Abstract
Both rapid eye movement (REM) sleep dreaming and non-rapid eye movement (NREM) sleep spindles have been linked to processes of memory consolidation. However, relationships between the two phenomena have yet to be explored. In a heterogeneous sample of 53 healthy subjects who had participated in a memory consolidation protocol and who varied in their self-reported recall of dreams and nightmares, we assessed overnight polysomnography, N2 sleep spindle density, REM dream word count, and retrospective estimates of how often they recall dreams, bad dreams (dysphoric dreams, no awakening), and nightmares (dysphoric dreams, with awakenings). Fast spindle density positively correlated with all measures of dream recall but was most robustly associated with bad dream recall and REM dream word count. Correlations with bad dream recall were particularly strong for spindles occurring in sleep cycles 2 and 3 and correlations with word count for cycles 1, 4, and 5. While slow spindle density showed opposite correlations with all of these measures, partialing out slow spindles attenuated, but did not eliminate, the fast spindle correlations. Results are largely consistent with the conclusion that fast sleep spindles are associated with a common trait factor that also influences dream recall. However, the results also raise the possibility that both spindles and dreaming are expressions of memory consolidation mechanisms, such as neural replay, that transcend sleep stage.
Highlights
- – Dream recall measures are correlated with the density of fast spindles from N2 sleep
- – The association is most prominent for measures of bad dream recall and REM dream word count
- – The two dream measures are associated with different N2 cycles of the night
- – Dreaming and spindles may reflect a shared trait or mechanism of sleep-dependent memory consolidation
Introduction
Two prevalent sleep phenomena – non-rapid eye movement (NREM) sleep spindles and rapid eye movement (REM) sleep dreaming – have been linked to the consolidation of new memories. Yet, direct relationships between the two phenomena have never been investigated. That these events occur in different sleep states might seem to preclude them being related, yet their joint association with memory consolidation processes and their correlations with other trait characteristics such as intelligence (Bodizs, Gombos, Ujma, and Kovacs, 2014; Foulkes, 1985) suggest that there may well be more direct links. Connections between thetwo are also suggested by many theories that consider memory consolidation to be a function of both REM and NREM sleep stages at different times of the sleep episode (Buzsaki, 1989; Ficca and Salzarulo, 2004; Fogel, Smith, and Beninger, 2009; Giuditta, 2014; Giuditta et al., 1995). Accordingly, the general goal of this study was to assess potential relationships between measures of N2 sleep spindles and dream recall.
Sleep spindles are a reliable index of intelligence and memory consolidation
The essential features of sleep spindles and methods for their detection are well described (for reviews, see Andrillon et al., 2011; O’Reilly, Godbout, Carrier, and Lina, 2015; O’Reilly and Nielsen, 2014; Wallant, Maquet, and Phillips, 2016). Their use as an index of intellect and memory consolidation has also been repeatedly validated, with robust associations being documented between sleep spindles and (a) trait measures of intelligence (for reviews, see Fogel and Smith, 2011) and (b) sleep-related measures of memory improvement (for reviews, see Astori, Wimmer, and Luthi, 2013; Fogel and Smith, 2011; Rasch and Born, 2013). As an indicator of trait abilities, spindle density correlates positively with intellectual competencies, such as general intelligence (Wechsler Memory Scale; Schabus et al., 2008), performance IQ (Multidimensional Aptitude Battery–II; Fogel, Nader, Cote, and Smith, 2007), fluid intelligence (Raven Progressive Matrices Test; Bódizs et al., 2005), verbal memory (Auditory Verbal Learning Test; Lafortune et al., 2014), and visuospatial memory (Rey–Osterrieth Complex Figure Test, ROCFT; Bódizs, Lazar, and Rigo, 2008).
As an indicator of memory processing during sleep, sleep spindles correlate with a number of different sleep-dependent memory consolidation measures. This is especially the case for higher frequency (“fast”) spindles (13–16 Hz), more prevalent over central, parietal, and occipital regions, than for lower frequency (“slow”) spindles (10–13 Hz), more prevalent over frontal regions (Andrillon et al., 2011; O’Reilly and Nielsen, 2014). For example, fast spindle density is elevated when participants show sleep-related improvement on tasks of verbal memory (Clemens, Fabo, and Halasz, 2005), visuospatial memory (Clemens, Fabo, and Halasz, 2006), and integration of new semantic information (Tamminen, Lambon Ralph, and Lewis, 2013; Tamminen, Payne, Stickgold, Wamsley, and Gaskell, 2010). A causal role for sleep spindles in memory is further demonstrated by protocols that enhance consolidation of new memories by increasing spindle density either pharmacologically (Kaestner, Wixted, and Mednick, 2013) or with transcranial alternating current stimulation (Lustenberger et al., 2016).
Neural theories of memory consolidation link sleep spindles to sleep “replay” events, that is, the reactivation of neural sequences learned during wake, by their observed associations with slow oscillations and sharp-wave ripples (Born, 2010; Ritter et al., 2015; Siapas and Wilson, 1998). Although most studies of memory replay find that neural replays occur during NREM sleep, which in humans is associated with less frequent and less vivid dreaming than is REM sleep, neural replay events also take place in REM sleep (Hennevin, Hars, Maho, and Bloch, 1995; Louie and Wilson, 2001; Poe, Nitz, McNaughton, and Barnes, 2000; Poe, Walsh, and Bjorness, 2010). Some characteristics of neural replay events during NREM sleep suggest a similarity to those seen during dreaming. For example, the finding that complete sequences are replayed much less often than that are short sequence fragments (Mahoney, Titiz, Hernan, and Scott, 2016) parallels the observation that dreaming about complete episodic memories is rare, whereas dreaming about short memory fragments is not (Fosse, Fosse, Hobson, and Stickgold, 2003; Malinowski and Horton, 2014). On the other hand, many features of NREM sleep replays have no clear phenomenological analog in human dreaming; for example, replays may occur in reverse order (Diba and Buzsaki, 2007) or accelerated by as much as 7–10 times (Euston, Tatsuno, and McNaughton, 2007; Kudrimoti, Barnes, and McNaughton, 1999).
Dreaming is a possible index of memory consolidation
Unlike findings for sleep spindles, a role for dreaming in memory consolidation remains uncertain. Theories have long held that dreaming serves a memory-enhancing function (Breger, 1969; Dewan, 1970; Palombo, 1976) with a wide variety of mechanisms having been proposed. Most theories implicate some form of replay of waking memory during dreaming; these may be quite general, as in the reprocessing of unassimilated “lessons” from waking experience (Dewan, 1970) or the realistic simulations of real-world threats (Revonsuo, 2000), or they may be relatively restrained and fragmented, such as the reactivation of fear elements in safe contexts (Nielsen and Levin, 2007) or the repetition of memory fragments in novel or bizarre contexts (Horton and Malinowski, 2015).
Empirical tests have supported this notion of replay during dreaming to a limited extent (for reviews, see Smith, 2010; Wamsley and Stickgold, 2011) with several studies supporting a role for dreaming in memory improvement (Cipolli, Fagioli, Mazzetti, and Tuozzi, 2004; De Koninck, Christ, Hebert, and Rinfret, 1990; De Koninck, Christ, Rinfret, and Proulx, 1988; De Koninck, Prévost, and Lortie-Lussier, 1996; Dumel et al., 2015; Erlacher and Schredl, 2010; Fiss, Kremer, and Lichtman, 1977; Pantoja et al., 2009; Schredl and Erlacher, 2010; Wamsley, Tucker, Payne, Benavides, and Stickgold, 2010) and several others finding no such effect (Cipolli, Bolzani, Tuozzi, and Fagioli, 2001; Nguyen, Tucker, Stickgold, and Wamsley, 2013; Nielsen et al., 2015; Plailly, Villalba, Nicolas, and Ruby, 2016; Sabourin, Forest, Hebert, and De Koninck, 2006; Schredl and Erlacher, 2010; Wamsley, Perry, Djonlagic, Reaven, and Stickgold, 2010). To illustrate one positive finding, in a frequently cited study (Wamsley, Tucker, et al., 2010), participants learned to navigate a virtual maze on a computer screen and were tested following a 90-min period of either napping or wakefulness. Participants who slept improved the most, and those who showed the biggest gains were also those who reported dreams in which the “the task was explicitly and unambiguously mentioned” (p. 853). Having had task-related thoughts in the waking condition did not predict improved performance. As compelling as such findings might be, the work has weaknesses such as the exclusive use of NREM dreams and very little dreaming about the task (n=4/50 subjects). However, unlike most others studying this question, Wamsley’s group attempted to replicate their findings using either the same (Wamsley, Nguyen, Tucker, Olsen, and Stickgold, 2012) or a similar virtual task (Nguyen et al., 2013; Wamsley, Perry, et al., 2010). They claim to have succeeded on one of these attempts (Wamsley, 2014) although a detailed report has not yet been published (Wamsley, Nguyen, et al., 2012).
