CC BY-NC-ND
0Neurobiology of Learning and Memory xxx (2015) xxx-xxx
Contents lists available at ScienceDirect
Neurobiology of Learning and Memory
journal homepage: www.elsevier.com/locate/ynlme
The effect of selective REM-sleep deprivation on the consolidation and affective evaluation of emotional memories
Christian D. Wiesnera'*, Julika Pulstb, Fanny Krause b, Marike Elsnerc, Lioba Bavinga, Anya Pedersenb, Alexander Prehn-Kristensen a, Robert Göderc
a Dept. of Child and Adolescent Psychiatry and Psychotherapy, Christian-Albrechts University School of Medicine, Kiel, Germany b Institute of Psychology, Christian-Albrechts University, Kiel, Germany
c Dept. of Psychiatry and Psychotherapy, Christian-Albrechts University School of Medicine, Kiel, Germany
ARTICLE INFO ABSTRACT
Emotion boosts the consolidation of events in the declarative memory system. Rapid eye movement (REM) sleep is believed to foster the memory consolidation of emotional events. On the other hand, REM sleep is assumed to reduce the emotional tone of the memory. Here, we investigated the effect of selective REM-sleep deprivation, SWS deprivation, or wake on the affective evaluation and consolidation of emotional and neutral pictures. Prior to an 9-h retention interval, sixty-two healthy participants (23.5 ± 2.5 years, 32 female, 30 male) learned and rated their affect to 80 neutral and 80 emotionally negative pictures. Despite rigorous deprivation of REM sleep or SWS, the residual sleep fostered the consolidation of neutral and negative pictures. Furthermore, emotional arousal helped to memorize the pictures. The better consolidation of negative pictures compared to neutral ones was most pronounced in the SWS-deprived group where a normal amount of REM sleep was present. This emotional memory bias correlated with REM sleep only in the SWS-deprived group. Furthermore, emotional arousal to the pictures decreased over time, but neither sleep nor wake had any differential effect. Neither the comparison of the affective ratings (arousal, valence) during encoding and recognition, nor the affective ratings of the recognized targets and rejected distractors supported the hypothesis that REM sleep dampens the emotional reaction to remembered stimuli. The data suggest that REM sleep fosters the consolidation of emotional memories but has no effect on the affective evaluation of the remembered contents. © 2015 The Authors. Published by Elsevier Inc. This is an open access article under the CCBY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4XI/).
Article history: Received 19 August 2014 Revised 4 February 2015 Accepted 12 February 2015 Available online xxxx
Keywords:
Rapid eye movement sleep Emotional memory Selective sleep deprivation
1. Introduction
Ever since the first description of REM sleep (Aserinsky & Kleitman, 1953), scientists have wondered about the function of this sleep stage and whether it is necessary at all (Dement, 1960). In an early experiment, Dement (1960) deprived participants of 65-75% of normal REM sleep for 5 consecutive nights. He characterized the occurring psychological disturbances as not catastrophic but encompassing anxiety, even panic, and irritability. Hence, emotional stability seemed to depend on REM sleep. A lot of studies conducted in the last 15 years document the role of REM sleep not only in the affective evaluation (Goldstein & Walker, 2014) but also the consolidation (Rasch & Born, 2013) of emotional memories. One theory about the role of REM sleep that has gotten a lot of attention in recent years is that of Walker and colleagues (Goldstein & Walker, 2014; Walker & van der Helm, 2009). The
* Corresponding author. E-mail address: ChristianDirk.Wiesner@uksh.de (C.D. Wiesner).
authors propose the so-called ''sleep to forget and sleep to remember" hypothesis (SFSR): REM sleep as compared to wake or other sleep stages is supposed to foster the memory consolidation of emotional events as compared to neutral events (''sleep to remember''). On the other hand, REM sleep is assumed to reduce the emotional tone of the memory (''sleep to forget''). That is, REM sleep is believed to attenuate the potential of a memory to elicit emotional reactions during recall. Note that the hypothesis only applies to declarative and especially episodic memory.
There are some studies supporting the ''sleep to remember''-part of the hypothesis (Groch, Wilhelm, Diekelmann, & Born, 2013; Groch, Zinke, Wilhelm, & Born, 2014; Nishida, Pearsall, Buckner, & Walker, 2009; Wagner, Gais, & Born, 2001). According to these studies, REM sleep indeed fosters the sleep-dependent consolidation of emotional memories. In an early, well controlled study, Wagner et al. (2001) used the split-night design to investigate the influence of REM sleep on emotional memory formation. In the split-night design, SWS-rich sleep in the first half of the night is compared with REM-sleep-rich sleep in the second half
http://dx.doi.org/10.1016/j.nlm.2015.02.008 1074-7427/© 2015 The Authors. Published by Elsevier Inc.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
of the night and corresponding wake intervals to control for circa-dian effects. The results showed that only late sleep rich in REM selectively fostered the consolidation of emotional in comparison to neutral texts. Nishida et al. (2009) compared the effects of a short nap and an interval spent awake on emotional memory consolidation. The nap design allows a strict control of circadian effects but has to rely on mere correlations of sleep stages during the nap with behavioral measures. The authors report that a nap fostered the consolidation of emotional but not neutral memories and that the amount of REM sleep correlated with the extent of emotional memory facilitation. Groch et al. (2013) conducted an experiment using the split-night paradigm but without wake control conditions (compare Wagner et al., 2001). They investigated the influence of early and late-night sleep on the consolidation of memories for negative and neutral pictures. The results supported the ''sleep to remember"-hypothesis: REM-sleep rich, late nocturnal sleep compared to SWS-rich, early sleep enhanced the consolidation of emotional pictures as compared to neutral pictures. Recently, Groch et al. (2014) published two studies employing the split-night paradigm. The studies elucidate on the critical question of which part of a declarative memory representation profits from REM sleep versus SWS. Concerning the memory for items such as pictures, only late-night REM-sleep rich sleep fostered the emotional memory enhancement. SWS-rich, early sleep helped to consolidate the memory for contexts in which the items were encountered during encoding. In summary, the most evidence in favor of the ''sleep to remember''-hypothesis (Goldstein & Walker, 2014; Walker & van der Helm, 2009) comes from experiments using the split-night design.
Although the SFSR-hypothesis has inspired a lot of studies, there is only one study supporting the notion of the ''sleep-to-for-get''-part, that REM sleep helps to tune down the emotional tone of negative memories. Van der Helm and colleagues reported a correlation between a REM sleep EEG parameter and the attenuation of the emotional reaction to repeatedly viewed pictures (van der Helm et al., 2011). The authors compared emotional reactivity after a night of sleep relative to a day of wake (i.e. a day-night design) and specifically looked at the role of REM sleep physiology. They focus on prefrontal gamma EEG activity as a marker of central adrenergic activity and report that this marker predicted emotional reactivity as measured by amygdala activity and behavioral reactivity. In particular, they found that anterior prefrontal gamma-activity during REM sleep was correlated with the decrease of emotional reactivity to pictures. Unfortunately the authors do not report correlations of the amount of REM sleep or SWS with emotional reactivity.
The present study aims to investigate the ''sleep-to-remember sleep-to-forget''-hypothesis. In a previous study using REM sleep deprivation, we were not able to replicate the ''sleep to remember'' effect (Morgenthaler et al., 2014). However, we did not measure the affective evaluation of the pictures after sleep and did not include a NREM-sleep deprivation condition as control for the effects of awakenings (Morgenthaler et al., 2014). Therefore, in the present study we compare the effects of selective, REM-sleep deprivation (REMD) and selective SWS deprivation (SWSD) as compared to wake on the consolidation and change of the emotional evaluation of neutral and negative (''emotional'') pictures. Selective sleep deprivation allows to experimentally manipulate sleep stages right up to almost complete elimination or REM sleep or SWS. This offers the opportunity to further elucidate the causal role of these sleep stages for the consolidation and affective evaluation of emotional memories.
