Edited by: Jennifer M. Windt, Johannes Gutenberg-University of Mainz, Germany
Reviewed by: Valdas Noreika, Medical Research Council, UK; Francesca Siclari, University of Wisconsin, USA
*Correspondence: Perrine Ruby, INSERM U1028, Centre Hospitalier Le Vinatier (Bât. 452), 95, Boulevard Pinel, 69675 Bron Cedex, France e-mail:
This article was submitted to Frontiers in Consciousness Research, a specialty of Frontiers in Psychology.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Studies in cognitive psychology showed that personality (openness to experience, thin boundaries, absorption), creativity, nocturnal awakenings, and attitude toward dreams are significantly related to dream recall frequency (DRF). These results suggest the possibility of neurophysiological trait differences between subjects with high and low DRF. To test this hypothesis we compared sleep characteristics and alpha reactivity to sounds in subjects with high and low DRF using polysomnographic recordings and electroencephalography (EEG). We acquired EEG from 21 channels in 36 healthy subjects while they were presented with a passive auditory oddball paradigm (frequent standard tones, rare deviant tones and very rare first names) during wakefulness and sleep (intensity, 50 dB above the subject's hearing level). Subjects were selected as High-recallers (HR, DRF = 4.42 ± 0.25 SEM, dream recalls per week) and Low-recallers (LR, DRF = 0.25 ± 0.02) using a questionnaire and an interview on sleep and dream habits. Despite the disturbing setup, the subjects' quality of sleep was generally preserved. First names induced a more sustained decrease in alpha activity in HR than in LR at Pz (1000–1200 ms) during wakefulness, but no group difference was found in REM sleep. The current dominant hypothesis proposes that alpha rhythms would be involved in the active inhibition of the brain regions not involved in the ongoing brain operation. According to this hypothesis, a more sustained alpha decrease in HR would reflect a longer release of inhibition, suggesting a deeper processing of complex sounds than in LR during wakefulness. A possibility to explain the absence of group difference during sleep is that increase in alpha power in HR may have resulted in awakenings. Our results support this hypothesis since HR experienced more intra sleep wakefulness than LR (30 ± 4 vs. 14 ± 4 min). As a whole our results support the hypothesis of neurophysiological trait differences in high and low-recallers.
Dreaming remains one of the great mysteries of human cognition. Just after waking up in the morning nearly everyone has experienced some bizarre representations from the night before. They are sometimes insightful and can result in great discoveries at the scientific level (like the periodic table or the ring-like structure of benzene, Maquet and Ruby,
However, despite recent advances (Voss et al.,
In the 1950s, rapid eye movement sleep (REM sleep) was considered as the neurophysiological state underlying dreaming (Aserinsky and Kleitman,
Interestingly, some researchers did investigate dreaming in several sleep stages. They used scalp electroencephalogram (EEG) and classified subject's awakenings (from REM and NREM sleep) according to the subsequent report (or not) of a dream. Then, they analyzed the EEG power in various frequency bands during both REM and NREM sleep in the few minutes preceding a dream report versus no dream report. They found that the sleep EEG preceding a dream report differed from the sleep EEG preceding no dream report (Takeuchi et al.,
Other researchers exploited one of the only current means to obtain real time information about dreaming i.e., lucid dreaming. Lucid dreamers can be aware that they are dreaming while asleep and sometimes they can also be in control of some part/elements of the dreamed story. In that case a lucid dreamer can indicate to an experimenter that he is dreaming, for example by moving his eyes according to a particular sequence (a code) determined previously. Using this strategy researchers could contrast REM with lucid dreaming to REM without lucid dreaming and investigate the cerebral correlates of lucidity (Voss et al.,
Finally, in light of the literature, it appears that no convincing paradigms are available to investigate the brain activity associated with dreaming in healthy subjects. The main reason which explains this lack and the paucity of experimental results about the cerebral correlates of dreaming is the difficulty in knowing when dreaming occurs in the subjects sleep cycle. Indeed, in this context it is not possible to scan the brain while dreaming versus while not dreaming.
