a strong negative correlation between sequenceof serial awakenings and sleep latency (rho=− .98, n=8, pb0.001)and between awakenings and duration of Stage 1 sleep (rho=− .74,n=8, pb0.05), with higher numbers of serial awakenings associatedwith a shorter duration of sleep latency and Stage 1 sleep (see Fig. 1).No association was detected between serial awakenings and durationof Stage 2 sleep (rho=− .21, n=8, p=0.61).

3.2. Inter-rater agreement

The entire set of sleep EEG was scored by a single judge. To verifythe precision of scoring, 14 periods of pre-awakening EEGwere scoredby two judges with an initial agreement on 78.6% of the scored sleepstages (κ=.83). Altogether, 27 awakenings were excluded from themain analysis due to EEG artifacts, technical problems or experimentalerrors, as well as 8 other awakenings that took place during sleeponset REM sleep. The remaining 137 NREM sleep reports were contentanalyzed. Inter-rater agreement on the scoring of dream reports was89.8% (κ=.85) for the recall categories and 78% (κ=.88) for thecategories of perceptual complexity.

3.3. EEG spectral power

Three separate 2×4×5 ANOVA tests for each Stage of sleep(Stage 2, transition, Stage 3) with awakenings (early ENSA, late ENSA),electrodes (C3, C4, O1, O2) and bands (delta, theta, alpha, beta1,beta2) as factors did not reveal significant changes in spectral power

Table 2Dream recall categories in different time windows of awakenings.

Reports N — % NREM sleep stages

Stage 2 Transiti

Dreamless 42–30.66 10–23.81 19–37.2White dream 54–39.42 22–52.38 18–35.2Dream 41–29.93 10–23.81 14–27.4Total 137 42 51

for the main effect of awakenings (Stage 2: F(1,4)=1.51, p=.29;transition: F(1,4)=1.93, p=.24; Stage 3: F(1,4)=3.38, p=.14), orfor the interactions between the awakenings and other factors.Conversely, significant main effects were observed for the bandsduring all sleep stages (Stage 2: F(1.04,4.16)=170.22, pb .001,η2=.98; transition: F(1.02,4.08)=107.22, pb .001, η2=.96; Stage 3:F(1.09,4.35)=287.03, pb .001, η2=.99) and for the electrodes duringtransition and Stage 3 (Stage 2: F(1.04,4.15)=4, p=.11; transition:F(1.23,4.93)=9.63, pb .05, η2=0.71; Stage 3: F(1.56,6.21)=7.16,pb .05, η2=.64). Pairwise comparisons of the electrodes revealedseveral significant differences (pb .05) in the spectral power betweenthe electrodes: C3bC4, O1bO2 during Stage 2 sleep; C3NO1, C3NO2,C4NO1, C4NO2 during transition of sleep stages, and C3bC4, C3NO2,C4NO1, C4NO2 during Stage 3 sleep. Interaction between the bandsand the electrodes was significant during Stage 2 and transition, andapproached significance during Stage 3 sleep (Stage 2: F(1.5,6.0)=5.81, pb .05, η2=0.59; transition: F(1.95,7.8)=8.23, pb .05, η2=.67;Stage 3: F(1.41,5.65)=5.69, p=.051). Interaction graphs showed thatthe central electrodes had higher spectral power than the occipitalelectrodes in the delta and beta1 frequency bands, but lower power inthe theta and alpha bands during Stage 2 and transitional sleep. Stage 3sleep implied a somewhat different interaction, with the centralelectrodes having higher spectral power in the delta, alpha and beta1frequency bands, and the occipital electrodes having higher thetapower.

3.4. Dream recall

There were no significant differences in the percentage of dreamlesssleep (t(4)=.81, p=.46), white dreams (t(4)=.25, p=.81) anddreaming (t(4)=1.95, p=.12) between the early and the late ENSAAwakenings. These comparisons were additionally controlled for theduration of sleep and the sleep stage of awakening. As expected, therewere no significant differences in the percentage of Stage 2, transitional,and Stage 3 sleep between the awakenings. Similarly, duration of sleepwas not significantly different between the early ENSA (M=8:32 min,SD=2:06 min) and the late ENSA (M=7:37 min, SD=1:27 min)awakenings. Paired samples t tests were also used to explore thedifferences in proportions of recall categories (dreamless, white dream,dream) during nights (first two nights, last two nights) (see Table 2).There were no significant changes in the percentage of dreamless sleepandwhite dreaming from the first two to the last two nights (t(4)=.84,p=.45; t(4)=1.27, p=.27), while an increase in the percentage ofdream reports was very close to being marginally significant (t(4)=2.68, p=.055), suggesting that proportion of dreaming may increaseduring several subsequent nights of serial awakenings.