The ''sleep-to-remember''-part of hypothesis states that REM sleep is necessary or at least advantageous to consolidate emotional memories. Therefore, we expected a REM-sleep-deprived group
to show worse memory performance, especially with the emotional pictures. Moreover, the (residual) time spent in REM sleep was expected to correlate with this emotional memory bias. The ''sleep-to-forget''-part of the hypothesis states that REM sleep decreases the potential of repeatedly encountered negative pictures to elicit an emotional response. Therefore, we expected that the arousal ratings of negative pictures would drop and that the valence rating would only shift to neutral during the night in the SWS-deprived group where REM sleep was present.
2. Materials and methods
2.1. Participants
Sixty-two healthy students (23.5 ± 2.5 years, 32 female, 30 male) were recruited at the University of Kiel campus and were paid 70 EUR for participation. All participants gave written informed consent before taking part. The study protocol was approved by the local ethics committee of the school of medicine of the Christian-Albrechts University of Kiel. Exclusion criteria were a history of neurological or psychiatric disorders, sleep disorders, medications (except contraceptives), nicotine dependence, left-handedness, defective vision, or a body-mass-index above 30. A telephone interview and questionnaires were used to assess these criteria. Psychiatric symptoms were assessed by a German version of the symptom checklist (SCL-90-R, Franke & Derogatis, 2002). Participants were screened for sleep disturbances by the Pittsburgh Sleep Quality Index (PSQI; Buysse, Reynolds, Monk, Berman, & Kupfer, 1989). The Fagerstrom test for nicotine dependence (Bleich, Havemann-Reinecke, & Kornhuber, 2002) was used to assure that no participant suffered from nicotine dependence. The Edinburgh-Handedness-Inventory (Oldfield, 1971) was used to ensure that all participants were consistent right-handers. All values of the respective questionnaires were in the normal range. All participants had normal or corrected-to-normal vision and reported no color blindness. To ensure that the participants of the sleep groups were able to sleep, we asked them to maintain a regular sleep schedule on the days before the experimental night and to fill out sleep protocols (Hoffmann, Müller, Hajak, & Cassel, 1997). Furthermore, they were asked to abstain from caffeine, nicotine, and alcohol consumption on the day before the experiment. One participant included in the original sample (SWSD group) had to be excluded due to questionable understanding of the instructions. The accuracy score for high-confidence recognition of neutral pictures was more than two standard deviations below the mean (.067 as compared to .492 ± .177). The whole sample of 62 participants was divided into a REM-sleep-deprivation group (REMD), a SWS-deprivation group (SWSD), and a Wake group (Wake). Note, that all other sleep stages, including SWS,
Table 1
Sleep parameters across groups.
REM-deprived mean (SEM) SWS-deprived mean (SEM) t df P
N1 (min) 67.89 (4.59) 72.35 (4.16) -0.718 39 .477
N2 (min) 139.12 (10.91) 166.51 (5.99) -2.200 30.9+ .035
N3 (min) 89.38 (5.62) 7.52 (0.71) 14.46 20.6+ <.001
REMS (min) 2.05 (0.55) 56.67 (5.15) -10.54 19.4+ <.001
TST (min) 293.67 (12.61) 303.09 (5.93) -0.676 28.4+ .505
TST/TIB (%) 68.05 (3.17) 68.00 (1.54) 0.01 28.8+ .989
Awakenings 12.33 (5.74) 22.20 (5.30) -5.712 39 <.001
Sleep stages according to the AASM classification: N1, N2, N3, non-REM sleep stages; REMS, REM sleep; TST, Total Sleep Time; TIB, time in bed; sleep efficiency = TST/TIB in %. + df adjusted due to significant Levene test for equal variances.
were present in the REMD group and that all other sleep stages, including REM, sleep were present in the SWSD group (see Table 1 and the results section). The REMD group consisted of 11 female and 10 male participants; the SWSD group of 10 female and 10 male participants; and the Wake group of 11 female and 10 male participants. We restricted the age range to 18-30 years to control for gross developmental changes. The mean age ± SEM was 23.5 ± 0.55 years in the REMD group, 23.5 ± 0.57 in the SWSD group, and 23.4 ± 0.55 in the Wake group and did not differ significantly between groups (F2;59 = 0.02, p = .981).
2.2. Memory task and affective ratings
The memory task consisted of three parts: (1) picture encoding and affective rating, (2) an immediate recognition control test, and (3) a delayed (9 h) recognition test and affective rating. The pictures were taken from the International Affective Picture System (80% of the pictures; Lang, Bradley, & Cuthbert, 2005) as well as an in-house picture set we had previously used in several studies (20%; Morgenthaler et al., 2014; Prehn-Kristensen et al., 2011).
During the encoding, the participants viewed 80 neutral and 80 negative (''emotional'') target pictures in a fixed, pseudo-random order on an LCD computer screen. They were instructed that memory for the pictures would be tested after the retention interval. Each presentation started with a fixation slide (1000 ms), followed by the picture (1500 ms), and valence and arousal rating scales (no time limit). The affective ratings were given on the Self-Assessment Manikin scales for valence and arousal (Bradley & Lang, 1994). The valence scale ranged from 0 (negative) to 4 (neutral) to 8 (positive) and the arousal scale from 0 (no arousal) to 8 (high arousal). During the encoding, the negative pictures were rated as more negative than the neutral pictures (negative means ±SEM: 2.02 ±0.06 vs. neutral: 4.69 ±0.04, t = -31.94, df = 61, p < .001) and the negative as more arousing than the neutral ones (negative: 3.98 ±0.16 vs. neutral: 1.30 ±0.11, t =19.96, df = 61, p < .001). The valence ranged from 0.19 to 4.27 in the negative stimuli and 3.51 to 6.48 in the neutral stimuli. The arousal ranged from 2.02 to 7.03 in the negative and 0.27 to 2.68 in the neutral stimuli.
During the immediate recognition test, the participants viewed 20 negative and 20 neutral targets mixed with 20 negative and 20 neutral new distractors matched for arousal and valence. The pictures were also matched to the pictures of the delayed recognition test. To limit the work load to an acceptable amount prior to the retention interval, to avoid emotional blunting, and to keep up the motivation of the participants, we used fewer items in the immediate test than in the delayed test. Each presentation started with a fixation slide (1000 ms), followed by the picture (1500 ms). Then, participants were asked, if the stimulus was an old one, which had been seen during encoding, or a new, as yet unseen one. They indicated their answer by clicking the right or left mouse button (time unlimited). Afterward, they were to indicate how confident they were that they had correctly recognized or rejected the stimulus on a four-point scale ranging from 1 ''very unsure'' to 4 ''very sure'' (time unlimited). According to signal detection theory, we computed the un-normalized hit rate (correctly identified as old), miss rate (wrongly identified as new), correct rejection rate (correctly identified as new), and false-alarm rate (wrongly identified as old) (Green & Swets, 1966). The accuracy, i.e. the difference between hit rate and false-alarm rate, was used as a bias-corrected and robust measure of memory performance. All measures were calculated for negative and neutral stimuli separately and used to base-line correct delayed memory performance.
During the delayed recognition test, the participants viewed the remaining 60 negative and 60 neutral targets mixed with 60 negative and 60 neutral new distractors. Procedure and measures were
the same as in the recognition control test described above. In addition, the participants gave an affective rating for all stimuli just like during encoding. We calculated the mean valence and arousal ratings for negative and neutral pictures as well as old and new pictures separately. For comparability with the study by Groch and colleagues, we also calculated the measures for delayed memory performance using only high-confidence items, i.e. ''very sure hits'' and ''very sure false alarms'' (Groch et al., 2013). On average, 38.2 ± 1.3 of 60 negative and 30.3 ± 1.3 of 60 neutral pictures were correctly recognized with very high confidence (mean ± SEM). Only 1.4 ± 0.2 of 60 negative and 0.9 ± 0.2 neutral pictures were wrongly recognized with high confidence. The purpose of this analysis is to focus on items reliably consolidated in memory.