An indirect connection between sleep spindles and dreaming is that both phenomena are related to trait measures of cognitive processes. First, like sleep spindles, dream recall has been linked to measures of intelligence. Laboratory dream recall through childhood and adolescence is quite reliably associated with high performance scores on the Block Design of the Wechsler scales (Foulkes, 1985). Second, intelligence (reasoning, verbal abilities) assessed by the Cambridge Brain Sciences Trials (Hampshire, Highfield, Parkin, and Owen, 2012) is correlated with the extent of dreaming about a newly learned task (Fogel, Ray, Sergeeva, and Owen, 2016). Third, like sleep spindles, dream recall frequency has been linked to visual memory in adults (Butler and Watson, 1985; Cory, Ormiston, Simmel, and Dainoff, 1975; Schredl and Montasser, 1996). Fourth, like sleep spindles, dream recall capacity may have trait-like electroencephalography (EEG) signatures, such as increased theta oscillations (Scarpelli et al., 2015) or delayed P3 reactivity (Eichenlaub, Bertrand, Morlet, and Ruby, 2013). Finally, the association of both dreaming and sleep spindles with various psychopathologies, such as schizophrenia (Lusignan et al., 2009; Wamsley, Tucker, et al., 2012), is consistent with the notion that these two phenomena are linked.
Briefly, whereas sleep spindles are a well-characterized, well-validated marker of intellectual abilities and sleep-dependent memory mechanisms, dreaming’s association with intelligence and memory is less certain. Contemporary theories are nonetheless consistent with the possibility that the two sleep phenomena are interrelated by virtue of neural replay mechanisms occurring in different forms during different sleep stages. Research with human subjects suggests a range of dream replay phenomena in which participants dream to different degrees about experimental stimuli (for reviews, see Arkin and Antrobus, 1991; Smith, 2010), whereas research with animals suggests that neural replays occur in both REM and NREM sleep and match the phenomenological features of dream replays in some respects (e.g., fragmentary nature) but not in others (e.g., reverse or accelerated replay). Thus, there are both empirical and conceptual grounds on which to expect that NREM sleep spindles may be related to dreaming. This study addresses this open question.
Objectives and hypotheses
This study aimed to examine relationships between sleep spindles and measures of dream recall. We focused on spindles recorded from six standard electrode derivations (F3, F4, C3, C4, O1, and O2) for the whole sleep period. And, because spindle attributes have been observed to change with increasing depth of sleep (Andrillon et al., 2011), we assessed spindles separately for each of the first five consecutive N2 sleep cycles. We used available polysomnography (PSG) recordings for a heterogeneous sample of healthy, non-clinical participants who varied in their self-reported frequencies of recalling dreams and nightmares. To assess differences in dream recall frequency, retrospective measures of recalling dreams, bad dreams, and nightmares were used. We hypothesized that fast spindle density in particular would correlate positively with measures of dream recall frequency.
Methods
Participants
A total of 53 participants [20 males (M) and 33 females (F) with the mean age of 24.2 ± 4.2 years] were available in the Dream & Nightmare Laboratory database who had all completed the same night-time protocol. These included 35 participants (17 M and 18 F with the mean age of 24.8 ± 4.1 years) whose results are published in a paper on sleep-related memory among high- and low-frequency dream recallers (Dumel et al., 2015) and 18 participants (3 M and 15 F with the mean age of 23.2 ± 4.1 years) whose results are unpublished; some of these participants reported high frequencies of dream recall, bad dreams, and/or nightmares, whereas the others reported low frequencies. All participants were initially recruited by word of mouth and advertisements on campus and on our Center website. Potential participants then completed a telephone screening interview that included reporting their retrospective monthly recall of dreams (Dream recall), bad dreams (Bad dream recall), and nightmares (Nightmare recall). To be included, participants had to be 18–35 years old and fluent in English or French. They were excluded if they reported a major sleep disorder, if their sleep efficiency score was <80% on the polysomnographic evaluation, or if they reported an emotional or psychological disorder, a major medical or psychiatric condition, excessive consumption of alcohol, cigarettes/nicotine, or recreational drugs or use of medications that affect sleep. In short, all participants reported themselves to be physically and psychologically fit – even if some recalled nightmares or bad dreams more often than others. Each participant gave written informed consent and received $110 plus compensation for transport and breakfast expenses. The research was approved by institutional scientific and ethical boards.
Procedures
Participants completed a sleep and dream log from home by telephone 1 week prior to and 1 week following their overnight stay in the laboratory. They also completed a variety of questionnaires that are not all reported here [Beck Depression Inventory-II (BDI), State–Trait Anxiety Inventory (STAI), Morningness–Eveningness Questionnaire, Boundaries Short Form, Inventory of Dreams: Experiences and Attitudes, Sleep Disorders Questionnaire, Pittsburgh Sleep Quality Index, and Toronto Alexithymia Scale-20]. Only the Beck Depression and State–Trait Anxiety measures are considered further in the present analyses.
Participants underwent a pre-sleep testing session during which four cognitive tasks were administered: the Mirror-Tracing Task (MTT), Tower of Hanoi (ToH), Corsi Block-Tapping (CBT) Task, and ROCFT. Two intellectual tasks from the Wechsler Adult Intelligence Scales were also given: Block Design and Digit Span (total). Participants were prepared for polysomnographic recordings and allowed to sleep uninterrupted through the night. In the morning, they were awakened from REM sleep and asked to report and self-rate their sleep and dream experiences using the same rating scales contained in the home logs. They were subsequently given a light breakfast and then retested on the MTT and ROCFT. They returned to the laboratory 1 week later and completed a retest of all four tasks. Pre-/post-sleep measures of task improvement were available for three of the four tasks: MTT, CBT, and ToH.
PSG
An electrode montage of six EEG (F3, F4, C3, C4, O1, and O2) referenced to A1 (including A2 for re-referencing offline to A1 + A2), four electrooculography (EOG; vertical, horizontal), and four electromyography (EMG; chin, finger extensor, toe flexor, and corrugator) channels was recorded using a Grass M12 or M15 Neurodata Acquisition System (−6 dB filters, 0.30 and 100 Hz cutoffs) with 16-bit precision under the control of Harmonie 6.2b (Natus/Stellate Systems Inc., Montreal, Canada). The sampling rate for EEG, EOG, EMG, and reference derivations was 256 Hz; for electrocardiography (ECG), it was 512 Hz. The F4 derivation was defective for one participant and removed from further analyses. PSG tracings were scored with American Academy of Sleep Medicine standards (Silber et al., 2007) by an experienced PSG technician. REM sleep for laboratory awakenings was determined online by an experimenter familiar with PSG scoring. Standard sleep variables (e.g., REM min, REM%, N2 min, N2%, and total sleep time) were calculated offline by in-house software.
Spindle detection
Each spindle was detected automatically on six artifact-free derivations (F3, F4, C3, C4, O1, and O2; re-referenced to A1 + A2 offline) for the entire night. For nights on which the first REM sleep episode was skipped (Nielsen et al., 2010), cycle timing was adjusted by inserting a faux epoch of REM sleep at the point in the hypnogram where an experienced technician familiar with partial REM sleep signs (descending N2, muscle atonia, phasic EMG activity, rapid eye movements, and saw-tooth waves) determined that the skipped REM sleep episode should have occurred. This allowed accurate segmentation of each night into relatively equal length cycles, of which the first five N2 episodes were selected for more detailed analysis of spindles. Raw digitized signals were bandpass filtered from 11 to 16 Hz using a linear phase finite impulse response filter (−3 dB at 11.1 and 15.9 Hz). Forward and reverse filtering was performed to obtain zero-phase distortion and double the filter order. The root mean square (RMS) of the filtered signal was then calculated with a 0.25-s time window and thresholded at the 95th percentile (Martin et al., 2013). A spindle was identified when at least two consecutive RMS time points exceeded this threshold and the event duration met the criterion of 0.5 s. The resulting output files were processed to verify the distribution of sleep spindle frequency; a cutoff of 12.99 Hz was used to distinguish slow (11.00–12.99 Hz) from fast (13.00–16.00 Hz) spindles; counts of slow and fast spindles were conducted automatically and spindle densities for the entire night and for each of the first five cycles were calculated by dividing each spindle count on each channel by the number of minutes of time elapsed in artifact free N2 over the entire night or over the corresponding sleep cycle as appropriate. In Fig. 1, distributions of the entire sample of recorded spindles for the 53 participants are plotted by 0.25 Hz frequency bin (11.0–16.0 Hz range) for each electrode derivation. The normal preponderance of slow spindles in frontal derivations is clearly evident from this plot as is the normal distribution of faster spindles in the central and parietal derivations.
Distributions of all-night spindle counts by frequency (bin=.25 Hz) and derivation for all participants. The predominance of slow spindles (11.00–12.99 Hz) in frontal derivations (F3, F4) and fast spindles (13.00–15.00 Hz) in all other derivations is apparent
Citation: Sleep Spindles & Cortical Up States 1, 1; 10.1556/2053.1.2016.003
Statistics
Since the dream recall measures were not normally distributed, they were log transformed. In addition, correlations between spindle densities and other measures were determined using the conservative non-parametric Spearman coefficients when possible (partial correlations were performed with Pearson correlations). Multiple linear regressions (forward stepwise) were used to assess which dream recall measures (predictors) correlated most strongly with fast spindle density (criterion). Task improvement differences on the MTT, CBT, and ToH tasks between groups high and low in dream recall were determined by defining High and Low groups (median splits) on each of the four dream recall measures and then comparing these groups using four separate multivariate ANOVAs each of which used Group (High, Low) as an independent variable and improvement scores on the MTT, CBT, and ToH tasks as multiple dependent measures. Follow-up ANCOVAs were applied to determine which dream recall measures best accounted for the group differences observed.
