The role of dreams in memory consolidation and reactivations during sleep

Memory is essential to humans throughout their lifespan enabling us to survive
in environments with both changing and static components. A broad range of
literature has shown that sleep is important for memory functions (for a review
see Rasch & Born, 2013). While encoding and retrieval of memories occur during
wake, sleep has been proposed to be the optimal state for memory consolidation
(Kirov et al., 2009). Memory consolidation encompasses both the strengthening
of memory (often measured as performance on a memory task) and the emotional
processing of memory (Walker & van Der Helm, 2009). Evidence is emerging that
the two sleep stages, rapid eye movement sleep (REM) and Non-REM sleep (NREM,
further divided into 3 stages), might each be important for one of these
functions. A broad range of studies has shown the effect of sleep on memory
strength for different types of memory (e.g. Plihal & Born, 1997, 1999) and
different age groups (e.g. Aly & Moscovitch, 2010; Seehagen et al., 2015). Many
studies have shown an association of NREM sleep with overnight changes in
memory strength (e.g. Holz et al., 2012; Tucker et al., 2006). The active
system consolidation hypothesis (Diekelmann & Born, 2010) proposes that sleep
plays an active role in memory consolidation by spontaneous and repeated neural
reactivations(i.e. activations of the same neurons in the same sequence)
particularly in the hippocampus, which have been measured in both rodents and
humans (Hirase et al., 2001; Pavlides & Winson, 1989; Peigneux et al., 2004).
The hypothesis assumes that during sleep, memories are redistributed from the
hippocampus to the neocortex. This redistribution is posited to be orchestrated
by slow waves and spindle-ripple events (Staresina et al., 2015); both features
of NREM sleep. Studies in humans have shown that these reactivations can also
be induced by presenting cues (e.g. sounds) associated with a specific memory
during sleep, so-called targeted memory reactivations (TMR, Rasch et al.,
2007). In rats, it has been shown that these cues lead to reactivation events
of the specific associated memory (Bendor & Wilson, 2012). The function of REM
sleep in memory consolidation is less clear. One proposed function could be the
removal of the associated affective tone of emotional memories (Walker & van
Der Helm, 2009). Repeated reactivations during REM sleep could lead to
depotentiation of emotional valence (i.e. positive or negative) of a memory. On
the other hand, sleep with more REM has been linked to increased valence
ratings on emotional images in the next morning (Lara*Carrasco et al., 2009).
Potentially, REM sleep increases emotionality short-term but decreases
long-term valence of emotional memories. One aspect of sleep has often been
overlooked in these studies: dreaming. This is partially because dream research
relies on the participants* reports of their dreams. This means that we cannot
disentangle a lack of dreams from a lack of remembering dreams and we cannot
study dreams unless they are remembered and reported. While research has
investigated when we dream (McNamara et al., 2010; Siclari et al., 2017), what
we dream about (Griffith et al., 1958; Schredl et al., 2004), the question of
potential functions of dreaming is still unsolved. Studies have shown, that
dreams often reflect recent waking-life experiences (Schredl, 2003), and
learning a game before sleep leads to incorporation of the game into dreams
(Stickgold et al., 2000).
Therefore, dreams could potentially reflect memory consolidation processes. If
that is true, then dreaming of a task should lead to better recall of said
task. However, it would make sense that the dual functions of the different
sleep stages for different aspects of memory consolidation also apply to dreams
from the respective stages. Humans report dreams when awoken from both NREM and
REM sleep, however, dream reports from REM are more frequent, longer, more
emotional and more vivid. This reflects the different neural environments
present during the different sleep stages. A recent article that reviewed 12
published studies on dreaming and memory consolidation has shown discordant
results (Plailly et al., 2019). Only five of the studies showed (at least
partial) associations between dreaming of a task and task performance. The
review included experiments using different types of memory, which rely on
different brain structures and therefore might be influenced by sleep and
dreams in different ways. Additionally, many of the studies included in the
review are underpowered (sample sizes less than 20, tasks with less than 10%
incorporation rate) and only our previous study (Schoch et al., 2019)
differentiated between NREM and REM dreams. Based on the findings of sleep and
memory research, we hypothesize, that only NREM dreams are associated with
memory strength, which is what we found in our previous study, which, however,
only included a small sample size (Schoch et al., 2019). In fact, when dividing
the studies in the review by sleep stage the dream reports are collected from,
a similar pattern appears: when only REM dreams are used positive findings are
rare (1/5 studies), but both studies with only NREM dreams show an association
(2/2). Therefore, it seems plausible that dreams reflect the specific
consolidation processes happening during each sleep stage. Additionally,
supporting this hypothesis is the finding from a study that shows that the
incorporation of stressful events into REM dreams is associated with better
mental health outcomes (Cartwright et al., 2006). TMR presents interesting
possibilities to not only study associations between dream content and memory
performance but also directly manipulate dream content using external cues.
Applying TMR and next to researching spontaneous incorporation will give a much
more detailed and mechanistic understanding of how dreams relate to declarative
memory processes in the night. Furthermore we are exploring how measurements of
autonomic activation (heart rate and stomach contractions) relate to sleep,
dreaming and memory consolidation processes during sleep.