2.3. Sleep recording and deprivation
The participants of the sleep groups spent two nights in the sleep laboratory. The first night's purpose was to exclude severe sleep disorders and to help participants adapt to the conditions in the sleep laboratory. Sleep during the experimental night was recorded by standard procedures using a digital electroencephalogram (EEG), electromyogram (EMG), electrooculogram (EOG), and electrocardiogram (EKG). To amplify and record the data, a SOMNOscreen PSG plus TM (SOMNOmedics, Randersacker, Germany) was used. The EEG montage according to the 10-20 system included single multi-use Ag/AgCl-electrodes attached to the positions C4 referenced to A1, O2 referenced to A1, and F4 referenced to A1. F3 referenced to A2, C3 referenced to A2, and O1 referenced to A2 were used as backup positions. Following the recommendations of the American Academy of Sleep Medicine (AASM; Berry et al., 2013), the EEG data were filtered with a 0.135 Hz band pass filter, the EOG data with a 0.1-10 Hz filter, and the EMG data with a high pass filter of p 10 Hz. All sleep data were analyzed on the next day according to the specifications provided in the revised manual of the AASM by a certified rater unaware of the hypotheses.
During the experimental nights selective deprivation of REM sleep or SWS was carried out. The sleep was monitored and classified according to standard criteria by two trained psychologists. REM sleep was indicated by rapid eye movements, low muscle tone, and rapid low-voltage EEG. Slow wave sleep (N3) was indicated by slow waves during at least 20% of an epoch. Slow waves are defined as waves of a frequency of 0.5-2 Hz and peak-to-peak amplitude greater than 75 iV. As soon as the first epoch of REM sleep in the REMD group or SWS in the SWS-deprivation group was identified, participants were awakened. The experimenters first used an intercom to address the participant and elicit a verbal response. If that was not sufficient to wake up the participant completely, an experimenter entered the room and addressed him or her directly. If that was not sufficient either, the participant was asked to sit up for a moment and solve a simple mathematical task (''subtract a constant amount from 100''). The criterium for wake was a high-frequency, low-voltage EEG for at least 1 min.
To control for tiredness and mood changes following the deprivation procedure, we used the German multidimensional state questionnaire (Mehrdimensionaler Befindlichkeitsfragebogen, MDBF; Steyer, Schwenkmezger, Notz, & Eid, 1997). The MDBF consists of 24 adjectives describing mental or affective states which can be rated on a five-point scale and belong to three sub-scales with 8 items each. The sub-scales range from 8 to 40 with a scale midpoint of 24. High values on the sub-scale ''good versus bad mood'' indicate a positive valence. High values on the sub-scale ''calm versus nervous states'' indicate a low arousal state. High values on the sub-scale ''wakeful versus tired states'' indicate wakefulness.
2.4. Procedure
The two sleep groups spent an adaptation night in the sleep laboratory the day before the experiment to familiarize them with the environment and the electrodes and to exclude sleep disorders. In the evening of the experimental nights, participants came to the sleep laboratory at 8 pm. They were familiarized with the procedure, and the electrodes were attached and tested. Around 9 pm they started with the memory encoding, followed by the immediate recognition control test. At 10:30 pm the lights were switched off, and the participants were asked to go to sleep. During the night, selective deprivation of REM sleep or SWS was carried out. Participants were finally awakened at 6 am, the electrodes were removed, and breakfast was offered. At 6:45 am the participants filled out the MDBF to describe their mental and affective state and started with the delayed recognition test. They left the laboratory at around 8 am. The participants of the Wake group came to the sleep laboratory at 9 am and performed the encoding and the immediate recognition test. The delayed recognition took place at 18:45 pm of that day. In between they stayed awake and went along with their usual daily routine. The duration of the delay between encoding and delayed recognition was approximately 9 h in all groups.
2.5. Statistical analysis
Sleep parameters, tiredness, and affective state ratings were compared across groups using two-sided t-tests. If a Levene-Test indicated inhomogeneous variances we used the Cochran and Cox adjustment for the standard error and corrected the degrees of freedom according to Satterthwaite. To evaluate the effects of sleep deprivation on memory performance and affective ratings, we applied ANOVAs with TIME (immediate versus delayed recognition) and EMOTION (high arousing negative vs. low arousing neutral) as within-subject factors and GROUP (REMD, SWSD, Wake) as a between-subject factor. Significant ANOVA effects were followed by post hoc t-tests. In one case, we calculated exploratory t-tests which were not protected by a significant interaction in the ANOVA to compare our results with a study by Groch et al. (2014). To test whether we could replicate the proposed positive correlation of REM sleep and emotional memory consolidation, we calculated one-tailed tests of the Pearson correlations. To evaluate the effects of sleep deprivation on the affective evaluation of memorized pictures, we computed ANOVAs with TIME (immediate versus delayed recognition) and EMOTION (high arousing negative vs. low arousing neutral) as within-subject factors and GROUP (REMD, SWSD, Wake) as a between-subject factor. Furthermore, we computed an ANOVA with EMOTION (high arousing negative vs. low arousing neutral) and RECOGNITION (hits vs. correct rejections) as within-subject factors and GROUP (REMD, SWSD, Wake) as a between-subject factor. Again, significant ANOVA effects were followed by t-tests. The level of significance was set <.05 for all analyses. Data analysis was performed with SPSS for Windows, version 22.0 (SPSS Inc., Chicago, IL, USA).
3. Results
3.1. Sleep
The goal of the deprivation procedure was to reduce REM sleep or SWS sleep as much and as selectively as possible while maintaining an equal amount of sleep in both deprivation groups. As Table 1 shows, the total sleep time (TST) was not significantly different (p = .510) and amounted to roughly 5 h in both groups. However, REM sleep was significantly reduced to 2.05 min on
average in the REMD group as compared to 56.67 min in the SWSD group (p <.001). Likewise, the amount of SWS (N3) was reduced to 7.52 min in the SWS-deprived participants as compared to 89.38 min in the REM-deprived (p < .001). The number of awakenings necessary to deprive SWS was almost double the number necessary to deprive REM sleep (p <.001). Furthermore, if SWS was deprived, lighter sleep (N2) automatically became more prevalent. This was reflected in a moderately enhanced (p = .035) amount of N2 in the SWSD group (166.51 min) as compared to the REMD group (139.12 min). N1 sleep, however, was equal in both groups (p = .477). The sleep efficiency was 68% in both groups (p = .989). In summary, the manipulation mainly reduced the targeted sleep stages while keeping the total sleep time and sleep efficiency nearly constant.
The deprivation of REM sleep or SWS sleep as compared to wake had no effect on the subjective mood state at recognition (Table 2). Neither the MDBF scale ''good versus bad mood'' (p = .804) nor the scale ''calm versus nervous states'' (p = .827) showed any significant differences across the three groups. The mood in all groups was in the positive range of the valence scale and in the low range of the arousal scale. Only the MDBF scale ''wakeful versus tired states'' showed significant group differences (F2;59 = 3.62, p = .033). The Wake group reported to be more wakeful (mean ± SEM: 28.48 ±1.52) than the REMD group (23.19 ±1.52; p = .017) and the SWSD group (23.75 ± 1.56; p = .034). However, the REM-deprived and the SWSD group did not differ concerning wakefulness (p = .798).