Results
Demographic and personality characteristics of the study cohort are shown in Table 1 and sleep architecture attributes are shown in Table 2.
Demographic and personality attributes of the cohort (n=53)
| Mean ± SD | Range | |
|---|---|---|
| Age | 24.2 ± 4.2 | 18–35 |
| Dreams/month | 16.4 ± 14.4 | 0.5–70 |
| Bad dreams/month | 5.2 ± 6. 7 | 0–28 |
| Nightmares/month | 2.1 ± 4.4 | 0–22 |
| #Words/report | 138.2 ± 146.1 | 10–650 |
| STAI: trait | 40.3 ± 8.9 | 24–60 |
| STAI: state | 33.8 ± 9.3 | 20–60 |
| BDI-II | 8.3 ± 7.0 | 0–33 |
Sleep characteristics of the cohort
| Mean ± SD | Range | |
|---|---|---|
| Sleep latency (min) | 15.0 ± 13.0 | 2.0–65.5 |
| Latency to persistent sleep (min) | 19.1 ± 15.0 | 2.0–75.0 |
| N1 latency (min) | 14.2 ± 12.5 | 2.0–63.0 |
| N2 latency (min) | 22.4 ± 15.1 | 6.5–83.0 |
| N3 latency (min) | 20.7 ± 22.5 | 5.0–153.5 |
| REM latency (min) | 79.1 ± 25.8 | 50.0–181.5 |
| Sleep duration (min) | 419.8 ± 49.0 | 279.5–513.0 |
| Wake duration (min) | 43.9 ± 39.0 | 7.5–187.0 |
| Number of awakenings (number) | 13.8 ± 8.2 | 1–37 |
| Sleep efficiency (%) | 90.6 ± 8.0 | 62.5–98.4 |
| Wake (min) | 43.9 ± 39.0 | 7.5–187.0 |
| N1 (min) | 35.8 ± 20.1 | 11.0–122.5 |
| N2 (min) | 189.1 ± 38.3 | 102.5–262.0 |
| N3 (min) | 109.1 ± 33.7 | 45.5–181.5 |
| REM (min) | 85.8 ± 24.0 | 42.0–153.5 |
| NREM (min) | 334.0 ± 41.5 | 233.5–418.5 |
| Sleep (min) | 419.8 ± 49.0 | 279.5–513.0 |
| Wake (%) | 9.3 ± 7.9 | 1.6–35.8 |
| N1 (%) | 8.7 ± 5.5 | 2.8–37.5 |
| N2 (%) | 44.9 ± 6.9 | 30.4–60.0 |
| N3 (%) | 26.0 ± 7.4 | 11.4–40.5 |
| REM (%) | 20.3 ± 4.8 | 65.2–88.1 |
| NREM (%) | 79.7 ± 4.8 | 11.9–34.8 |
| Number of sleep cycles (number) | 4.7 ± 1.0 | 3–8 |
Dream recall measures
The three retrospective measures of dream recall frequency were very highly correlated with coefficients between .57 and .67 (all p < .000001; Table 3) but only Dream recall was moderately correlated with the #Words/report measure (r=.341, p=.013). Furthermore, although all three retrospective dream recall measures correlated negatively with depression and trait anxiety, these were significant only for Dream recall (all p < .01) and Bad dream recall (all p < .05); none were significant for Nightmare recall or #Words/report.
Spearman correlations (upper right) and p-values (lower left) between age, dream recall measures, anxiety, and depression scores
| Age | Dream recall | Bad dream | Nightmare | #Words/report | STAI: trait | STAI: state | BDI | |
|---|---|---|---|---|---|---|---|---|
| Age | −.090 | −.168 | −.173 | .056 | .295* | .212 | .039 | |
| Dream recall | .521 | .631* | .572* | .341* | −.360* | −.125 | −.431* | |
| Bad dream recall | .229 | .000 | .672* | .121 | −.284* | .076 | −.362* | |
| Nightmare recall | .217 | .000 | .000 | .098 | −.149 | .067 | −.165 | |
| #Words/report | .691 | .013 | .386 | .487 | −.034 | .107 | −.196 | |
| STAI: trait | .034 | .009 | .041 | .293 | .811 | .644* | .719* | |
| STAI: state | .132 | .376 | .592 | .638 | .448 | .000 | .428* | |
| BDI | .786 | .001 | .008 | .243 | .164 | .000 | .002 |
Note. Dream recall: log (dreams/month + 1); Bad dream recall: log (bad dreams/month + 1); Nightmare recall: log (nightmares/month + 1); #Words/report: number of content-bearing words in laboratory dream report; STAI: State–Trait Anxiety Scale, trait and state measures; BDI: Beck Depression Inventory-II.
*Correlations surpassing p < .05.
Fast versus slow sleep spindle densities
Spearman correlations revealed that the within-derivation densities of fast and slow spindles were inversely related (Table 4, diagonal). The average coefficient for the six derivations was r=−.875 (p < .001) and substantially larger than the average of between-derivation correlations (r=.759, p < .001). The highest within-derivation correlations occurred centrally (C3: −.919; C4: −.914), the lowest frontally (F3: −.828; F4: −.806), and intermediate values occipitally (O1: −.890; O2: −.895). To control for this covariance, partial correlations were conducted in examining the relationships between spindle densities and dream recall.
Spearman correlations between all-night fast and slow spindle densities.
| F3-fast | F4-fast | C3-fast | C4-fast | O1-fast | O2-fast | Mean | |
|---|---|---|---|---|---|---|---|
| F3-slow | −0.828 | −0.783 | −0.748 | −0.729 | −0.584 | −0.588 | −0.710 |
| F4-slow | −0.808 | −0.806 | −0.732 | −0.737 | −0.576 | −0.586 | −0.708 |
| C3-slow | −0.892 | −0.835 | −0.919 | −0.856 | −0.747 | −0.752 | −0.834 |
| C4-slow | −0.883 | −0.869 | −0.875 | −0.914 | −0.739 | −0.759 | −0.840 |
| O1-slow | −0.727 | −0.684 | −0.780 | −0.780 | −0.890 | −0.900 | −0.794 |
| O2-slow | −0.719 | −0.682 | −0.769 | −0.786 | −0.875 | −0.895 | −0.788 |
| Mean | −0.810 | −0.777 | −0.804 | −0.800 | −0.735 | −0.747 | −0.779 |
Note. Values in bold indicate within-derivation correlations; Values in italics denote derivation mean correlations; All p < .001.
Relation of Sleep Spindles to Dreaming
As shown in Fig. 2 and Table 5, all-night spindle density scores were correlated with the four dream recall measures. Fast spindle correlations were uniformly positive and robust for both Bad dream recall and #Words/report, where all six derivations surpassed at least the p < .05 significance threshold, and Dream recall, where five of the six correlations did. Correlations were least robust for Nightmare recall, where only C3 surpassed the p < .05 threshold.
Spearman correlations between (A) all-night fast (13.00–15.00 Hz) and (B) all-night slow (11.00–12.99 Hz) spindle densities and measures of Dream recall, Bad dream recall, Nightmare recall, and #Words/dream report. Scatter plots and Spearman correlations with Bad dream recall for (C) C4 derivation and (D) O2 derivation. *p < .05; **p < .01
Citation: Sleep Spindles & Cortical Up States 1, 1; 10.1556/2053.1.2016.003
Spearman coefficients and two-tailed p-values for all-night fast (13.00–16.00 Hz) and slow (11.00–12.99 Hz) spindle density correlations with four dream recall measures
| Spearman r | p (two-tailed) | |||||||
|---|---|---|---|---|---|---|---|---|
| Dream recall | Bad dream recall | Nightmare recall | #Words/report | Dream recall | Bad dream recall | Nightmare recall | #Words/report | |
| F3-fast | 0.275* | 0.342* | 0.243 | 0.367* | 0.047 | 0.012 | 0.079 | 0.007 |
| F4-fast | 0.220 | 0.334* | 0.223 | 0.314* | 0.118 | 0.016 | 0.112 | 0.023 |
| C3-fast | 0.362* | 0.381* | 0.281* | 0.348* | 0.008 | 0.005 | 0.041 | 0.011 |
| C4-fast | 0.337* | 0.362* | 0.211 | 0.424* | 0.014 | 0.008 | 0.130 | 0.002 |
| O1-fast | 0.370* | 0.365* | 0.198 | 0.321* | 0.006 | 0.007 | 0.154 | 0.019 |
| O2-fast | 0.354* | 0.369* | 0.220 | 0.330* | 0.009 | 0.006 | 0.114 | 0.016 |
| F3-slow | −0.301* | −0.222 | −0.217 | −0.331* | 0.029 | 0.111 | 0.119 | 0.015 |
| F4-slow | −0.286* | −0.222 | −0.235 | −0.291* | 0.040 | 0.114 | 0.093 | 0.036 |
| C3-slow | −0.323* | −0.313* | −0.272* | −0.275* | 0.018 | 0.022 | 0.049 | 0.046 |
| C4-slow | −0.319* | −0.316* | −0.228 | −0.317* | 0.020 | 0.021 | 0.101 | 0.021 |
| O1-slow | −0.366* | −0.331* | −0.200 | −0.211 | 0.007 | 0.016 | 0.151 | 0.129 |
| O2-slow | −0.351* | −0.336* | −0.199 | −0.186 | 0.010 | 0.014 | 0.154 | 0.183 |
*Coefficients with p < .05.