The percentage of reports of dreamless sleep and dreams wereexactly the same in Stage 2 awakenings (dreamless: 23.8%, dreams:23.8%), and approximately of the same magnitude in transition(dreamless: 37.3%, dreams: 27.5%) and Stage 3 (dreamless: 29.6%,dreams: 38.6%) NREM sleep (see Table 2). One-way ANOVA withNREM stages of sleep (Stage 2, transition, Stage 3) as factors didnot reveal significant differences in the percentage of dreamless sleep(F(1.89,7.57)=1.85, p=.22), whereas the percentage of white dream

Experimental nights

onal Stage 3 1–2 nights 3–4 nights

5 13–29.55 26–34.67 16–25.819 14–31.82 33–44.00 21–33.875 17–38.64 16–21.33 25–40.32

44 75 62

17V. Noreika et al. / International Journal of Psychophysiology 74 (2009) 14–18

reports significantly differed across the sleep stages (F(1.88,7.53)=7.15, pb .05, ηp2=.64). Pairwise comparisons showed a significant(pb .05) decrease of white dreaming from Stage 2 to transition, andfrom Stage 2 to Stage 3. The opposite increase in dreaming wasobserved from Stage 2 to transition to Stage 3, yet the differences didnot reach significance (F(1.88,7.51)=3.83, p=.073).

3.5. Dream complexity

Forty-one NREM dreams (N=41) were scored according to thefollowing complexity categories of Orlinsky's Modified Scale: category1=5 dreams, category 2=2 dreams, category 3=18 dreams,category 4=14 dreams, and category 5=2 dreams. Friedman'sANOVA test showed that complexity of dreaming, i.e. percentage ofsimple (categories 1–3) vs. complex (categories 4–7) dreams, did notdiffer between the sleep stages (χ2(2)=2.21, p=.39). Similarly,complexity of dreaming did not vary between the awakenings (earlyENSA, late ENSA) (t(4)=.27, p=.80).

4. Discussion

Results indicate that ENSA appeared to be an efficient paradigm forobtaining large samples of dream reports: 164 awakenings wereperformed in just 20 sessions with approximate 3 h duration, which isa significant increase compared to a typical four or five awakeningsper single night of 8 h. All five participants managed with theprocedure relatively well, and none of them dropped out of the studybecause of discomfort or difficulties in falling asleep. Contrary,correlation analysis showed that sleep latency shortens during thecourse of ENSA session. On average, ENSA allowed to acquire at least 8awakenings per single session. In some sessions, participants wereserially awakened up to 12 times, after which the session wasterminated by the experimenters, as we wanted to restrict serialawakenings to early-night sleep, without extending them to middle-or late-night sleep. Presumably, participants with sleeping difficultieswould not tolerate ENSA procedure very well. In fact, during pre-piloting phase of this study, one or two participants could not fallasleep easily after being awakened for several times. Perhaps astandard sleep quality questionnaire, such as the Pittsburgh SleepQuality Index (Buysse et al., 1989), should be used to select goodquality sleepers for efficient use of the ENSA paradigm in futurestudies.

EEG spectral power measurements remained stable across thewhole session. It has been reported previously that the rise latency fordelta (0.78–3.9 Hz) power is longer, and the latencies for alpha (8.2–11.7 Hz) and low sigma (12.1–13.7 Hz) power are shorter in the earlyNREM sleep episodes (Tagaya et al., 2000). Yang and Wu (2007)reported that several NREM sleep-specific ERP components, e.g. N350(250–475ms) and P900 (600–1000 ms), decreased during the secondhalf of the night, which was related to the partial satiation of thehomeostatic NREM sleep drive. It seems that ENSA during approxi-mately 3 h of total recording time is too short to produce the “time ofnight” effect for the sleep EEG. Alternatively, spectral power couldhave normally changed during 3 h of uninterrupted sleep, but ENSAdid not let this change to take place and kept EEG power at the samelevel. In any case, ENSA provided stable EEG power and dream recallfrequency across series of awakenings, which shows that the sug-gested paradigm can be used for studies on the early-night NREMdreams without losing data homogeneity.