3.2. Recognition memory
First, we compared the accuracy during the immediate and the delayed recognition tests using an ANOVA with the within-subject factors TIME (immediate vs. delayed) and EMOTION (neutral vs. negative) and the between-subject factor GROUP (REMD, SWSD, Wake). The descriptive statistics are reproduced in Table 3. The results of the ANOVA are illustrated in Fig. 1 in a condensed form as forgetting rates (delayed accuracy - immediate accuracy). The analysis revealed that a significant amount of pictures were forgotten during the retention interval (TIME: F1;59 = 268.28, p <.001) and that the retention of emotional pictures was better than for neutral pictures (EMOTION: F1;59 = 15.68, p < .001). There was no main effect of the GROUP (F2;59 = 1.76, p = .181). Emotional pictures were forgotten less often than neutral ones (t = -4.85, df = 61, p < .001) as revealed by a significant interaction of TIME and EMOTION (F1;59 = 23.74, p < .001). Also, there was a significant interaction of TIME and GROUP (F2;59 = 3.85, p = .027): Independent of the emotional content, the Wake group forgot significantly more pictures than the SWSD group (t = 2.73, df=39, p = .009) and, by trend, the REMD group (t =1.75, df = 40, p = .087). However, the interactions of EMOTION and GROUP (F2;59 = 0.39, p = .678) and TIME, EMOTION, and GROUP (F2;59= 0.93, p = .399) were not significant. To compare our results with the studies by Groch et al. (2013, 2014), we calculated exploratory t-tests to further explore whether emotional pictures were forgotten less than neutral ones in all groups. This emotional memory facilitation was only significant in the Wake group (t = -4.47, df = 20, p <.001) and in the SWSD group (t = -4.43, df = 19, p <.001) but not in the REMD group (t = -1.41, df = 20, p = .175). Furthermore, we had a closer look at the forgetting rates for negative pictures (Fig. 1): The REMD group forgot more emotional pictures over the retention interval than the SWSD group (t = -2.05, df = 39, p = .048). Also, the Wake group forgot more emotional pictures than the SWSD group (t = -3.37, df=39, p = .002). However, the forgetting rate for emotional pictures did not differ between the REMD and the Wake group (t = 1.17, df = 40, p = .250). Although the 3-way interaction was not
Table 2
Mood state and wakefulness measured by the MDBF.
REMD SWSD Wake ANOVA
Mean ± SEM Mean ± SEM Mean ± SEM df F P
Good mood (GB) 33.2 ±1.0 33.8 ±1.1 34.1 ±1.0 2; 59 0.22 .804
Calm state (CN) 32.0 ±1.1 33.0 ±1.1 32.3 ±1.1 2; 59 0.19 .827
Wakefulness (WT) 23.2 ±1.5* 23.8 ±1.6* 28.5 ±1.5 2; 59 3.62 .033
The MDBF sub-scales range from 8 to 40 with a scale midpoint of 24. High values on the sub-scale ''good versus bad mood'' (GB) indicate a positive valence. High values on the sub-scale ''calm versus nervous states'' (CN) indicate a low arousal state. High values on the sub-scale ''wakeful versus tired states'' (WT) indicate wakefulness. The asterisk indicates a significant (p < .05) difference between one of the sleep groups and the Wake group in a post hoc t-test.
Table 3
Memory performance.
Neutral Mean ± SEM
Emotional Mean ± SEM
Neutral Mean ± SEM
Emotional Mean ± SEM
Neutral Mean ± SEM
Emotional Mean ± SEM
All trials
Immediate
Delayed
Delayed-immediate
High-confidence trials Delayed
.902 ± .021 .721 ±.021 -.181 ±.031
.521 ± .033
.912 ±.018 .783 ±.019 -.129 ±.020
.623 ± .039
.855 ± .022 .672 ± .028 -.183 ± .024
.508 ± .039
.853 ± .025 .774 ± .020 -.078 ±.015
.631 ± .033
.907 ±.015 .673 ± .023 -.234 ±.019
.469 ± .040
.900 ±.011 .739 ± .024 -.161 ± .020
.586 ± .038
Memory performance of the REM-sleep deprived, the SWS-deprived group and the Wake group is given as recognition accuracy (hit-rate - false- alarm-rate) in means ± SEM. The standard signal detection analysis encompassed all trials. The immediate and the delayed recognition accuracy for neutral and emotional pictures, as well as the difference (forgetting), are given in the upper three rows. The accuracy values in the lower row are based only on responses to stimuli rated with the highest confidence during the delayed recognition.
Fig. 1. Memory performance. Accuracy of the delayed recognition test minus the accuracy of the immediate recognition test for neutral and emotional pictures. REMD, REM deprived group; SWSD, slow-wave sleep deprived group, WAKE, Wake group. The p-values correspond to t-tests reported in the text.
significant (p = .399) and the t-tests have to be interpreted carefully, the correlation analysis supports the proposed role of REM sleep. We calculated the difference of the delayed recognition accuracy for emotional and neutral pictures as a measure of the emotional memory bias. In the SWSD group, the amount of REM sleep correlated positively with the emotional memory bias (r = .417, n = 20, p = .034 one-tailed). The more REM sleep the participants had, the more emotional pictures were remembered compared to neutral pictures (see Fig. 2A). In the REM-sleep-deprived group, the amount of residual REM sleep only correlated by trend with the emotional memory bias (r = .317, n = 21, p = .081 one-tailed, see Fig. 2B). The respective correlations with SWS were not significant (all p > .218 one-tailed). Thus, REM-sleep seemed to foster the consolidation of emotional memories in the sleep groups.
To compare our results with the study by Groch and colleagues (Groch et al., 2013), we computed two additional ANOVAs with the factors EMOTION (neutral vs. negative) and GROUP (REMD, SWSD, Wake). In the first analysis, we refrained from using the immediate recognition test and simply compared the memory performance in the delayed recognition test. We found a significant main effect of EMOTION (F1;59 = 44.57, p <.001), meaning that the retention of emotional pictures was better than for neutral ones. However, there was no main effect of GROUP (F2;59 = 1.36, p = .265) nor an interaction of EMOTION and GROUP (F2;59 = 1.24, p = .296). In the second analysis, we only included responses to stimuli rated with the highest confidence. The recognition accuracy for high-confidence responses revealed a superior recognition of emotional as compared to neutral pictures (EMOTION: F1;59 = 65.89, p <.001), but this effect was independent of the sleep or Wake groups (EMOTION / GROUP: F2;59 = 0.18, p = .835), and there was no main effect of GROUP either (F2;59 = 0.51, p = .603).
3.3. Affective ratings
Firstly, we compared the affective ratings of target pictures during encoding and during the delayed recognition test using an ANOVA with the factors TIME (encoding vs. recognition), EMOTION (neutral vs. negative), and GROUP (REMD, SWSD, Wake) (see Table 4). Indeed, we found a significant decrease in the subjective arousal over TIME (F1;59 = 32.06, p < .001). Also, the emotional content of the pictures had an effect on the arousal ratings (EMOTION: F1;59 = 414.92, p <.001), meaning that negative pictures were rated as more arousing. There was no main effect of GROUP (F2;59 = 0.50, p = .609). The decrease in arousal over time did not depend on sleep or Wake (TIME / GROUP: F2;59 = 0.34, p = .710), and the groups did not differ in their overall evaluation of emotional versus neutral pictures with regards to arousal (EMOTION / GROUP: F2;59 = 1.39, p = .256). There was a trend toward an interaction of TIME and EMOTION (F1;59 = 3.12, p = .083), meaning that the arousal ratings for negative pictures
C.D. Wiesner et al. /Neurobiology of Learning and Memory xxx (2015) xxx-xxx
Fig. 2. (A) Correlation between REM sleep in minutes and emotional memory bias in the SWS-deprived group. (B) Correlation between REM sleep in minutes and emotional memory bias in the REM-sleep-deprived group. The difference of the delayed recognition accuracy for emotional and neutral pictures was calculated as a measure of the emotional memory bias.
Table 4
Affective ratings.