Slow spindle correlations with the four dream recall measures were uniformly negative, although these were consistently beyond the p < .05 threshold only for Dream recall. Correlations between fast spindle density and Bad dream recall were attenuated but not eliminated for four derivations by partialing out slow spindle densities: F3 (r52=.259, p=.064), C4 (r52=.287, p=.039), O1 (r52=.264, p=.059), and O2 (r52=.326, p=.018); correlations were more severely reduced for the remaining two derivations: F4 (r51=.149, p=.298) and C3 (r52=.247 p=.078). For #Words/report, partialing out slow spindle densities reduced or eliminated all fast spindle correlations except O2, that is, F3 (r52=.114, p=.423), F4 (r51=.110, p=.441), C3 (r52=.065 p=.645), C4 (r52=.238, p=.089), O1 (r52=.207, p=.140), and O2 (r52=.272, p=.051). For Dream recall and Nightmare recall, partialing out slow spindle densities eliminated all correlations with fast spindle densities (all p > .120 and >.414, respectively). In all cases, the reverse procedure – partialing out fast spindle densities from correlations with slow spindle densities – completely eliminated the slow spindle correlations (all p > .258). Thus, there were unique relationships with fast spindle densities only for Bad dream recall and, to a lesser extent, for #Words/report.
A second set of partial correlations was, therefore, conducted to remove possible modulating effects of depression and trait anxiety from the Bad dream recall × fast spindle correlations because, as shown in Table 3, these two pathology measures were also correlated with Bad dream recall. Partialing out BDI scores attenuated but did not eliminate the correlations for five of the six derivations, that is, F3 (r52=.265, p=.061), F4 (r51=.206, p=.152), C3 (r52=.287 p=.041), C4 (r52=.322, p=.021), O1 (r52=.323, p=.021), and O2 (r52=.342, p=.014). Similarly, partialing out STAI-trait scores attenuated but did not eliminate the correlations for the same five derivations, that is, F3 (r52=.278, p=.048), F4 (r51=.215, p=.133), C3 (r52=.307 p=.028), C4 (r52=.344, p=.014), O1 (r52=.349, p=.012), and O2 (r52=.369, p=.008).
To determine which of the dream recall measures was most highly correlated with fast spindle density, first, a series of six linear multiple regressions was conducted in which the three retrospective recall measures – Dream recall, Bad dream recall, and Nightmare recall – served as predictor variables and all-night fast spindle densities for each of the six derivations served as the criterion variables. As shown in Table 6A, Bad dream recall was the sole significant predictor for five of the six derivations (F3, C3, C4, O1, and O2), but for F4, it was as predictive as were the other two measures.
Forward stepwise multiple regression solutions and 95% confidence intervals for dream recall predictors of all-night fast spindle density
| Predictor | R | R 2 | F | p | B | 95% CI | SE | |
|---|---|---|---|---|---|---|---|---|
| (A) Dream recall, Bad dream recall, and Nightmare recall | ||||||||
| F3 | Bad dream recall | .287 | .082 | 4.571 | .037* | 0.562 | 0.03–1.09 | 0.263 |
| F4a | Bad dream recall | .230 | .053 | 2.794 | .101 | 0.420 | −0.09–0.93 | 0.765 |
| C3 | Bad dream recall | .306 | .093 | 5.258 | .026* | 0.773 | 0.10–1.45 | 0.337 |
| C4 | Bad dream recall | .341 | .116 | 6.723 | .012* | 0.817 | 0.18–1.45 | 0.315 |
| O1 | Bad dream recall | .354 | .126 | 7.322 | .009* | 0.908 | 0.23–1.58 | 0.336 |
| O2 | Bad dream recall | .370 | .137 | 8.084 | .006* | 0.906 | 0.27–1.55 | 0.319 |
| (B) A + #Words/report | ||||||||
| F3 | #Words/report | .334 | .111 | 6.392 | .015* | 0.002 | 0.00–0.003 | 0.001 |
| F4 | #Words/report | .333 | .111 | 6.224 | .016* | 0.002 | 0.00–0.003 | 0.001 |
| C3 | Bad dream recall | .306 | .093 | 5.258 | .026* | 0.773 | 0.10–1.45 | 0.337 |
| C4 | 1. #Words/report 2. Bad dream recall |
.356 .452 |
.127 .204 |
7.399 4.882 |
.009* .032* |
0.002 0.680 |
0.00–0.004 0.06–1.30 |
0.001 0.308 |
| O1 | Bad dream recall | .354 | .126 | 7.322 | .009* | 0.908 | 0.23–1.58 | 0.336 |
| O2 | Bad dream recall | .370 | .137 | 8.084 | .006* | 0.906 | 0.27–1.55 | 0.319 |
Forced solution: three predictors contributed equally to the regression solution.
*Coefficients with p < .05.
Second, a similar set of six linear regressions was conducted in which #Words/report was added to the previous three predictors (Table 6B). In this analysis, Bad dream recall again predicted spindle density for C4, O1, and O2 derivations, whereas #Words/report predicted spindle density for F3 and F4. Both #Words/report and Bad dream recall independently predicted spindle density for C3.
Sleep spindle relationships by cycle of night
To assess if the observed relationships between dreaming and sleep spindles were uniform for spindles sampled from different times of night, Spearman correlations were calculated between the best two predictive measures – Bad dream recall and #Words/report – and fast spindle densities for each of the first five NREM sleep periods of the night. Some participants lacked cycles 4 (n=5) and 5 (n=22), whereas some lacked a recalled dream (n=8), and thus sample sizes were reduced for these cycles. As shown in Fig. 3, correlations were uniformly positive for all cycles and for both measures. For Bad dream recall, correlations surpassed the p < .05 threshold consistently only for cycles 2 and 3. For #Words/report, they surpassed p < .05 for all cycles on at least one derivation, especially F3, but never on all derivations in any one cycle. The largest number of significant correlations here (four of six derivations) was seen for cycle 4, but the largest coefficients (F3, C4; r > .40) were seen for cycle 5.
Correlations between fast spindle density and self-reported frequencies of Bad dream recall (top panel) and #Words/report (bottom panel) for the first five NREM cycles of the night. N by cycle for Bad dream recall: 53, 53, 53, 48, and 31; for #Words/report: 45, 45, 45, 43, and 30. *p < .05; **p < .01
Citation: Sleep Spindles & Cortical Up States 1, 1; 10.1556/2053.1.2016.003
Relation of dream recall and spindle density to task improvement
To determine whether dream recall frequencies and spindle densities were associated with improved performance on any of the three administered tasks, we calculated (a) Spearman correlations between improvement scores on the MTT, ToH, and CBT tasks and all-night fast spindle densities for the six derivations and (b) task improvement scores for participants scoring High and Low on the four dream recall measures as determined by median splits of the distributions. First, no substantial correlations were observed between task improvement scores and all-night spindle densities (all p > .14). Even when correlations were calculated for the five sleep cycles separately, correlations were minimal for the most part; the only exception was between improvement on the CBT task and fast spindle densities calculated for the fifth cycle. For this cycle, positive correlations were observed for all six derivations: F3 (r31=.388, p=.031), F4 (r31=.393, p=.029), C3 (r31=.393, p=.029), C4 (r30=.421, p=.018), O1 (r31=.412, p=.024), and O2 (r31=.353, p=.051). However, none of these correlations would withstand a conservative correction for multiple testing at .05/30 correlations or p < .002.
Second, multivariate ANOVAs revealed that the largest multivariate effect was for Bad dream recall (Hotelling’s T=.259, F4,48=3.121, p=.023, η2=.206) which included univariate effects for both MTT improvement (F1,51=8.357, p=.006, η2=.141) and CBT improvement (F1,51=4.318, p=.043, η2=.078). As shown in Fig. 4, High Bad dream recall participants showed more improvement on the MTT (M: 22.44 ± 18.42) and the CBT (M: 1.46 ± 2.77) than did Low Bad dream recall participants (MTT: 11.40 ± 7.27; CBT: 0.04 ± 2.19). Entering Dream recall as a covariate in this analysis reduced, but did not completely eliminate, these two effects, that is, neither MTT (F1,50=4.005, p=.051, η2=.074) nor CBT (F1,50=3.313, p=.075, η2=.062). However, entering Nightmare recall as a covariate eliminated the MTT effect (F1,50=0.674, p=.416, η2=.013) and attenuated the CBT effect (F1,50=3.204, p=.079, η2=.060).
Overnight levels of improvement on (A) Mirror-Tracing Task, (B) Corsi Block-Tapping Task, and (C and D) Tower of Hanoi #moves and time to completion for participants’ self-reporting high and low levels of recalling dreams, bad dreams, and nightmares. *p < .05; **p < .01
Citation: Sleep Spindles & Cortical Up States 1, 1; 10.1556/2053.1.2016.003
For the Dream recall measure, a multivariate main effect (T=.227, F4,48=2.721, p=.040, η2=.185) revealed only a single univariate difference: High Dream recall participants showed greater MTT improvement (M: 22.73 ± 18.79) than did Low participants (M: 11.53 ± 7.05; F1,51=8.602, p=.005, η2=.144). This effect was eliminated by entering either Bad dream recall (F1,50=1.761, p=.191, η2=.034) or Nightmare recall (F1,50=1.161, p=.286, η2=.023) as covariates.