The observed tendency for the increase in dream recall during thelater experimental nights indicates a practice effect. Arguably, traineddream reporters should participate in future studies, especially if thesample of participants is small and the reliability of dream recall isindispensable, e.g., during PET neuroimaging studies of sleep (Maquetet al., 1996). As Schredl and Fulda (2005) reported, a period of twoweeks seems to be long enough to obtain stable dream recall

frequencies at home; in the sleep laboratory, a period of more thanfive nightsmight also be needed for gettingmaximally reliable rates ofrecall.

Most of the NREM dreams were scored in the Orlinsky categories1–4, which are distinguished by the lack of temporal dynamics, andonly two dreams received category 5 scoring that involves thepresence of at least one perceptual change. Such simple early-nightNREM dreams seem to contrast with more complex REM dreams,which have been reported to fall to the highest categories in theOrlinsky's scale (Kamiya, 1961), and with the late-night NREMdreams, which typically resemble REM sleep dreaming (Cicogna etal., 1998). In the future, it would be useful to carry out direct within-participant comparisons between NREM dreams elicited by ENSA andlate-night NREM as well as REM sleep dreams. There may be several“levels” or “forms” of dreams involving distinct neurophysiologicalmechanisms, and, presumably, different experimental approachesthan ENSA might be required to investigate the psychophysiology ofmore complex dreaming.

Dissociation between dreamless sleep and dreaming during NREMsleep constitutes a theoretically interesting contrast between thepresence and the absence of subjective experience or “phenomenalconsciousness” during sleep (Revonsuo, 2006). Of course, assumingthat dreaming occurs throughout the night, but its recall is not alwayssuccessful, would be an alternative interpretation of the absence ofdream reports. In order to determine psychophysiological correlatesthat are specific to dreaming independently of the sleep stage,dreaming should be contrasted with dreamless sleep withoutvariation in sleep stage (Esposito et al., 2004). Our results show thatreports of both dreaming and dreamless sleep can be obtained withENSA from various stages of NREM sleep. Even though ENSA fromStage 2 resulted in the most balanced proportion of dream anddreamless reports, this stage of sleep also produced a relatively highproportion of white dreaming, a type of report that is not useful instudies comparing dreaming and dreamless sleep. The transitionalperiod and Stage 3 sleep provided almost equal proportions of reportsof dreamless sleep and dreaming, while the rate of white dreamingstayed low. As the scoring of the transitional period is less established,ENSA from Stage 3 can be regarded as the optimal time for awakeningsfor psychophysiological studies aiming at roughly equal proportions ofdreaming and dreamless sleep without variation in sleep stage.

The suggested ENSA paradigm has some limitations, which need tobe addressed carefully. Several participants exhibited symptoms ofsleep inertia following the last few awakenings of a session andtended to fall asleep while reporting. It is known that sleep inertia isespecially strong after awakenings from the suggested Stage 3 sleep(Tassi and Muzet, 2000). Consequently, the presence of confoundingvariables may be implicated in these late reports, such as blendeddream reports, especiallywhen two awakenings are very close in time.Thus, we propose that EEG should be constantly monitored afterawakenings and that drowsy participants should be additionallyaroused if there are indications of wakefulness turning into Stage 1sleep. Dream content analysis could also include a rating question onwhether the same, similar or different content is reported in severaladjacent awakenings.

The average dream recall frequency was relatively low in thepresent study (30%) when compared to NREM sleep studies that usedconventional awakenings. Nielsen (2000) reviewed 25 studiespublished since 1962 and reported the mean percentage of NREMdream recall to be 50.9% (SD=15.5, range: 23–75). However, in mostof these studies participants spent the entire night in the laboratory.Nevertheless, several studies have reported recall frequency close toour findings. It remains to be investigated in future, whether ENSAreduces recall rate; alternatively, our participants were not highdream recallers. Ideally, within-participant comparisons should becarried out between ENSA nights and nights with conventional NREMsleep awakenings.