REMD SWSD Wake
Neutral Emotional Neutral Emotional Neutral Emotional
Mean ± SEM Mean ± SEM Mean ± SEM Mean ± SEM Mean ± SEM Mean ± SEM
Arousal
Targets Pre 1.27 ±0.15 4.22 ± 0.22 1.25 ±0.25 3.66 ± 0.30 1.28 ±0.20 4.25 ± 0.31
Targets Post 0.90 ±0.16 3.77 ± 0.23 0.97 ± 0.24 3.35 ± 0.34 1.10 ±0.21 3.79 ± 0.36
Post - Pre -0.38 ±0.10 -0.45 ±0.14 -0.28 ± 0.07 -0.31 ±0.12 -0.19 ±0.09 -0.46 ±0.17
Old (Hits) 0.95 ±0.17 3.81 ± 0.23 1.01 ±0.24 3.36 ± 0.35 1.14 ±0.22 3.82 ± 0.36
New (CR) 0.89 ±0.17 3.44 ± 0.22 1.01 ±0.26 3.13 ±0.31 1.10 ±0.21 3.56 ± 0.35
Old - New 0.06 ± 0.07 0.37 ± 0.09 0.00 ± 0.06 0.23 ± 0.09 0.04 ± 0.05 0.25 ± 0.09
Valence
Targets Pre 4.65 ± 0.08 2.00 ± 0.09 4.76 ± 0.07 1.99 ±0.11 4.54 ± 0.07 2.04 ±0.12
Targets Post 4.52 ± 0.07 2.18 ±0.08 4.58 ± 0.08 2.11 ±0.12 4.47 ± 0.08 2.32 ±0.11
Post - Pre -0.14 ±0.04 0.19 ±0.05 -0.18 ±0.05 0.12 ±0.05 -0.07 ± 0.08 0.28 ± 0.07
Old (Hits) 4.56 ± 0.09 2.16 ±0.08 4.56 ± 0.08 2.11 ±0.12 4.48 ± 0.07 2.32 ±0.11
New (CR) 4.53 ± 0.08 2.41 ± 0.08 4.62 ± 0.09 2.19 ±0.12 4.53 ± 0.09 2.36 ±0.12
Old - New 0.02 ± 0.06 -0.25 ± 0.05 -0.05 ± 0.04 -0.08 ± 0.04 -0.06 ± 0.05 -0.04 ± 0.04
Affective ratings of the REM-sleep-deprived and the SWS-deprived group are given as means and SEM. Arousal and valence ratings for neutral and negative pictures were obtained during encoding (targets pre) and delayed recognition (targets post) and for old pictures (hits) and new ones (correct rejections, CR) during delayed recognition. The arousal scale ranged from 0 (''no arousal'') to 8 (''very high arousal''). The valence scale ranged from 0 (''very negative valence'') to 8 (''very positive valence'').
decreased slightly more over time than for neutral ones (t = 1.78, df=61, p = .080). Most importantly, the 3-way interaction of TIME, EMOTION and GROUP, which would substantiate the claim that specifically REM sleep decreases emotional arousal, was not significant (F2;5g = 1.13, p = .330).
We see a similar picture with the valence ratings. The negative pictures were rated as more negative than the neutral ones (EMOTION: F1;59 = 944.85, p < .001). There were no significant main effects of TIME (F1;59 = 1.57, p = .216) or GROUP (F2;59 = 0.05, p = .947), but the valence ratings became less extreme over time, as was reflected by a significant interaction of TIME and EMOTION (Fi;59 = 54.65, p <.001). That is, the valence ratings of neutral pictures became less positive (t =3.59, df=61, p = .001) and the valence ratings of the negative pictures became less negative (t = -5.82, df = 61, p <.001). The interactions of TIME and GROUP (F2;59 = 1.91, p = .157), EMOTION and GROUP (F2;59 = 1.13, p = .330), as well as TIME, EMOTION and GROUP (F2;59 = 0.01, p = .910), were not significant.
Secondly, we compared the affective ratings of correctly recognized targets (hits) and correctly rejected distractors (CR) from the delayed recognition test (see Fig. 3A). This comparison is based on the supposition that only items already consolidated in memory and only after REM-sleep may be rated as less arousing. An ANOVA of the arousal ratings with the factors RECOGNITION (hits vs. correct rejections), EMOTION (neutral vs. negative), and GROUP (REMD, SWSD, Wake) revealed no significant main effect of GROUP (F2;59 = 0.34, p = .712). However, we found significant main effects of EMOTION (F1;59 = 3 5 5.20, p <.001) and RECOGNITION (Fi;59 = 23.47, p < .001), meaning that emotional pictures were rated as more arousing than neutral ones, and correctly recognized targets were rated as more arousing than correctly rejected dis-tracters. These main effects were qualified by an interaction of EMOTION and RECOGNITION (F^ = 19.12, p <.001). The difference of arousal ratings of hits and correct rejections was greater in the negative pictures than in the neutral ones (t = -4.43, df=61, p <.001). Note, however, that this interaction effect did
C.D. Wiesner et al. /Neurobiology of Learning and Memory xxx (2015) xxx-xxx
Fig. 3. Affective ratings. (A) Arousal ratings during recognition for hits and correct rejections (CR). The scale ranged from 0 ("no arousal'') to 8 (''very high arousal''). (B) Valence ratings during recognition for hits and correct rejections (CR). The scale ranged from 0 (''very negative valence'') to 8 (''very positive valence''). The lower p-values correspond to the t-tests comparing the ratings of emotional and neutral pictures. The p-values in the middle correspond to t-tests resolving the interaction of EMOTION and RECOGNITION. The p-values at the top correspond to the 3-way interaction of EMOTION, RECOGNITION, and GROUP.
not depend on sleep (EMOTION / RECOGNITION / GROUP: F2;59 = 0.28, p = .757). As illustrated in Fig. 3A and emphasized by additional t-tests, the arousal ratings of pictures held in memory (hits) were significantly higher than the arousal ratings of pictures encountered the first time (correct rejections) in all three groups (REMD: t = 4.10, df=20, p = .001; SWSD: t = 2.68, df= 19, p = .015; Wake: t = 2.76, df = 20, p = .012). Additionally, the difference of arousal ratings of hits and correct rejections in all three groups was greater for emotional than neutral pictures (REMD: t = -2.98, df = 20, p = .007; SWSD: t = -2.20, df=19, p = .040; Wake: t = -2.39, df =20, p = .027). Furthermore, the GROUP factor did not interact with RECOGNITION (F2;59 = 0.88, p = .419) or EMOTION (F2;59 = 1.08, p = .346). In summary, the results oppose the hypothesis that sleep or REM sleep in particular reduces the arousal of emotional memories.
The results for the valence ratings do not support the hypothesis either (see Fig. 3B). The ANOVA of the valence ratings with the factors RECOGNITION (hits vs. correct rejections), EMOTION (neutral vs. negative), and GROUP (REMD, SWSD, Wake) revealed significant main effects of RECOGNITION (F1;59 = 15.24, p < .001), EMOTION (F1;59 = 668.01, p <.001), and a significant interaction of RECOGNITION and EMOTION (F1;59 = 6.47, p = .014), meaning that especially negative and correctly recognized targets were rated as more negative than negative, but correctly rejected distrac-tors (t = 2.39, df = 61, p = .020). The main effect of GROUP (F2;59 = 0.27, p = .762) and the interactions of GROUP with RECOGNITION (F2;59 = 0.88, p = .419) and with EMOTION (F2;59 = 0.81, p = .450) were not significant. Finally, we found a three-way interaction of RECOGNITION, EMOTION and GROUP (F2;59 = 5.79, p = .005), but the effect depended on the new pictures
C.D. Wiesner et al./Neurobiology of Learning and Memory xxx (2015) xxx-xxx
(correct rejections) instead of the old ones (hits) (see Fig. 3B). As confirmed by t-tests, only negative, correctly rejected distractors were rated as less negative than negative, correctly recognized targets (t = -4.69, df=61, p <.001). This effect was stronger (t = -2.60, df = 39, p = .013) in the REMD group (t = -4.99, df = 20, p <.001) than in the SWS-deprived one (t = -2.00, df=19, p = .06) and stronger (t = -3.33, df=40, p = .002) in the REMD group than in the Wake group (t = -1.24, df =20, p = .231). In other words, the REMD group showed a slightly less negative affect to new negative pictures as compared to the SWSD group or the Wake group (see Table 4).