Finally, for the Nightmare recall measure, a significant multivariate effect (T=.224, F4,48=2.683, p=.042, η2=.183) revealed one univariate effect for MTT improvement (F1,51=7.728, p=.008, η2=.132). As for the other two measures, High Nightmare recall participants showed more MTT improvement (M: 21.86 ± 17.94) than did Low participants (11.16 ± 7.28). This effect was not eliminated by covarying Dream recall (F1,50=4.791, p=.033, η2=.087) but was by covarying Bad dream recall (F1,50=2.273, p=.138, η2=.043). Finally, participants with High and Low recall on any measure did not improve differentially on the ToH improvement measures.
Briefly, retrospective dream recall measures were associated with improvements on MTT (all three measures) and CBT (Bad dream recall only) with the most robust differences being shown by participants differing in High and Low Bad dream recall.
For the prospective dream recall measure, in contrast, the results were consistently negative. Participants High (≥91 words) and Low (<91 words) on #Words/report did not differ on any of the task improvement scores (all p > .301).
Relation of dream recall and spindle density to intellectual abilities
Spearman correlations between dream recall measures and intellectual abilities (Block Design and Digit Span) are shown in Table 7. For Block Design, we replicated, albeit marginally, an expected (Foulkes, 1985) positive correlation with #Words/report (r=.282, p=.061) but also observed a negative, again marginal, correlation with Bad dream recall (r=−.265, p=.055). There were no correlations with Digit Span; Block Design and Digit Span were also not correlated (r=.067, p=.635).
Spearman correlations between dream recall measures and Wechsler Adult Intelligence Scales (Block design and Digit Span)
| Dream recalla | Bad dream recalla | Nightmare recalla | #Words/reportb | ||
|---|---|---|---|---|---|
| Block design | r | −.137 | −.265 | −.112 | .282 |
| p | .329 | .055 | .426 | .061 | |
| Digit span | r | .144 | −.075 | .053 | .136 |
| p | .305 | .595 | .705 | .373 |
n=53; bn=45 due to lack of dream recall from REM awakening.
Correlations between all-night fast spindle densities and Block Design and Digit Span were consistently null (Block Design: all p > .265; Digit span: all p > .210). Similarly, when calculated by sleep cycle, there were no significant correlations for Block Design (all p > .154) and only two significant correlations for Digit Span, both in cycle 4 (C3, r=.319, p=.027; C4, r=.294, p=.043). Neither survives a correction for multiple analyses.
Discussion
Overall, these results confirm our general expectation of a relationship between measures of dream recall and fast N2 sleep spindles. On three retrospective and one prospective measures of dream recall frequency, we observed positive correlations between levels of recall and the density of fast spindles. The correlations were somewhat stronger for the Bad dream recall measure in that only this measure withstood partialing out of the slow spindle density measures for most derivations, but positive correlations were nonetheless obtained for all recall measures. The correlations were also independent of depression and trait anxiety scores. Furthermore, the correlations were apparent for spindles assessed at different points in the sleep period – but especially in cycles 2 and 3 for Bad dream recall but in all cycles except 2 for #Words/report. While there was some evidence that retrospective dream recall measures were related to learning (two of the three tasks), there was no relationship between task improvement and spindle density scores.
A possible trait factor common to spindles and dreaming
While there might seem to be no reason to expect an association between sleep spindles, a predominantly NREM sleep phenomenon, and dreaming, a predominantly REM sleep phenomenon, there is ample evidence of the interdependence of NREM and REM sleep processes, as suggested by sequential (Giuditta, 2014; Giuditta et al., 1995), “two-stage” (Buzsaki, 1989; Fogel et al., 2009), and “sleep organization” (Ficca and Salzarulo, 2004) models of memory consolidation. As described in the Introduction section, such interrelatedness is suggested by the fact that measures of cognitive abilities that are correlated with sleep spindles are also correlated with REM sleep measures – and in some cases with dreaming measures as well. For example, Auditory Verbal Learning Test scores are correlated with REM sleep time as well as with NREM sleep spindles (Lafortune et al., 2014). Similarly, IQ is associated with both REM sleep eye movements (Smith, Nixon, and Nader, 2004) and sleep spindle density (Fogel et al., 2007), whereas intelligence is also correlated with some measures of dreaming (Fogel et al., 2016; Foulkes, 1999; Sándor, Szakadát, and Bódizs, 2016). In light of such findings, the most parsimonious explanation for the present results may be that sleep spindles and dream recall are related by virtue of a common cognitive trait. While the precise nature of this trait remains unknown for the present, several possibilities might be considered. The trait of general intelligence is a salient candidate, although we found no correlations between spindle density and the Block Design or Digit Span intelligence subtests. However, these two subtests tap a relatively restricted range of cognitive abilities (visuospatial competence and working numeric memory, respectively) and more thorough testing may well find links with general intelligence. Other possible trait factors include an individual’s ability to either generate visuospatial imagery or to accurately recall visuospatial images and memories (e.g., dreams). The latter abilities may be reflected in our #Words/report measure, which could conceivably reflect the typical salience of visuospatial imagery or an ability to describe it in a detailed narrative form. Further study could readily clarify this issue.
A possible sleep-dependent consolidation mechanism common to spindles and dreaming
The present findings are also consistent with the possibility that sleep spindles and dream recall are affected by a shared mechanism for sleep-related memory consolidation that spans both NREM and REM sleep, rather than – or perhaps in addition to – a trait factor explanation. This possibility is supported by several findings. First, spindle density was reliably correlated with the #Words/report variable, which is widely viewed as a measure of local REM sleep memory or arousal processes (Palagini, Gemignani, Feinberg, Guazzelli, and Campbell, 2004; Rosenlicht, Maloney, and Feinberg, 1994). Furthermore, when considered by sleep cycle, the correlations between #Words/report and spindles were largely different from those between Bad dream recall and spindles, suggesting a different mechanism. The former correlations were more pervasive for cycles 1, 4, and 5, whereas the latter correlations were more pervasive for cycles 2 and 3. Similarly, when considered by derivation, the two types of correlations differed: multiple regressions showed spindle correlations with #Words/report to be strongest for F3 and F4, whereas correlations with Bad dream recall were predominant for C3, O1, and O2. Finally, elevated dream recall was found to be associated with better improvement on two of the three consolidation tasks; high Bad dream recall in particular was associated with improvement on MTT and CBT tasks. However, this possible mechanistic explanation is not entirely consistent with either the finding that the #Words/report measure was not associated with task improvements or the finding that spindle densities were uncorrelated with such improvements.
Nonetheless, the findings taken together may belie a more complex interaction between trait and sleep state factors. Such an interaction was demonstrated in a recent memory consolidation study in which participants high on a general cognitive trait factor (“baseline memory performance”) had higher spindle densities but also demonstrated the strongest sleep-dependent memory consolidation effects (Wislowska, Heib, Griessenberger, Hoedlmoser, and Schabus, 2016). The fact that our multiple regression analyses found at least one derivation (C4) for which spindle density independently predicted both the trait (Bad dream recall) and sleep state (#Words/report) measures supports such an interaction between trait and sleep-related mechanisms.
The possibility that sleep spindles are associated with dreaming due to a shared sleep-dependent mechanistic factor is also broadly consistent with sequential (Giuditta, 2014; Giuditta et al., 1995), two-stage (Buzsaki, 1989; Fogel et al., 2009), and sleep organization (Ficca and Salzarulo, 2004) models of NREM–REM sleep states, and has parallels in previous research. First, it is consistent with the finding that overnight improvements in visual discrimination are linked most strongly to a combination of NREM and REM sleep stages, that is, early-night NREM sleep and late-night REM sleep percentages, rather than to either measure alone (Stickgold, Whidbee, Schirmer, Patel, and Hobson, 2000). Second, it is consistent with the finding that targeted memory activation during NREM sleep leads to functional magnetic resonance imaging-detected brain changes in both NREM and REM sleep (Cousins, El-Deredy, Parkes, Hennies, and Lewis, 2016). The latter study found that targeted stimulation in NREM sleep produced REM sleep-related increases in bilateral superior parietal cortex, cerebellum, premotor cortex, sensorimotor cortex, and left dorsolateral prefrontal cortex, as well as changes in NREM sleep. Third, the notion that spindles and dreaming share a common mechanism is consistent with clinical observations; patients with auto-activation deficit are deficient in sleep spindles and report infrequent dreams and dreams that are short, simple, and devoid of normal bizarre and emotional elements, even though measures of sleep duration, continuity, and stage percentages do not differentiate patients from healthy controls (Leu-Semenescu et al., 2013). Finally, some neuroanatomical evidence points to brain regions that are implicated in both spindles and dreaming. For example, whereas spindles in one study were accompanied by activations in thalamus, anterior cingulate cortex, insular cortex, and superior temporal gyrus (Schabus et al., 2007), a review of imaging studies (Hobson, Stickgold, and Pace-Schott, 1998) found activation in the same regions during REM sleep, when dreaming is most abundant. Similarly, while fast spindles are associated with increased gray matter volume in bilateral hippocampus (Saletin, van der Helm, and Walker, 2013), hippocampal volume is also associated with increased dream bizarreness (De Gennaro et al., 2011) and REM sleep hippocampal connectivity is associated with both successful dream recall (Fell et al., 2006) and visual dream vividness (De Gennaro et al., 2011).