18 V. Noreika et al. / International Journal of Psychophysiology 74 (2009) 14–18

Our participants were serially awakened during the first hoursof several non-consecutive nights. Further studies are needed toanswer whether ENSA would also be successful when applied duringconsecutive nights. ENSA paradigm could be also extended to middle-night serial awakenings (MiNSA) and to late-night serial awakenings(LaNSA). Presumably, such experiments would be able to observe,with a higher precision than previous studies, how the recall rate andEEG activity change throughout the course of night, and which sleepprocesses transform the early-night static NREM dreaming to complexREM-like dreams of the late-night hours of NREM sleep.

Acknowledgements

This study was supported by the Academy of Finland (projects Nr.8110957 and 8124623), and grants from the Signe and Ane GyllenbergFoundation and the Turku University Foundation (V. Noreika).Authors wish to thank L. Móró, J. Windt, D. Orlinsky, M. Oravecz, H.Iivonen andM. Koivisto for their help at different stages of this project,and anonymous reviewers for their helpful comments.

References

Bértolo, H., Paiva, T., Pessoa, L., Mestre, T., Marques, R., Santos, R., 2003. Visual dreamcontent, graphical representation and EEG alpha activity in congenitally blindsubjects. Cogn. Brain Res. 15, 277–284.

Buysse, D.J., Reynolds III, C.F., Monk, T.H., Berman, S.R., Kupfer, D.J., 1989. The PittsburghSleep Quality Index: a new instrument for psychiatric practice and research.Psychiat. Res. 28, 193–213.

Cicogna, P.C., Natale, V., Occhionero, M., Bosinelli, M., 1998. A comparison of mentalactivity during sleep onset and morning awakening. Sleep 21, 462–470.

Esposito, M.J., Nielsen, T.A., Paquette, T., 2004. Reduced Alpha power associated withthe recall of mentation from Stage 2 and Stage REM sleep. Psychophysiology 41,288–297.

Hori, T., Hayashi, M., Morikawa, T., 1994. Topographical EEG changes and the hypnagogicexperience. In: Ogilivie, R.D., Harsh, J.R. (Eds.), Sleep Onset: Normal and AbnormalProcesses. American Psychological Association, Washington, DC, pp. 237–253.

Kamiya, J., 1961. Behavioral, subjective, and physiological aspects of drowsiness andsleep. In: Fiske, D.W., Maddi, S.R. (Eds.), Functions of Varied Experience. DorseyPress, Homewood, IL, pp. 145–174.

Maquet, P., Péters, J.-M., Aerts, J., Delfiore, G., Degueldre, C., Luxen, A., Franck, G., 1996.Functional neuroanatomy of human rapid-eye-movement sleep and dreaming.Nature 383, 163–166.

Nielsen, T.A., 2000. A review of mentation in REM and NREM sleep: “Covert” REM sleepas a possible reconciliation of two opposing models. Behav. Brain Sci. 23, 851–866.

Orlinsky, D.E., 1962. Psychodynamic and cognitive correlates of dream recall. Un-published doctoral dissertation, The University of Chicago, Chicago.

Rechtschaffen, A., Kales, A., 1968. A Manual of Standardized Terminology, Techniquesand Scoring System for Sleep Stages of Human Subjects. Brain Information Service,Los Angeles.

Revonsuo, A., 2006. Inner Presence: Consciousness as a Biological Phenomenon. TheMIT Press, Cambridge, MA.

Schredl, M., Fulda, S., 2005. Reliability and stability of dream recall frequency. Dreaming15, 240–244.

Tagaya, H., Trachsel, L., Murck, H., Antonijevic, I., Steiger, A., Holsboer, F., Friess, E., 2000.Temporal EEG dynamics of non-REM sleep episodes in humans. Brain Res. 861,233–240.

Tassi, P., Muzet, A., 2000. Sleep inertia. Sleep Med. Rev. 4, 341–353.Wittmann, L., Palmy, C., Schredl, M., 2004. NREM sleep dream recall, dream report

length and cortical activation. Sleep Hypn. 6, 43–47.Yang, C.-M., Wu, C.-S., 2007. The effects of sleep stages and time of night on NREM sleep

ERPs. Int. J. Psychophysiol. 63, 87–97.