Primary Objective: This study aims to investigate the role of dreams in memory
consolidation during sleep. These are our 2 primary objectives and the
associated hypothesis:

1) To replicate the result of our previous study in large sample size: Only
incorporations in NREM dreams are related to memory strength and to elicit the
role of REM dreams in memory consolidation. Our first research hypothesis is
the following:
H1) NREM and REM dreams reflect different aspects of memory consolidation
pro-cesses during sleep.
A) Incorporations of the picture categories of the memory task into NREM
dreams, but not REM dreams, are associated with improved performance on the
memory task the next morning and 4-days later.
B) Incorporating the picture categories of the memory task during REM dreams is
associated with lower emotional and arousal ratings the next morning and at a
4-day follow-up.
2) To use TMR to verify that dreams reflect memory reactivations during sleep.
The memory task we have used in my previous experiment and plan to re-use has
been designed for use with targeted memory reactivations (Lehmann et al.,
2016). We will use this to cue specific images during sleep and then verify
with subsequent dream reports if they were incorporated into dreams. My second
research hypothesis is the following:
H2) TMR leads to subsequent incorporation of the associated image categories
into dreams during NREM and REM sleep stages.









Secondary Objective(s):
We have three more objectives we want to investigate with the dataset we plan
to collect:

4) To research the body-brain axis during sleep and dreaming. This will be done
using three methods: 1) ECG: We will investigate the associations between heart
rate variability as well as heartbeat evoked potentials and dreaming as well as
dream experience. 2) EGG: We will ex-amine the amplitude and regularity of
gastric rhythms in relation to sleep and dreaming, as well as the coupling of
the gastric rhythm with the EEG. 3) Gut microbiota: We will examine the
relationship between gut microbiota composition and sleep parameters collected
using the Fitbit over 4 weeks.

5) To generate a large dataset of dream reports and corresponding neural
activity. This will be used to apply machine learning to decode dream content
and emotion. We will use the EEG data leading up to the dream reports (2 min
interval) and use a convolutional neural network as well as Phase Space
Reconstruction and Complex Networks to see if it can predict from the neural
activity if the dream was neutral, positive or negative and what the dream
content was (based on the 6 topics used).

6) To establish the reliability of dream reports of dreams collected in
multiple conditions. Because the current project will utilize different methods
for the study of dreaming, it is a unique opportunity to reflect on the
trustworthiness of dream reports allowing us to further optimize the conditions
and methods for their collection. For example, targeted memory reactivation and
decoding provide an additional source of evidence that in combination with
dream reports might allow our team to triangulate underlying aspects of dream
experience. We will also address specific questions related to this project,
for example whether a potential mismatch between decoding results and
descriptions in dream reports is due to technical failures or systematic
weaknesses in dream reports.

Healthy adult volunteers will be recruited by advertisement. All subjects will
visit the EEG lab of the Donders Centre for Cognitive Neuroimaging (DCCN) for
one intake session, one adaptation night, and to two experimental sessions,
additionally, there will be two recalls the participants can fill out online at
home. Before and after each experimental session, participants will be asked to
write down their dream reports for 7 days prior and after (online
questionnaire, approximately 2-3 minutes per day) and measure their sleep using
a Fitbit and a sleep diary.