Together, these findings illustrate that consolidation mechanisms, such as those underlying both sleep spindles and dreaming, may well span NREM and REM sleep stages. There may be a neural substrate for memory consolidation that implicates both sleep stages and, within these stages, both spindles and dream imagery.
Spindles, dreaming, and memory replay
A consolidation mechanism that may link sleep spindles directly to dreaming is that of memory replays that manifest in different forms in NREM and REM sleep. According to one influential model (Born, 2010; Ritter et al., 2015; Siapas and Wilson, 1998), NREM replay events, as indexed by hippocampal sharp-wave ripples and sleep spindles, are synchronized by a third wave form: neocortical slow oscillations (0.5–1.0 Hz). Slow oscillations cause sharp-wave ripples in the hippocampal CA1 to occur in close temporal proximity to spindles recorded in neocortical regions such as medial prefrontal cortex (Peyrache, Battaglia, and Destexhe, 2011; Siapas and Wilson, 1998). This produces spindle–ripple events (Siapas and Wilson, 1998) that facilitate information transfer between the hippocampal and neocortical cell assemblies (Sirota, Csicsvari, Buhl, and Buzsaki, 2003). Critically, during sleep spindles, the cortex is “functionally deafferented” from hippocampal inputs, presumably through the strong recruitment of inhibitory interneurons (Peyrache et al., 2011). There is anatomical evidence supporting such a view (Varela, Kumar, Yang, and Wilson, 2014) as well as evidence of the expected directionality of activity from hippocampus to prefrontal cortex (Wierzynski, Lubenov, Gu, and Siapas, 2009).
Although this specific mechanism may be suppressed during REM sleep (Wierzynski et al., 2009), when in humans dreaming is more abundant, neural replay events also do take place at this time (Hennevin et al., 1995; Louie and Wilson, 2001; Poe et al., 2000, 2010). Furthermore, it is generally thought that the cortex is functionally deafferented from the hippocampus during REM sleep and that this disengagement supports consolidation (Diekelmann and Born, 2010). Thus, as inhibitory processes, NREM sleep spindles and tonic REM sleep may function similarly to enable replay-related consolidation. However, it remains to be shown more precisely how the neural bases of NREM sleep replay events differ from those of REM sleep and how these may implicate dream formation processes.
A privileged role for bad dreams?
The fact that it was Bad dream recall, rather than either Dream recall or Nightmare recall, that most reliably correlated with both spindles and task improvement in this study requires further clarification. If bad dreams are a sign of pathology, it may seem counterintuitive for them to be associated with neural indicators of adaptive memory consolidation. However, the negative correlations we observed between Bad dream recall and STAI and BDI scores are opposite to what is typically reported in studies of nightmare populations (Nielsen and Levin, 2007). This may be because our laboratory volunteers were not clinical patients and were not seeking treatment; thus, their associated suffering may have been low, despite the high frequency of their nightmares. In other words, the participants with bad dreams and nightmares in this study may have been high functioning and resilient to comorbidities to different degrees. Another possibility is that bad dreams may, in fact, be an adaptive form of emotional dreaming. That bad dreams represent dysphoric emotions without sleep being disturbed by awakenings, which is the definition of bad dreams, might point to these dreams having successfully consolidated a dysphoric emotion by an ostensible adaptive function. Dreaming has been theorized to modify emotion and memory through various mechanisms (for review, see Nielsen and Carr, 2016), including by modulating fear extinction (Nielsen and Levin, 2007) and affect depotentiation (Walker and van der Helm, 2009). From this perspective, our finding that Dream recall had large and significant negative correlations with both depression (r=−.431, p=.001) and anxiety (r=−.360, p=.009) is consistent with the possibility that normal dreaming helps down-regulate negative effect. At the other extreme, however, that frequent Nightmare recall was non-significantly correlated in the same negative direction with the two measures (r=−.165, p=.243 and r=−.149, p=.293) suggests that nightmares may have little or no regulatory effect on negative emotions. Finally, our findings show that correlations for Bad dream recall were also significantly negative, but smaller than those for Dream recall (r=−.362, p=.008 and r=−.284, p=.041), suggesting that bad dreams may be more like normal dreams than they are like nightmares in having a somewhat smaller regulatory effect on negative emotions. Nonetheless, in such a scenario, it remains unclear why Nightmare recall and Bad dream recall were so highly correlated (r=.672, p < .00001) and why Nightmare recall, like Bad dream recall, also correlated positively with sleep spindles in the present analyses. It also remains unclear why Bad dream recall was negatively associated with Block Design scores, while Dream recall was positively correlated, even though both measures were positively associated with sleep spindles. More comparative study of these measures is clearly required.
Authors’ contribution
TN and TP designed the experiment, analyzed the data, and wrote the manuscript; MC, CB-C, and TP conducted the spindle analyses, analyzed the data, and edited the manuscript; all other authors conducted PSG recordings, completed neurocognitive testing, performed artifact rejection, collected dream reports, and entered/checked the data.
Conflict of interest
The authors declare no conflict of interest.
Ethics
This study was approved by Research Ethics Committee of the Hôpital du Sacré-Coeur de Montréal (#2012-749).
Acknowledgements
This study was financially supported by Canadian Institutes of Health Research (CIHR; Nielsen) and Natural Sciences and Engineering Research Council of Canada grant (NSERC; Nielsen).
References
Andrillon T , Nir Y , Staba RJ , Ferrarelli F , Cirelli C , Tononi G , and Fried I (2011). Sleep spindles in humans: insights from intracranial EEG and unit recordings. J Neurosci 31(49):17821–17834. doi:10.1523/JNEUROSCI.2604-11.2011
Arkin AM and Antrobus JS (1991). The effects of external stimuli applied prior to and during sleep on sleep experience. In SJ Ellman and JS Antrobus (Eds.), The mind in sleep, 2nd edition, (pp. 265–307). New York, NY: John Wiley & Sons (Reprinted from: In File).
Astori S , Wimmer RD , and Luthi A (2013). Manipulating sleep spindles – expanding views on sleep, memory, and disease. Trends Neurosci 36(12):738–748. doi:10.1016/j.tins.2013.10.001
Bodizs R , Gombos F , Ujma PP , and Kovacs I (2014). Sleep spindling and fluid intelligence across adolescent development: sex matters. Front Hum Neurosci 8:952. doi:10.3389/fnhum.2014.00952
Bódizs R , Kis T , Lazar AS , Havran L , Rigo P , Clemens Z , and Halasz P (2005). Prediction of general mental ability based on neural oscillation measures of sleep. J Sleep Res 14(3):285–292. doi:10.1111/j.1365-2869.2005.00472.x
Bódizs R , Lazar AS , and Rigo P (2008). Correlation of visuospatial memory ability with right parietal EEG spindling during sleep. Acta Physiol Hung 95(3):297–306. doi:10.1556/APhysiol.95.2008.3.5
Born J (2010). Slow-wave sleep and the consolidation of long-term memory. World J Biol Psychiatry 11(Suppl 1):16–21. doi:10.3109/15622971003637637
Breger L (1969). Dream function: an information processing model. In L. Breger (Ed.), Clinical-cognitive psychology, (pp. 182–227). Englewood Cliffs, NJ: Prentice-Hall.
Butler SF and Watson R (1985). Individual differences in memory for dreams: the role of cognitive skills. Percept Mot Skills 61(3):823–828. doi:10.2466/pms.1985.61.3.823
Buzsaki G (1989). Two-stage model of memory trace formation: a role for “noisy” brain states. Neuroscience 31(3):551–570. doi:10.1016/0306-4522(89)90423-5
Cipolli C , Bolzani R , Tuozzi G , and Fagioli I (2001). Active processing of declarative knowledge during REM-sleep dreaming. J Sleep Res 10(4):277–284. doi:10.1046/j.1365-2869.2001.00268.x
Cipolli C , Fagioli I , Mazzetti M , and Tuozzi G (2004). Incorporation of presleep stimuli into dream contents: evidence for a consolidation effect on declarative knowledge during REM sleep? J Sleep Res 13(4):317–326. doi:10.1111/j.1365-2869.2004.00420.x
Clemens Z , Fabo D , and Halasz P (2005). Overnight verbal memory retention correlates with the number of sleep spindles. Neuroscience 132(2):529–535. doi:10.1016/j.neuroscience.2005.01.011
Clemens Z , Fabo D , and Halasz P (2006). Twenty-four hours retention of visuospatial memory correlates with the number of parietal sleep spindles. Neurosci Lett 403(1–2):52–56. doi:10.1016/j.neulet.2006.04.035
Cory TL , Ormiston DW , Simmel E , and Dainoff M (1975). Predicting the frequency of dream recall. J Abnorm Psychol 84(3):261–266. doi:10.1037/h0076653
Cousins JN , El-Deredy W , Parkes LM , Hennies N , and Lewis PA (2016). Cued reactivation of motor learning during sleep leads to overnight changes in functional brain activity and connectivity. PLoS Biol 14:e1002451. doi:10.1371/journal.pbio.1002451
Dewan EM (1970). The programing (P) hypothesis for REM sleep. Int Psychiatry Clin 7:295–307.