4. Discussion
In the present experiment, we investigated the ''sleep to remember - sleep to forget'' hypothesis using either REM-sleep-or SWS-deprivation as compared to wake (Goldstein & Walker, 2014; Walker, 2009). We found some evidence for the ''sleep to remember''-part of the hypothesis. Sleep fostered the consolidation of neutral and emotional pictures. Note that neither the awakenings nor the lower subjective wakefulness in the sleep groups had a detrimental effect on the sleep-dependent consolidation. Also emotional arousal helped to consolidate the pictures. This emotional memory bias was only significant in the SWSD group, where a normal amount of REM sleep was present, and in the Wake group. One would expect a stronger emotional memory bias in the SWSD group than in the Wake group, but one should keep in mind that this emotional memory bias occurred in the SWSD group on top of a significantly higher, overall memory retention. The interaction of GROUP, EMOTION, and TIME, however, failed to reach significance. Therefore, we cannot prove that the emotional memory bias was greater when slow wave sleep and REM sleep was present (SWSD) as compared to slow wave sleep without REM sleep (REMD) or no sleep at all (Wake). The lack of a significant emotional memory bias in the REMD group could be due to a lack of power (false negative) or, as intended, the result of the REM-sleep deprivation (true negative). However, a closer look at only the emotional pictures revealed that the retention of emotional pictures was best when REM sleep was present, such as in the SWSD group, and worse when REM sleep was deprived or when the subjects stayed awake. Thus, this exploratory analysis confirmed the prediction of the ''sleep-to-remember''-hypothesis. The same pattern occurs in the correlations of the amount of REM sleep with the extent of the emotional memory bias. Again, only in the SWSD group was more REM sleep associated with the preferred consolidation of emotional as opposed to neutral memories. Note, however, that the narrower range of REM sleep durations in the REM-sleep-deprived group could account for the low correlation in this group. The results can be interpreted in terms of the sequential hypothesis of memory consolidation (Giuditta et al., 1995; Walker& Stickgold, 2010). It seems that the succession of sleep containing slow wave sleep and REM sleep is advantageous for the consolidation of emotional memories. This was only true in the SWSD group which also showed a significant emotional memory bias. The participants of this group had 167 min of N2 on average which still may have contained enough slow waves (although less than 20% per epoch). Also, only these participants had a substantial amount of REM sleep (56 min compared to 2 min in the REM-sleep-deprived group), and the amount correlated with the emotional memory bias. In our previous study, the undisturbed sleep group had 57.79 ± 6.34 min REM sleep under the same laboratory conditions (Morgenthaler et al., 2014). In this sense, the SWS-deprived group had a ''normal'' amount of REM sleep. Furthermore, the amount of 56 min REM sleep is still well in the range of normal sleep in healthy adults (Redline et al.,
2004). On the other hand, the effect of REM-sleep-deprivation on the consolidation of emotional memories was not catastrophic but only strong enough to weaken the emotional memory bias.
Our results are in accordance with the studies by Wagner et al. (2001), Groch et al. (2013) and Groch et al. (2014). All three of these studies compared the effects of SWS-rich sleep in the first half of the night with REM-sleep-rich sleep in the second half (split-night design) and revealed that REM-sleep-rich late sleep selectively fostered the consolidation of emotional memories. Furthermore, the study by Nishida et al. (2009) matches our results. The authors compared the effects of a short nap and an interval spent awake (nap-design) on emotional memory consolidation. The authors report that a nap fostered the consolidation of emotional but not neutral memories and that the amount of REM sleep correlated with the extent of emotional memory facilitation.
However, there are also four studies whose results contradict ours. Baran and colleagues used the day-night design with subsequent correlations (Baran, Pace-Schott, Ericson, & Spencer, 2012). Sleep fostered the consolidation of neutral and emotional pictures as well, and consolidation of emotional pictures did not correlate with the amount of REM sleep. Cairney and colleagues used a correlative design to investigate the effect of sleep on the consolidation of positive, neutral, and negative pictures (Cairney, Durrant, Power, & Lewis, 2014). The study was not able to show any effect of valence category on sleep-dependent consolidation. However, SWS positively correlated with the memory for negative pictures. REM sleep, as opposed to the ''sleep to remember''-hypothesis, correlated negatively with the memory for positive pictures. Using the day-night design, we previously demonstrated a preferred consolidation of emotional pictures during sleep in healthy children but not in adults (Prehn-Kristensen et al., 2013). Although theta power during REM sleep correlated with the emotional memory bias, the amounts of REM sleep and SWS did not. The lack of a correlation with sleep stages might be due to the picture set which had been selected to be appropriate for children. Extremely arousing pictures depicting death, open wounds, or extreme violence, for example, had to be omitted for ethical reasons. In a recent study, we used a picture set suitable only for adults and compared the effect of selective REM-sleep deprivation, undisturbed sleep, and wakefulness on the consolidation of neutral and negative pictures (Morgenthaler et al., 2014). Sleep fostered the consolidation of neutral and negative pictures, and negative pictures were consolidated better than neutral ones. Although the deprivation reduced REM sleep to 5 min per night as compared to 58 min in the undisturbed sleep condition, the deprivation had no effect whatsoever. That study did not employ a control condition with non-REM awakenings and did not encompass affective ratings at the delayed recognition test. In the better-controlled, present study we did find evidence that emotional memories are preferably consolidated when REM sleep is present, even when SWS is massively reduced. However, the interaction was not significant, leaving open whether or not REM sleep is necessary but still beneficial for the consolidation of emotional memories.
We did not find any evidence in favor of the ''sleep to forget''-part of the hypothesis. Neither the comparison of the affective ratings during encoding and recognition nor of the affective ratings of recognized targets and rejected distractors supported the hypothesis that REM sleep dampens the emotional reaction to remembered stimuli and that this effect would be related to the consolidation of memories. Only new, negative pictures were rated slightly less negative by the REMD group. On the other hand, the paradigm produced reliable enough ratings to show the difference between neutral and negative pictures.
The only study at odds with our results is the study by van der Helm et al. (2011). The authors reported a correlation between a
REM sleep EEG parameter and the attenuation of the emotional reaction to repeatedly viewed pictures. They used a day-night comparison and specifically looked at the role of REM sleep physiology for post-sleep emotional reactivity. They focus on pre-frontal gamma EEG activity as a marker of central adrenergic activity and report that this marker predicted emotional reactivity measured by amygdala activity and behavioral reactivity. Specifically, they found that anterior prefrontal gamma activity during REM sleep was correlated with a decrease in emotional reactivity to pictures. Unfortunately, the authors do not report correlations of the amount of REM sleep or SWS with emotional reactivity.