Recruitment: Participants will initiate the first contact following the study
advertisement (see E1). Study advertisement will be posted on SONA as well as
distributed on social media (fa-cebook, twitter and Instagram) and via mailing
lists and hung up as flyers. Additionally, a short recruitment video to be
posted on social media. After this, interested participants will receive
detailed information about the experiment and example informed consent forms
will be sent via email (see E2 and E3). In a short telephone call, the study
details and inclusion criteria will be discussed. They will be explicitly asked
whether they fulfill the inclusion criteria. This proce-dure is added to
prevent possible dropouts based on our exclusion criteria. If the inclusion
criteria are fulfilled, and they have considered the participation and wish to
participate, an in-take session is scheduled. Because we only recruit
participants with high English language proficiency, all recruitment letters
and information will only be provided in English.

Intake session: In a one-hour intake session, a brief recap of the study
procedure will be given. Participants will also be informed that they will be
excluded from participation in case they (i) do not fit one of the inclusion
criteria, (ii) fit any exclusion criteria, or (iii) when no data of sufficient
quality can be acquired due to any unforeseen reasons. This explicit
declaration is followed by the opportunity for the participant to ask any
remaining questions. Once all questions are answered, the participants will
sign the informed consent agreement (5 minutes). Then they will fill out all
questionnaires and tasks used to screen eligibility for the study (BNT, PSQI,
BDI, BAI, MCTQ, MRI, Dream Recall Frequency). The questionnaires will be
presented digitally using Castor EDC and are then checked for exclusion
criteria. If a par-ticipant meets one of the exclusion criteria, they will be
excluded from participation (and paid 6 ¤), and a replacement participant will
be recruited. If all criteria are fulfilled, the participants will do a
structural T1 and T2 Magnetic Resonance Imaging (MRI) scan on a Prisma or
PrismaFit (3T) (20 minutes). Then the three nights in the sleep laboratory
(adaptation and both experi-mental nights) are scheduled. The participants will
start collecting sleep data using a sleep tracker (Fitbit Inspire 2) and a
sleep diary, as well as a dream diary for one week before the first
experimental session. Both are presented digitally and can be filled out on a
computer or phone. The sleep and dream tracking procedure is explained in
detail, and participants can ask questions (10 minutes). Participants will get
a reminder on their phones to fill out their questionnaires each morning.


Adaptation night: The adaptation night is scheduled as closely as possible to
the first exper-imental night (the night before the first experimental night,
maximally seven nights before). Participants will be invited to the Donders EEG
laboratory at 21:30. They will be asked to re-frain from any alcohol/drug
intake during the study day, caffeine intake after lunch (maximum of 2 coffees
in the morning according to their usual intake) and get up at or before 08:00
(checked with participant report and sleep tracker). The participants will get
a short descrip-tion to read of the adaptation night. Then after the
participants are bed-ready, we will apply the EEG cap and EOG, EMG, ECG, and
EGG (if opted in) electrodes. During this time, the partic-ipants will fill out
the following questionnaires: a check on alcohol/drug/caffeine intake (2
minutes) the *Schlaffragebogen A* (sleep questionnaire A, lab translated from
German, SQQ 10 minutes) about the previous night and the *Mehrdimensionaler
Befindlichkeitsfragebogen* (multidimensional mood questionnaire, lab translated
from German, MDBF, 3 minutes), a lab-developed dream memory questionnaire (30
minutes on project OSF), and the daydreaming frequency scale (DDFS, 5 minutes).
They will complete a color-naming Stroop task across one practice and five
experimental blocks (24 congruent, 12 incongruent trials, 10 minutes).
Additionally, they will complete the trail-making test (TMT, 5 minutes).
At 23:00, participants will go to bed and be able to sleep until 07:00. An
investigator will always be present, and participants are instructed to knock
on the wall if they need anything (e.g., go to the toilet). If participants
cannot fall asleep (either after 1.5 hours or when participants re-quest it),
we will first remove the EGG. If they still cannot sleep (after 3 h or when
they re-quest it), we will remove all electrodes and discontinue the study
(they will have the option to sleep in the laboratory or go home). At 07:00,
the sleep opportunity will end. They will fill out a questionnaire about their
sleep quality (SQQ) and recall their dreams. Then the EEG and oth-er electrodes
will be removed, and participants can shower and get dressed. Afterward, we
will confirm that they want to continue the study and are eligible based on the
sleep quality. At around 7:40, the adaptation night will be done.