Studies in cognitive psychology showed that personality (openness to experience, thin boundaries, absorption), creativity, nocturnal awakenings, and attitude toward dreams were significantly related to dream recall frequency (DRF) (Schredl et al.,
In a first study, we showed that event related potentials (ERPs) elicited by first names, presented rarely and randomly among pure tones (Eichenlaub et al.,
In order to pursue the investigation of the neurophysiological differences between HR and LR we re-analysed the data of this EEG study (Eichenlaub et al.,
In the sleep literature, the presence of alpha rhythm in the EEG is one of the criteria signaling the wakefulness state (Rechtschaffen and Kales,
In the wakefulness literature, ongoing alpha rhythms were first discovered by Berger (
In the visual modality, decreases in the power of alpha rhythms (also called event-related desynchronization) were observed over regions involved in the task realized by the subject. By contrast increases were found over regions irrelevant to the task (Klimesch et al.,
In the auditory modality, few EEG studies investigated alpha oscillations. Taken together, they brought evidence that sounds induce an alpha decrease with a parietal topography (Yordanova et al.,
Therefore, during wakefulness, alpha rhythms seem to play an important inhibitory role involved in the selection of stimuli to be processed and brain regions to be activated (or not inhibited) (Klimesch et al.,
Thus, in light of the functional role attributed to the alpha band modulation during sleep and wakefulness, and according to our ERP results one may expect that auditory stimuli induce different pattern of alpha activity in HR and LR. However, few studies investigated alpha activity in HR and LR during sleep (Goodenough et al.,
In the present study, we tried to fill in this gap. We investigated alpha oscillatory activity in response to complex sounds (first names) during wakefulness and sleep in HR and LR. We used time-frequency analysis to investigate oscillatory power in the frequency-band 8–12 Hz [we ran an oscillatory analysis on the EEG data previously used for ERPs analysis in Eichenlaub et al. (
Approximately 1000 persons interested in participating in this study filled out a questionnaire concerning sleep and dreaming habits (the subjects were unaware that DRF was a criterion for subject selection). Subsequently, the subjects were contacted by telephone and asked “on average, how many mornings per week do you wake up with a dream in mind?”. A dream was previously defined as a long and bizarre story, an image that vanishes rapidly, or a feeling of having dreamt. Subjects were selected as High-recallers upon confirming dream recall (long stories or images) on more than three mornings per week, and as Low-recallers upon confirming dream recall (long stories, images or even a feeling of having dreamt) on less than two mornings per month. Eighteen High-recallers (mean DRF = 4.42 ± 0.25 SEM, dream recalls per week) and 18 Low-recallers (mean DRF = 0.25 ± 0.02) were selected. The following parameters did not differ between the groups: gender, age, habitual sleep duration, habitual sleep time, education level and the size of the place of residence (Schredl,
The auditory stimuli were spectrally rich tones with a main frequency of 800 Hz and two harmonic partials (1600 and 3200 Hz), the subject's own first name (OWN) and an unfamiliar first name (OTHER). First names were digitally recorded by a neutral masculine voice using Adobe Audition 1.5 (Adobe software). After recording, maximum amplitudes of all stimuli were normalized. The mean durations of OWN (581 ms ± 86) and OTHER (598 ms ± 78) were not significantly different (Eichenlaub et al.,
The presentation of the four types of auditory stimuli obeyed the rules of a novelty oddball paradigm. Tones lasting 75 and 30 ms (including 5 ms rise/fall times) were used respectively as Standards (
Subjects arrived in the lab at 7.00 p.m. after they had eaten. During ~1 h and a half, electrodes were fixed on their head and face. The subjects selected a movie among a choice of comedy or action movies. Then, subjects were installed in an acoustically dampened and electrically shielded room, earphones were inserted in their ears, and their hearing threshold was assessed using standard stimuli. The evening recording session started at 10.24 p.m. ± 45 min (duration, 1 h and 6 ± 9 min). Stimuli (around 120 Novels were presented in average) were presented binaurally at 50 dB above the subject's hearing level using the Presentation software (Neurobehavioral Systems). Subjects were instructed to watch the movie (silenced with subtitles) and to ignore the auditory stimuli (Eichenlaub et al.,