De Gennaro L , Cipolli C , Cherubini A , Assogna F , Cacciari C , Marzano C , Curcio G , Ferrara M , Caltagirone C , and Spalletta G (2011). Amygdala and hippocampus volumetry and diffusivity in relation to dreaming. Hum Brain Mapp 32:1458–1470. doi:10.1002/hbm.21120
De Koninck J , Christ G , Hebert G , and Rinfret N (1990). Language learning efficiency, dreams and REM sleep. Psychiatr J Univ Ott 15:91–92.
De Koninck J , Christ G , Rinfret N , and Proulx G (1988). Dreams during language learning: when and how is the new language integrated? Psychiatr J Univ Ott 13:72–74.
De Koninck J , Prévost F , and Lortie-Lussier M (1996). Vertical inversion of the visual field and REM sleep mentation. J Sleep Res 5(1):16–20. doi:10.1046/j.1365-2869.1996.00001.x
Diba K and Buzsaki G (2007). Forward and reverse hippocampal place-cell sequences during ripples. Nat Neurosci 10:1241–1242. doi:10.1038/nn1961
Diekelmann S and Born J (2010). The memory function of sleep. Nat Rev Neurosci 11:114–126. doi:10.1038/nrn2762
Dumel G , Carr M , Marquis LP , Blanchette-Carriere C , Paquette T , and Nielsen T (2015). Infrequent dream recall associated with low performance but high overnight improvement on mirror-tracing. J Sleep Res 24:372–382. doi:10.1111/jsr.12286
Eichenlaub JB , Bertrand O , Morlet D , and Ruby P (2013). Brain reactivity differentiates subjects with high and low dream recall frequencies during both sleep and wakefulness. Cereb Cortex 24(5):1206–1215. doi:10.1093/cercor/bhs388
Erlacher D and Schredl M (2010). Practicing a motor task in a lucid dream enhances subsequent performance: a pilot study. Sport Psychol 24(2):157–167. doi:10.1123/tsp.24.2.157
Euston DR , Tatsuno M , and McNaughton BL (2007). Fast-forward playback of recent memory sequences in prefrontal cortex during sleep. Science 318(5853):1147–1150. doi:10.1126/science.1148979
Fell J , Fernandez G , Lutz MT , Kockelmann E , Burr W , Schaller C , Elger CE , and Helmstaedter C (2006). Rhinal-hippocampal connectivity determines memory formation during sleep. Brain 129, 108–114. doi:10.1093/brain/awh647
Ficca G and Salzarulo P (2004). What in sleep is for memory. Sleep Med 5(3):225–230. doi:10.1016/j.sleep.2004.01.018
Fiss H , Kremer E , and Lichtman J (1977). The mnemonic function of dreaming. Sleep Res 6:122.
Fogel SM , Nader R , Cote KA , and Smith CT (2007). Sleep spindles and learning potential. Behav Neurosci 121(1):1–10. doi:10.1037/0735-7044.121.1.1
Fogel SM , Ray LB , Sergeeva V , and Owen AM (2016, September 13–16). Smart dreams: do smart people dream smarter? 23rd Congress of the European Sleep Research Society, Bologna, Italy.
Fogel SM and Smith CT (2011). The function of the sleep spindle: a physiological index of intelligence and a mechanism for sleep-dependent memory consolidation. Neurosci Biobehav Rev 35(5):1154–1165. doi:10.1016/j.neubiorev.2010.12.003
Fogel SM , Smith CT , and Beninger RJ (2009). Evidence for 2-stage models of sleep and memory: learning-dependent changes in spindles and theta in rats. Brain Res Bull 79(6):445–451. doi:10.1016/j.brainresbull.2009.03.002
Fosse MJ , Fosse R , Hobson JA , and Stickgold RJ (2003). Dreaming and episodic memory: a functional dissociation? J Cogn Neurosci 15(1):1–9. doi:10.0.1162/089892903321107774
Foulkes D (1985). Dreaming: A cognitive-psychological analysis. Hillsdale, NJ: Lawrence Erlbaum Associates.
Foulkes D (1999). Children’s dreaming and the development of consciousness. Cambridge: Harvard University Press.
Giuditta A (2014). Sleep memory processing: the sequential hypothesis. Front Syst Neurosci 8:219. doi:10.3389/fnsys.2014.00219
Giuditta A , Ambrosini MV , Montagnese P , Mandile P , Cotugno M , Grassi ZG , and Vescia S (1995). The sequential hypothesis of the function of sleep. Behav Brain Res 69(1–2):157–166. doi:10.1016/0166-4328(95)00012-I
Hampshire A , Highfield RR , Parkin BL , and Owen AM (2012). Fractionating human intelligence. Neuron 76:1225–1237. doi:10.1016/j.neuron.2012.06.022
Hennevin E , Hars B , Maho C , and Bloch V (1995). Processing of learned information in paradoxical sleep: relevance for memory. Behav Brain Res 69(1–2):125–135. doi:10.1016/0166-4328(95)00013-J
Hobson JA , Stickgold R , and Pace-Schott EF (1998). The neuropsychology of REM sleep dreaming. Neuroreport 9(3):R1–R14. doi:10.1097/00001756-199802160-00033
Horton CL and Malinowski JE (2015). Autobiographical memory and hyperassociativity in the dreaming brain: implications for memory consolidation in sleep. Front Psychol 6:874. doi:10.3389/fpsyg.2015.00874
Kaestner EJ , Wixted JT , and Mednick SC (2013). Pharmacologically increasing sleep spindles enhances recognition for negative and high-arousal memories. J Cogn Neurosci 25:1597–1610. doi:10.1162/jocn_a_00433
Kudrimoti HS , Barnes CA , and McNaughton BL (1999). Reactivation of hippocampal cell assemblies: effects of behavioral state, experience, and EEG dynamics. J Neurosci 19:4090–4101.
Lafortune M , Gagnon JF , Martin N , Latreille V , Dube J , Bouchard M , Bastien C , and Carrier J (2014). Sleep spindles and rapid eye movement sleep as predictors of next morning cognitive performance in healthy middle-aged and older participants. J Sleep Res 23(2):159–167. doi:10.1111/jsr.12108
Leu-Semenescu S , Uguccioni G , Golmard JL , Czernecki V , Yelnik J , Dubois B , Forgeot d’Arc B , Grabli D , Levy R , and Arnulf I (2013). Can we still dream when the mind is blank? Sleep and dream mentations in auto-activation deficit. Brain 136(10):3076–3084. doi:10.1093/brain/awt229
Louie K and Wilson MA (2001). Temporally structured replay of awake hippocampal ensemble activity during rapid eye movement sleep. Neuron 29(1):145–156. doi:10.1016/S0896-6273(01)00186-6
Lusignan FA , Zadra A , Dubuc MJ , Daoust AM , Mottard JP , and Godbout R (2009). Dream content in chronically-treated persons with schizophrenia. Schizophr Res 112(1–3):164–173. doi:10.1016/j.schres.2009.03.032
Lustenberger C , Boyle MR , Alagapan S , Mellin JM , Vaughn BV , and Frohlich F (2016). Feedback-controlled transcranial alternating current stimulation reveals a functional role of sleep spindles in motor memory consolidation. Curr Biol 26:2127–2136. doi:10.1016/j.cub.2016.06.044
Mahoney JM , Titiz AS , Hernan AE , and Scott RC (2016). Short-Range temporal interactions in sleep; hippocampal spike avalanches support a large milieu of sequential activity including replay. PLoS One 11:e0147708. doi:10.1371/journal.pone.0147708
Malinowski JE and Horton CL (2014). Memory sources of dreams: the incorporation of autobiographical rather than episodic experiences. J Sleep Res 23:441–447. doi:10.1111/jsr.12134
Martin N , Lafortune M , Godbout J , Barakat M , Robillard R , Poirier G , Bastien C , and Carrier J (2013). Topography of age-related changes in sleep spindles. Neurobiol Aging 34(2):468–476. doi:10.1016/j.neurobiolaging.2012.05.020
Nguyen ND , Tucker MA , Stickgold R , and Wamsley EJ (2013). Overnight sleep enhances hippocampus-dependent aspects of spatial memory. Sleep 36, 1051–1057. doi:10.5665/sleep.2808
Nielsen T and Carr M (2016). Nightmares and nightmare function. In M Kryger, B Roth, and W Dement (Eds.), Principles and practice of sleep medicine, (Vol. 6, pp. 546–554). New York, NY: Elsevier.
Nielsen T and Levin R (2007). Nightmares: a new neurocognitive model. Sleep Med Rev 11(4):295–310. doi:10.1016/j.smrv.2007.03.004
Nielsen T , O’Reilly C , Carr M , Dumel G , Godin I , Solomonova E , Lara-Carrasco J , Blanchette-Carriere C , and Paquette T (2015). Overnight improvements in two REM sleep-sensitive tasks are associated with both REM and NREM sleep changes, sleep spindle features, and awakenings for dream recall. Neurobiol Learn Mem 122:88–97. doi:10.1016/j.nlm.2014.09.007
Nielsen TA , Paquette T , Solomonova E , Lara-Carrasco J , Popova A , and Levrier K (2010). REM sleep characteristics of nightmare sufferers before and after REM sleep deprivation. Sleep Med 11(2):172–179. doi:10.1016/j.sleep.2008.12.018
O’Reilly C , Godbout J , Carrier J , and Lina JM (2015). Combining time-frequency and spatial information for the detection of sleep spindles. Front Hum Neurosci 9:70. doi:10.3389/fnhum.2015.00070
O’Reilly C and Nielsen T (2014). Assessing EEG sleep spindle propagation. Part 2: experimental characterization. J Neurosci Methods 221, 215–227. doi:10.1016/j.jneumeth.2013.08.014
Palagini L , Gemignani A , Feinberg I , Guazzelli M , and Campbell IG (2004). Mental activity after early afternoon nap awakenings in healthy subjects. Brain Res Bull 63(5):361–368. doi:10.1016/j.brainresbull.2003.12.008
Palombo SR (1976). The dream and the memory cycle. Int Rev Psychoanal 3:65–83.