Experimental Sessions
The two experimental sessions will be counterbalanced between the participants
with random assignment and additional counterbalancing of the memory task
categories (random number generator (sample in R) will be used for each
participant). Participants are blinded to the con-dition. The two experimental
conditions are scheduled at least 14 days apart. Participants are instructed to
abstain from alcohol and drugs on experimental days and to get up before 08:00.
No caffeine intake is allowed after lunch, with a maximum of two coffees in the
morning. Alco-hol and caffeine intake is checked with a questionnaire.
Furthermore, sleep tracker data will be checked to confirm that no sleep nights
have been skipped in the previous week. A stool sample is collected by the
participant with a kit (OM-NIgene•GUT | OM-200) on the day of the experimental
session (not analyzed within this study, opt-in by participants). The
experimental sessions will start at 19:30. The participants will get a written
instruction explaining the experimental session. Afterward, they will get ready
for bed. Then the EEG will be applied.

Session A: Awakenings
During the EEG application, the participants will be able to ask any questions
about the awak-ening protocol (the same questions as used at home). For the
remaining time during EEG application, the participant will fill out the
following questionnaires: the alcohol/coffee check (2 minutes), the Mannheim
Dream Questionnaire (MADRE, 10 minutes), the Brief-COPE ques-tionnaire (10
minutes), the MDBF (3 minutes), the need for closure scale (NFCS, 15 minutes),
and the Freiburg Mindfulness Inventory (FMI, 5 minutes). Afterward, the
participants will un-dergo the learning blocks of the memory task. Between the
learning blocks and the recall, there will be a short 10 minutes break during
which the participants will fill out the MDBF again and the SQQ for the
previous night. Recall happens in 2 blocks which take approximately 30 minutes.
At 23:00, participants will go to bed. When the participant is lying in bed, we
will do a resting-state EEG measurement (1.5 min eyes open, 1.5 min eyes
closed, 1.5 min eyes o

Considering the extensive exclusion criteria, the screening procedure, constant
monitoring of the subjects we do not expect (S)AE (serious adverse effects)
side effects. EEG, EGG and MRI measurements themselves do not pose any risk, if
appropriate precautions are made. However, for the MRI the noise and the
relative confined space of the MRI scanner may cause discomfort to some
subjects. However, the scan will only be very short (< 20 mins).

In case of an incidental finding concerning a deviation in the MRI scan, step
1, the researcher will immediately contact the MRI lab manager:
a. If it concerns a healthy subject the MRI lab manager will send the images
out for assessment by a radiologist. At this stage the participant is not
informed, but the incidental finding will be documented (proceed to step 2).
b. If it concerns a patient and/or minor the researcher is required to consult
the responsible study MD.
* The responsible MD is trained to judge the images adequately proceed step 3.
* The responsible MD is not trained to judge the images proceed step 1a.
Step 2, if, according to a written report of the radiologist,
a. No clinically relevant finding is obtained, the participant will not be
informed, but the docu- mentation updated.
b. A clinical relevant finding is obtained; the home physician will receive a
letter. Also the par- ticipant will receive a letter, which, however, does not
reveal any diagnosis but which asks the participant to contact his/ her home
physician.
c. The incidental finding will be centrally documented at the DCCN.
Step 3, if the responsible MD considers the abnormality as clinically relevant
the home physi- cian or the treating specialist will be personally contacted by
the responsible MD. (for details see K6b).
Because of the repeated awakenings, participants will be asked to not drive
after the study because of the awakenings. We also recommend not to use the
bicycle, but ideally to take the public transport or be picked up by someone.
If this is not possible for the participants, we will offer to arrange a taxi
to go home.
The consent discussion starts sufficiently in advance of the initiation of
study-related procedures to allow potential subjects time to reflect on the
potential benefits and risks and possible discomforts. Participants are
informed about our standard studies per email before they agree to sign up and
they are able to see a consent form and take time to reflect if they want to
participate in the study. Participants are informed that the risk associated
with participation in this study in total can be regarded as minimal. The
results of this study will provide us with better insight into dreams
specifically in relation to memory consolidation but additionally also about
the associated neural activity and dream reports.