However, there are several studies in accordance with our results which contradict the ''sleep to forget''-hypothesis. Wagner, Fischer, and Born (2002) used the split-night design to investigate the influence of SWS versus REM sleep on changes in emotional responses to aversive pictures. The results oppose the ''sleep-to-forget''-hypothesis formulated years later by Walker (2009). What Wagner and colleagues found in this well controlled experiment was an enhanced subsequent arousal rating of the pictures following sleep. Moreover, REM-rich late sleep shifted the valence ratings in the negative direction, whereas SWS-rich early sleep tended to shift the valence to the positive side. In a REM-sleep-deprivation study, Lara-Carrasco and colleagues investigated the emotional adaptation to repeatedly evaluated neutral and negative pictures as proposed by the ''sleep to forget''-hypothesis (Lara-Carrasco et al., 2009). Contrary to the hypothesis, the emotional adaptation was greater in the group deprived of REM sleep than in the undisturbed group. REM-sleep deprivation caused less arousal to negative pictures. Also, REM sleep enhanced the reactivity to negative stimuli in the morning. However, the deprivation was not very rigid, leaving 53 min of REM sleep on average in the deprived group. In a nap study, Pace-Schott and colleagues looked for inter-session habituation to emotional stimuli (Pace-Schott et al., 2011). In the study by Pace-Schott and colleagues, a nap of 84 min did not change the subjective emotional evaluations of the pictures. Of note, the authors used only 6 negative and 6 neutral pictures. However, as opposed to the ''sleep to forget''-hy-pothesis, SWS during the nap was significantly associated with greater habituation of the electromyographic response and REM sleep with lesser habituation of skin-conductance response to negative pictures. Baran and colleagues performed an experiment comparing the effect of a night of sleep and a day of wakefulness on emotional reactivity to and incidental memory for neutral and emotional pictures (Baran et al., 2012). The emotional reactivity to the negative pictures decreased more during wakefulness and was protected by sleep. As opposed to the ''sleep to forget''-hy-pothesis, the more REM sleep a participant had, the more the emotional reactivity to negative pictures was conserved. The study by Groch et al. (2013) also investigated the influence of early and late-night sleep on the evaluation of negative and neutral pictures. Although REM sleep and SWS differentially influenced memory consolidation, no effect on emotional evaluation was apparent. However, the authors did not obtain affective ratings prior to early or late sleep, although another study showed that emotional stories elicited higher arousal when encoded in the middle of the night (late sleep condition) as compared to the evening (early sleep condition) (Wagner et al., 2001). Therefore, the change of emotional evaluation might have been obscured. In summary, there are now six studies which either were not able to replicate (Groch et al., 2013; present study) the effect of REM sleep on the emotional evaluation of repeatedly presented stimuli proposed by van der Helm et al. (2011) or even found contradicting results (Baran et al., 2012; Lara-Carrasco et al., 2009; Pace-Schott et al., 2011; Wagner et al., 2001). Moreover, the only studies using experimental manipulations of REM sleep as compared to post hoc correlational
analysis oppose the ''sleep to forget''-hypothesis (Baran et al., 2012; Groch et al., 2013; Lara-Carrasco et al., 2009; Wagner et al., 2002), including the present study.
What, then, are the limitations and alternative explanations of our results? The main limitation of our results is that the 3-way interaction of GROUP; EMOTION, and TIME was not significant. Emotional facilitation was not only present in the SWSD group with intact REM sleep but also in the Wake group. The effect was smaller and not significant in the REMD group. Note, however, that both sleep groups showed better consolidation of neutral and emotional pictures than the Wake group. It seems that other aspects of sleep which are also present in N1 and N2 are helpful to foster the consolidation of declarative memories and that the role of REM sleep is to modulate this consolidation, i.e. strengthening emotional memory representations and weakening neutral ones. This is supported by two aspects of our results: Firstly, the amount of REM sleep in the SWSD group correlated with the emotional memory bias. Secondly, the retention of emotional pictures only was best in the SWSD group with REM sleep present.
Another limitation might be the higher amount of N2 sleep in the SWSD group (166.51 min) as compared to the REMD group (139.12 min). It has been argued that sleep spindles in N2 foster the consolidation of declarative memories (Fogel & Smith, 2011) and might enhance the emotional memory bias in sleep-dependent consolidation (Kaestner, Wixted, & Mednick, 2013). Kaester et al. (2013) observed an emotional memory bias only in a zolpidem condition but not in a placebo control condition. Zolpidem pharmacologically increases sleep spindle activity leading to the assumption that spindles are causally linked to emotional memory. However, Zolpidem simultaneously increased SWS duration which in turn weakened the interpretation that spindles caused the emotional memory bias. Furthermore, selective deprivation of REM sleep or SWS does not increase spindle density or absolute spindle activity (Genzel et al., 2009). Moreover, we only found a correlation of sleep spindles in S2 with the recognition accuracy for neutral but not emotional pictures in one of our previous studies (Goder et al., in press). The study is in line with another recent study showing that the correlation of S2 sleep spindles with memory performance was neither specific for emotional memories nor for sleep spindles during the consolidation night (Ward, Peters, & Smith, 2014). This corresponds to the notion that sleep spindles are a trait that correlates with intelligence (Fogel & Smith, 2011). However, this does not rule out that the higher amount of S2 in our SWSD group could explain the more pronounced sleep-dependent emotional memory bias in this group.
It has been repeatedly argued that REM-sleep awakenings might be stressful and therefore hinder the consolidation of memories. There are three reasons why this argument cannot explain our results. First, the number of awakenings was even higher in the group with deprivation of SWS. Nevertheless, the consolidation-enhancing effect of emotion was significant in the SWSD group but not in the REM-sleep-deprived group which encountered fewer awakenings. Therefore, awakenings per se cannot be the reason for the missing effect in the REM-sleep deprived group. Second, despite the selective deprivation of REM sleep or SWS and a slightly lower subjective wakefulness, both sleep groups showed better retention of pictures than the Wake group. Therefore, awakenings per se do not hinder consolidation. Furthermore, overall memory performance in the SWSD group was on the same level as in the REMD group, although the SWSD group had been awakened more often. Third, after the deprivation night both groups were very comparable regarding tiredness, good versus bad mood, and calm versus nervous state as measured by a reliable and valid questionnaire (Het & Wolf, 2007). Therefore, it is reasonable to assume that our results cannot be explained by the effect of awakenings per se. In future studies it would be advisable to measure
C.D. Wiesner et al./Neurobiology of Learning and Memory xxx (2015) xxx-xxx
the level of stress hormone, such as cortisol to further rule out the notion that awakenings cause intense stress and thereby influence consolidation.
Another question often discussed is the possible influence of the circadian rhythm. The time of day of encoding and recognition was different in the sleep groups (evening/morning) as compared to the Wake group (morning/evening). However, other studies showed that sleep-dependent consolidation of declarative memory is independent of time of day (Gais, Lucas, & Born, 2006) and that the emotional memory bias in sleep-dependent consolidation can also be demonstrated when controlling for circadian effects (Wagner et al., 2001).
A further limitation of our study is that we did not use arousing stimuli with positive valence. Therefore, our results cannot be generalized to all arousing events. Furthermore, the SFSR hypothesis might only be true for very intense or even traumatic negative events (compare Goldstein & Walker, 2014). At least for the ''sleep to remember''-part of the hypothesis, Menz et al. (2013) found some correlational evidence that REM sleep is important for the consolidation of fear memory.
Another limitation might be that the immediate recognition test encompassed fewer items than the delayed recognition test. However, more pictures would have prolonged the encoding phase unreasonably and increased the risk of emotional blunting toward the end of the picture series. Moreover, if the reliability of the immediate recognition test had been too low, we would not have found such a strong main effect of TIME (g2 = .798) or an interaction of TIME and EMOTION (g2 = .278) in the ANOVA.
In summary, our data suggest three conclusions. First, REM sleep has no effect on the change of the emotional evaluation of memorized stimuli (''sleep to forget''). Second, sleep fosters the consolidation of declarative memories even when the amount of REM sleep or SWS is greatly reduced. Third, REM sleep modulates the consolidation of declarative memories by increasing the emotional facilitation (''sleep to remember''), but the effect is not very strong and the consequences of an almost complete deprivation of REM sleep are far from devastating.
Funding
This study was supported by a grant (SFB 654 ''Plasticity and Sleep'') from the Deutsche Forschungsgemeinschaft. However, neither the Deutsche Forschungsgemeinschaft nor any of its associates had any role in the design of the study, in the collection, analysis, or interpretation of data, in the writing of the paper, or in the decision to submit the paper for publication.
Acknowledgments
We would like to thank Nicola Wendisch for technical assistance.
References
Aserinsky, E., & Kleitman, N. (1953). Regularly occurring periods of eye motility, and
concomitant phenomena, during sleep. Science, 118, 273-274. Baran, B., Pace-Schott, E. F., Ericson, C., & Spencer, R. M. (2012). Processing of emotional reactivity and emotional memory over sleep. Journal of Neuroscience, 32,1035-1042.