Pantoja AL , Faber J , Rocha LH , Ferro D , Silvestre-Souca RC , Dias G , Araujo JF , Nicolelis MA , and Ribeiro S (2009). Assessment of the adaptive value of dreams. Sleep 32:A421.
Peyrache A , Battaglia FP , and Destexhe A (2011). Inhibition recruitment in prefrontal cortex during sleep spindles and gating of hippocampal inputs. Proc Natl Acad Sci U S A 108(41):17207–17212. doi:10.1073/pnas.1103612108
Plailly J , Villalba M , Nicolas A , and Ruby P (2016, September 13–16). Recalling a dream related to a recent experience: does it help episodic memory consolidation? 23rd Congress of the European Sleep Research Society, Bologna, Italy.
Poe GR , Nitz DA , McNaughton BL , and Barnes CA (2000). Experience-dependent phase-reversal of hippocampal neuron firing during REM sleep. Brain Res 855(1):176–180. doi:10.1016/S0006-8993(99)02310-0
Poe GR , Walsh CM , and Bjorness TE (2010). Cognitive neuroscience of sleep. Prog Brain Res 185:1–19. doi:10.1016/B978-0-444-53702-7.00001-4
Rasch B and Born J (2013). About sleep’s role in memory. Physiol Rev 93(2):681–766. doi:10.1152/physrev.00032.2012
Revonsuo A (2000). The reinterpretation of dreams: an evolutionary hypothesis of the function of dreaming. Behav Brain Sci 23(6):877–901. doi:10.1017/S0140525X00004015
Ritter P , Born J , Brecht M , Dinse HR , Heinemann U , Pleger B , Schmitz D , Schreiber S , Villringer A , and Kempter R (2015). State-dependencies of learning across brain scales. Front Comput Neurosci 9:1. doi:10.3389/fncom.2015.00001
Rosenlicht N , Maloney T , and Feinberg I (1994). Dream report length is more dependent on arousal level than prior REM duration. Brain Res Bull 34(2):99–101. doi:10.1016/0361-9230(94)90004-3
Sabourin C , Forest G , Hebert G , and De Koninck J (2006). Intensive language learning and episodic memory in dreams. Sleep 29:A48.
Saletin JM , van der Helm E , and Walker MP (2013). Structural brain correlates of human sleep oscillations. Neuroimage 83:658–668. doi:10.1016/j.neuroimage.2013.06.021
Sándor P , Szakadát S , and Bódizs R (2016). The development of cognitive and emotional processing as reflected in children’s dreams: active self in an eventful dream signals better neuropsychological skills. Dreaming 26:58–78. doi:10.1037/drm0000022
Scarpelli S , Marzano C , D’Atri A , Gorgoni M , Ferrara M , and De Gennaro L (2015). State- or trait-like individual differences in dream recall: preliminary findings from a within-subjects study of multiple nap REM sleep awakenings. Front Psychol 6:928. doi:10.3389/fpsyg.2015.00928
Schabus M , Hoedlmoser K , Pecherstorfer T , Anderer P , Gruber G , Parapatics S , Sauter C , Kloesch G , Klimesch W , Saletu B , and Zeitlhofer J (2008). Interindividual sleep spindle differences and their relation to learning-related enhancements. Brain Res 1191:127–135. doi:10.1016/j.brainres.2007.10.106
Schabus M , Dang-Vu TT , Albouy G , Balteau E , Boly M , Carrier J , Darsaud A , Degueldre C , Desseilles M , Gais S , Phillips C , Rauchs G , Schnakers C , Sterpenich V , Vandewalle G , Luxen A , and Maquet P (2007). Hemodynamic cerebral correlates of sleep spindles during human non-rapid eye movement sleep. Proc Natl Acad Sci U S A 104(32):13164–13169. doi:10.1073/pnas.0703084104
Schredl M and Erlacher D (2010). Is sleep-dependent memory consolidation of a visuo-motor task related to dream content? Int J Dream Res 3:74–79. doi:10.11588/ijodr.2010.1.486
Schredl M and Montasser A (1996). Dream recall: state or trait variable? Part I: model, theories, methodology and trait factors. Imag Cogn Pers 16(2):181–210. doi:10.2190/RCAG-NY96-3D99-KA0G
Siapas AG and Wilson MA (1998). Coordinated interactions between hippocampal ripples and cortical spindles during slow-wave sleep. Neuron 21(5):1123–1128. doi:10.1016/S0896-6273(00)80629-7
Silber MH , Ancoli-Israel S , Bonnet MH , Chokroverty S , Grigg-Damberger MM , Hirshkowitz M , Kapen S , Keenan SA , Kryger MH , Penzel T , Pressman MR , and Iber C (2007). The visual scoring of sleep in adults. J Clinical Sleep Med 3:121–131.
Sirota A , Csicsvari J , Buhl D , and Buzsaki G (2003). Communication between neocortex and hippocampus during sleep in rodents. Proc Natl Acad Sci U S A 100(4):2065–2069. doi:10.1073/pnas.0437938100
Smith C (2010). Sleep states, memory processing, and dreams. Sleep Med Clin 5(2):217–228. doi:10.1016/j.jsmc.2010.01.002
Smith CT , Nixon MR , and Nader RS (2004). Posttraining increases in REM sleep intensity implicate REM sleep in memory processing and provide a biological marker of learning potential. Learn Mem 11(6):714–719. doi:10.1101/lm.74904
Stickgold R , Whidbee D , Schirmer B , Patel V , and Hobson JA (2000). Visual discrimination task improvement: a multi-step process occurring during sleep. J Cogn Neurosci 12(2):246–254. doi:10.1162/089892900562075
Tamminen J , Lambon Ralph MA , and Lewis PA (2013). The role of sleep spindles and slow-wave activity in integrating new information in semantic memory. J Neurosci 33(39):15376–15381. doi:10.1523/JNEUROSCI.5093-12.2013
Tamminen J , Payne JD , Stickgold R , Wamsley EJ , and Gaskell MG (2010). Sleep spindle activity is associated with the integration of new memories and existing knowledge. J Neurosci 30(43):14356–14360. doi:10.1523/JNEUROSCI.3028-10.2010
Varela C , Kumar S , Yang JY , and Wilson MA (2014). Anatomical substrates for direct interactions between hippocampus, medial prefrontal cortex, and the thalamic nucleus reuniens. Brain Structure Function 219:911–929. doi:10.1007/s00429-013-0543-5
Walker MP and van der Helm E (2009). Overnight therapy? The role of sleep in emotional brain processing. Psychol Bull 135(5):731–748. doi:10.1037/a0016570
Wallant DC , Maquet P , and Phillips C (2016). Sleep spindles as an electrographic element: description and automatic detection methods. Neural Plast 2016:6783812. doi:10.1155/2016/6783812
Wamsley EJ (2014). Dreaming and offline memory consolidation. Curr Neurol Neurosci Rep 14(3):433. doi:10.1007/s11910-013-0433-5
Wamsley EJ , Nguyen N , Tucker M , Olsen A , and Stickgold R (2012). EEG correlates of overnight memory consolidation in a virtual navigation task. Sleep 35:A86.
Wamsley EJ , Perry K , Djonlagic I , Reaven LB , and Stickgold R (2010). Cognitive replay of visuomotor learning at sleep onset: temporal dynamics and relationship to task performance. Sleep 33:59–68.
Wamsley EJ and Stickgold R (2011). Memory, sleep and dreaming: experiencing consolidation. Sleep Med Clinics 6(1):97–108. doi:10.1016/j.jsmc.2010.12.008
Wamsley EJ , Tucker M , Payne JD , Benavides JA , and Stickgold R (2010). Dreaming of a learning task is associated with enhanced sleep-dependent memory consolidation. Curr Biol 20(9):850–855. doi:10.1016/j.cub.2010.03.027
Wamsley EJ , Tucker MA , Shinn AK , Ono KE , McKinley SK , Ely AV , Goff DC , Stickgold R , and Manoach DS (2012). Reduced sleep spindles and spindle coherence in schizophrenia: mechanisms of impaired memory consolidation? Biol Psychiatry 71(2):154–161. doi:10.1016/j.biopsych.2011.08.008
Wierzynski CM , Lubenov EV , Gu M , and Siapas AG (2009). State-dependent spike-timing relationships between hippocampal and prefrontal circuits during sleep. Neuron 61(4):587–596. doi:10.1016/j.neuron.2009.01.011
Wislowska M , Heib DPJ , Griessenberger H , Hoedlmoser K , and Schabus M (2016). Individual baseline memory performance and its significance for sleep-dependent memory consolidation. Sleep Spindles Cortical Up States, 1–12. doi:10.1556/2053.1.2016.001