Twenty-one Ag/AgCl scalp electrodes were manually positioned according to the extended International 10–20 System (Fz, Cz, Pz; FP1, F3, FC1, C3, T3, CP1, P3, M1, O1 and their counterparts on the right hemiscalp). This relatively small number of electrodes was both compatible with sleep recording and with the use of scalp potential (SP) maps. We concentrated the electrodes around the sites expected to show between first name effects i.e., central and parietal sites. Contact between skin and electrodes was made using EC2 electrode cream Pactronic (Grass Product Group) and electrodes were fixed on the scalp using the paste TENSIVE (Parker Laboratories, Inc.). The reference electrode was placed on the tip of the nose, and the ground electrode on the forehead. The electro-oculogram (EOG) was recorded from 2 electrodes placed on the supraorbital ridge of the left eye and on the infraorbital ridge of the right eye. Muscle activity (EMG) was recorded from 2 electrodes attached to the chin. Electrode impedance was kept below 5 kΩ. The electrophysiological data (EEG, EOG, and EMG) were continuously recorded via a BrainAmp system (Brain Products GmbH, Germany) with an amplification gain of 12,500, a high-pass filter of 0.1 Hz and a sampling rate of 1000 Hz with an anti-aliasing low-pass filter (Eichenlaub et al.,
Sleep stages were scored off-line, visually by JBE according to standard criteria (Silber et al.,
Analysis was focused on responses to first names (novel stimuli). Oscillatory activities induced by the novel stimuli were characterized in each vigilance stage, separately. Trials were automatically excluded from analysis when the overall electrophysiological signal amplitude exceeded 150 μV during wakefulness and 400 μV in REM sleep, N2 and N3, during the [−500; 1500 ms] time-window. We investigated oscillatory activities by means of wavelet decomposition, which provides a good compromise between time and frequency resolutions. We used complex Gaussian Morlet's wavelets (complex waves with a Gaussian shape in the time- and frequency- domains) with a ratio f/σf = 7, f being the central frequency of the wavelet and σf the standard deviation of the Gaussian envelop in the frequency domain (Tallon-Baudry and Bertrand,
We computed the oscillation power on the [−500; 1500 ms] time-window around first names onset in each group (HR and LR). A baseline correction was applied by subtracting the mean power between −300 and −100 ms before stimuli onset, in each frequency band.
As we were interested in modulations of alpha oscillations, we performed statistical analysis in the frequency-window from 8 to 12 Hz. Moreover, as the Standard after a novel stimulus occurred 1260 ms after its onset we performed statistics on the time-window from 0 to 1200 ms.
Alpha reactivity to the first names was detected in the time–frequency domain at each electrode with successive non-parametric signed-rank Wilcoxon tests comparing the mean power in the 8–12 Hz frequency band over 200 ms windows moving by step of 100 ms (from 0 to 1200 ms) with the pre-stimulus power (mean power over a window between −300 and −100 ms before stimulus onset). These time-parameters are justified by the 220 ms duration of the 10 Hz wavelet. We used the following strategy to correct for multiple comparisons. In the temporal dimension we applied the Bonferroni correction (Abdi,
To compare alpha power induced in high and low recallers, non-parametric Mann–Whitney tests were performed at the electrodes showing a significant alpha reactivity to the first names. Mann–Whitney tests were performed on the mean power in the 8–12 Hz frequency band over 200 ms windows moving by step of 100 ms (from 0 to 1200 ms).
All analyses were performed with the ELAN Pack software developed at the Brain Dynamics and Cognition Team of the Lyon Neuroscience Research Center, Lyon, France (
Despite the uncomfortable nature of the experimental setup, sleep quality was generally preserved. For both groups, all sleep parameters evaluated were in the normal range. The sleep parameters did not differ between HR and LR with the exception of the duration of intra-sleep wakefulness (computed through the number of epochs scored as wakefulness during the sleep period; this measure did not include arousals or micro-arousals lasting less than 15 s). HR demonstrated longer intra-sleep wakefulness than LR (~15 min more on average). The number of awakenings (the number of phases composed of consecutive pages of awakening) was not significantly different between the two groups (HR, 17.5 ± 8.7; LR, 12.1 ± 11.9;
The dream reports obtained immediately after awakening in the morning confirmed a large DRF difference between the two groups. Although the subjects were in most cases awakened during REM sleep, only 33% of Low-recallers reported a dream while 94% of High-recallers did (see Eichenlaub et al.,
The average number of accepted trials (first names) was 127 ± 39 and 130 ± 39 during wakefulness in HR and LR respectively, 183 ± 52 and 214 ± 65 during sleep stage N2, 135 ± 32 and 133 ± 37 during sleep stage N3 and 114 ± 34 and 125 ± 35 during REM sleep.