Berry, R B., Brooks, R., Gamaldo, C. E., Harding, S. M., Lloyd, R. M., Marcus, C. L., et al. (2013). The AASM manual for the scoring of sleep and associated events: rules terminology and technical specifications (2.0.2 ed.). Darien, Illinois, USA: American Academy of Sleep Medicine. Bleich, S., Havemann-Reinecke, U., & Kornhuber, J. (2002). FTNA. Fagerstrom-Test fur
Nikotinabhangigkeit. Gottingen, Germany: Hogrefe. Bradley, M. M., & Lang, P. J. (1994). Measuring emotion: The self-assessment manikin and the semantic differential. Journal of Behavior Therapy and Experimental Psychiatry, 25, 49-59.
Buysse, D. J., Reynolds, C. F., 3rd, Monk, T. H., Berman, S. R., & Kupfer, D. J. (1989). The Pittsburgh sleep quality index: A new instrument for psychiatric practice and research. Psychiatry Research, 28,193-213.
Cairney, S. A., Durrant, S. J., Power, R., & Lewis, P. A. (2014). Complementary roles of slow-wave sleep and rapid eye movement sleep in emotional memory consolidation. Cereb Cortex.
Dement, W. (1960). The effect of dream deprivation. Science, 131,1705-1707.
Fogel, S. M., & Smith, C. T. (2011). The function of the sleep spindle: A physiological index of intelligence and a mechanism for sleep-dependent memory consolidation. Neuroscience and Biobehavioral Reviews, 35,1154-1165.
Franke, G. H., & Derogatis, L. R (2002). SCL-90-R: Symptom-Checkliste von L.R. Derogatis -Deutsche Version. Göttingen, Germany: Hogrefe.
Gais, S., Lucas, B., & Born, J. (2006). Sleep after learning aids memory recall. Learning & Memory, 13, 259-262.
Genzel, L., Dresler, M., Wehrle, R., Grozinger, M., & Steiger, A. (2009). Slow wave sleep and REM sleep awakenings do not affect sleep dependent memory consolidation. Sleep, 32, 302-310.
Giuditta, A., Ambrosini, M. V., Montagnese, P., Mandile, P., Cotugno, M., Grassi Zucconi, G., et al. (1995). The sequential hypothesis of the function of sleep. Behavioural Brain Research, 69,157-166.
Göder, R., Graf, A., Ballhausen, F., Weinhold, S., Baier, P. C., Junghanns, K., et al. (in press). Impairment of sleep-related memory consolidation in schizophrenia: Relevance of sleep spindles? Sleep Medicine.
Goldstein, A. N., & Walker, M. P. (2014). The role of sleep in emotional brain function. Annual Review of Clinical Psychology, 10, 679-708.
Green, D. M., & Swets, J. A. (1966). Signal detection theory and psychophysics. New York, USA: Wiley.
Groch, S., Wilhelm, I., Diekelmann, S., & Born, J. (2013). The role of REM sleep in the processing of emotional memories: Evidence from behavior and event-related potentials. Neurobiology of Learning and Memory, 99, 1-9.
Groch, S., Zinke, K., Wilhelm, I., & Born, J. (2014). Dissociating the contributions of slow-wave sleep and rapid eye movement sleep to emotional item and source memory. Neurobiology of Learning and Memory.
Het, S., & Wolf, O. T. (2007). Mood changes in response to psychosocial stress in healthy young women: Effects of pretreatment with cortisol. Behavioral Neuroscience, 121,11-20.
Hoffmann, R. M., Müller, T., Hajak, G., & Cassel, W. (1997). Abend-Morgenprotokolle in Schlafforschung und Schlafmedizin - Ein Standardinstrument für den deutschsprachigen Raum. Somnologie, 1 , 103-109.
Kaestner, E. J., Wixted, J. T., & Mednick, S. C. (2013). Pharmacologically increasing sleep spindles enhances recognition for negative and high-arousal memories. Journal of Cognitive Neuroscience, 25,1597-1610.
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (2005). International Affective Picture System (IAPS): Affective ratings of pictures and instruction manual. Technical Report A-6. Gainsville, USA: University of Florida.
Lara-Carrasco, J., Nielsen, T. A., Solomonova, E., Levrier, K., & Popova, A. (2009). Overnight emotional adaptation to negative stimuli is altered by REM sleep deprivation and is correlated with intervening dream emotions. Journal of Sleep Research, 18,178-187.
Menz, M. M., Rihm, J. S., Salari, N., Born, J., Kalisch, R., Pape, H. C., et al. (2013). The role of sleep and sleep deprivation in consolidating fear memories. Neuroimage, 75, 87-96.
Morgenthaler, J., Wiesner, C. D., Hinze, K., Abels, L. C., Prehn-Kristensen, A., & Goder, R. (2014). Selective REM-sleep deprivation does not diminish emotional memory consolidation in young healthy subjects. PLoS ONE, 9, e89849.
Nishida, M., Pearsall, J., Buckner, R. L., & Walker, M. P. (2009). REM sleep, prefrontal theta, and the consolidation of human emotional memory. Cerebral Cortex, 19, 1158-1166.
Oldfield, R. C. (1971). The assessment and analysis of handedness: The edinburgh inventory. Neuropsychologia, 9, 97-113.
Pace-Schott, E. F., Shepherd, E., Spencer, R. M., Marcello, M., Tucker, M., Propper, R. E., et al. (2011). Napping promotes inter-session habituation to emotional stimuli. Neurobiology of Learning and Memory, 95, 24-36.
Prehn-Kristensen, A., Göder, R., Fischer, J., Wilhelm, I., Seeck-Hirschner, M., Aldenhoff, J., et al. (2011). Reduced sleep-associated consolidation of declarative memory in attention-deficit/hyperactivity disorder. Sleep Medicine, 12, 672-679.
Prehn-Kristensen, A., Munz, M., Molzow, I., Wilhelm, I., Wiesner, C. D., & Baving, L. (2013). Sleep promotes consolidation of emotional memory in healthy children but not in children with attention-deficit hyperactivity disorder. PLoS ONE, 8, e65098.
Rasch, B., & Born, J. (2013). About sleep's role in memory. Physiological Reviews, 93, 681-766.
Redline, S., Kirchner, H. L., Quan, S. F., Gottlieb, D. J., Kapur, V., & Newman, A. (2004). The effects of age, sex, ethnicity, and sleep-disordered breathing on sleep architecture. Archives of Internal Medicine, 164, 406-418.
Steyer, R., Schwenkmezger, P., Notz, P., & Eid, M. (1997). Der Mehrdimensionale Befindlichkeitsfragebogen (MDBF). Handanweisung. Göttingen, Germany: Hogrefe.
van der Helm, E., Yao, J., Dutt, S., Rao, V., Saletin, J. M., & Walker, M. P. (2011). REM sleep depotentiates amygdala activity to previous emotional experiences. Current Biology, 21 , 2029-2032.
Wagner, U., Fischer, S., & Born, J. (2002). Changes in emotional responses to aversive pictures across periods rich in slow-wave sleep versus rapid eye movement sleep. Psychosomatic Medicine, 64, 627-634.
Wagner, U., Gais, S., & Born, J. (2001). Emotional memory formation is enhanced across sleep intervals with high amounts of rapid eye movement sleep. Learn & Memory, 8, 112-119.
Walker, M. P. (2009). The role of sleep in cognition and emotion. Annals of the New York Academy of Sciences, 1156,168-197.
Walker, M. P., & Stickgold, R (2010). Overnight alchemy: Sleep-dependent memory evolution. Nature Reviews Neuroscience, 11,218. author reply 218.
Walker, M. P., & van der Helm, E. (2009). Overnight therapy? The role of sleep in
emotional brain processing. Psychological Bulletin, 135, 731-748. Ward, M. P., Peters, K. R., & Smith, C. T. (2014). Effect of emotional and neutral declarative memory consolidation on sleep architecture. Experimental Brain Research, 232,1525-1534.