No results are provided from NREM sleep due to the strong overlap of the spectrum of K complexes, spindles, slow waves with the alpha band (8–12 Hz). As a consequence the genuine alpha reactivity to first names could not be correctly identified.
The present study aimed at better understanding the neurophysiological differences between HR and LR. In 18 High-recallers and 18 Low-recallers, we investigated alpha oscillatory activity (8–12 Hz) in response to first names presented as auditory novel stimuli in a stream of simple tones during sleep and wakefulness. First names were presented at a low frequency (4% of stimuli) in a novelty oddball paradigm while subjects were watching a silent movie with subtitles in the evening and during sleep at night. Our objective was to compare between HR and LR the reactivity of alpha rhythms to novel stimuli during wakefulness and sleep.
According to the dominant hypothesis, alpha rhythms would be involved in the active inhibition of the brain regions not involved in the current brain operations (reviews: Klimesch et al.,
Contrary to our hypothesis we found no group difference in induced alpha power during REM sleep. One possibility to explain this negative result is that increases in alpha power in HR may have resulted in awakenings. Indeed, contrary to wakefulness, in REM sleep, stimulations induce an increase in alpha power and not a decrease. As alpha is a rhythm characteristic of the wake state, a strong increase in alpha power may destabilize REM sleep and induce a transition from sleep to wakefulness. If so, the stimuli showing a great alpha effect have been excluded from the sleep analysis because they pertained to a page showing signs of wakefulness. Our results support this hypothesis since HR experienced more intra sleep wakefulness than LR (30 ± 4 vs. 14 ± 5 min in HR vs. LR on average, see Eichenlaub et al.,
Two interpretations can be put forward to explain that alpha reactivity to stimuli has two opposite directions during wakefulness (decrease) and REM sleep (increase) (data analysis in all subjects,
Cantero et al. (
Yet, the functional role attributed to alpha rhythms in REM sleep by Cantero et al.'s hypothesis is not consistent with their functional role during wakefulness according to the inhibition-timing hypothesis (Klimesch et al.,
The comparison of the brain activity in HR and LR provides argument rather in favor to the hypothesis of an activating role of alpha rhythm during sleep. Indeed, our results showed increased reactivity in HR be it with ERPs during sleep and wakefulness, or with alpha reactivity during wakefulness. In addition intra-sleep wakefulness was longer in HR than in LR. All these results argue in favor of increased reactivity in HR as compared to LR. As a consequence one may interpret the absence of alpha difference between the 2 groups during REM sleep as a false negative. Indeed, the increased reactivity in HR resulted in increased post-stimulus ERPs amplitude. If it also induced a post-stimulus increase in alpha power, such increase may have destabilized sleep and may have led to awakenings. If true, awakening would systematically mask the increase in alpha power in HR since trials followed by awakening are excluded from the analysis of sleep.
Our results (ERPs and alpha) as well as previous ones (Solms,
At the theoretical level, contrasting HR and LR cannot help to resolve the memory/production issue i.e., disentangle between the respective involvement of dream production and dream memorization in the ability to recall dreams. This paradigm may nonetheless help to progress providing some clues.
Our results showing intra-sleep wakefulness differences between HR and LR, strongly suggest that memory processes participate in the DRF difference between groups in the context of the arousal-retrieval model (Koulack and Goodenough,
Contrasting HR and LR with neuroimaging techniques (functional Magnetic Resonance Imaging—fMRI, or PET may also help to progress in the understanding of the relationship between dream production and dream recall. Indeed, identifying the brain regions showing the greatest activity difference between HR and LR during sleep may provide clues. If, for example, the brain regions differentiating the two groups are involved in memory processing (HR showing more activity than LR in the hippocampus during sleep for example), this would be an argument in favor of the hypothesis proposing that “no dream recall” result from no memory of the dream (encoding or recall) rather than from no production of the dream.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
This work was supported by a grant awarded to Dr. Perrine Ruby from the French “Agence Nationale pour la Recherche” (ANR-07-JCJC-0095). This work was performed within the framework of the LABEX CELYA (ANR-10-LABX-0060) of Université de Lyon, within the program “Investissements d'Avenir” (ANR-11-IDEX-0007) operated by the French National Research Agency (ANR). All authors (Ruby, Blochet, Eichenlaub, Bertrand, Morlet, Bidet-Caulet) have no conflicts of interest.