Dream Recall and Content in Different Stages of Sleep and Time-of-Night Effect

Full text

Dream Recall and Content Carr & Solomonova
1
Dream Recall and Content in Different Sleep Stages
Michelle Carr1 & Elizaveta Solomonova2
1Sleep Laboratory, Swansea University, Swansea, United Kingdom
2Dream and Nightmare Laboratory, Center for Advanced Research in Sleep Medicine; and
Psychiatry Department, McGill University, Montreal, Quebec, Canada
To appear in: Dreams: Biology, Psychology and Culture. Valli, K., and Hoss, R. editors. Greenwood
Publishing Group.
Many researchers have attempted to characterize the association between sleep physiology
and dreaming, with some consensus that both dream recall frequency and dream content fluctuate
in predictable ways across a night of sleep. Dream recall and content seem to be tempered, on the
one hand, by the cycle of sleep stages which occurs approximately every 90-minutes, termed the
ultradian rhythm, as well as by the total duration of sleep and the time of night, linked to the
circadian rhythm, the approximately 24-hour long cycle.
The REM sleep = dreaming paradigm
The study of dreams in scientific research laboratories began with the discovery of rapid-
eye-movement (REM) sleep by Aserinsky and Kleitman (1953), who are credited with having the
first sleep research laboratory that discovered the “rapid, jerky, binocularly symmetrical
movements” of the eyes, characteristic of that sleep stage, and intuited that these eye movements
correspond with the visual activity of dreaming—what’s now known as the “scanning hypothesis”.
They further associated dreams with these eye movements, finding that 74.1% of participants
recalled dreams from REM sleep, whereas only 17.4% recalled dreams from Non-REM sleep
(NREM). In 1957, researchers Dement and Kleitman conducted a similar study eliciting 351
awakenings in nine participants during different stages of sleep, which corroborated earlier
findings of a very high percentage of participants recalling dreams from REM sleep (80%),
whereas only 6.9% recalled dreams from NREM sleep. These early studies instituted the notion
that dreaming may be a REM sleep phenomenon and research continued for many years under that
assumption with several other laboratory studies observing similarly high recall from REM sleep
and low recall from NREM sleep. Some researchers even referred to REM sleep as dream sleep or
D-sleep. To date, researchers find the correlation of dream recall and REM sleep to be consistently
high (around 80% on average), suggesting that the neurophysiology of REM sleep is a strong
enabling condition for vivid dreaming and for cognitive processes, such as attention and memory,
that underlie dream recall upon awakening.
Despite the evidence that dreams could be recalled at least some of the time from NREM
sleep, initially it was presumed that the relatively infrequent NREM dream reports were actually
due to memories of prior REM sleep dreams occurring earlier in the sleep cycle, as it was thought
that NREM sleep itself could not give rise to dreaming. Nevertheless, in 1962, researcher David

Dream Recall and Content Carr & Solomonova
2
Foulkes conducted a study collecting 134 dream reports across the night, which demonstrated that
in fact dream recall can be observed from any stage of sleep and at any time of the night, including
in a NREM sleep stage that preceded any REM sleep. This meant that the NREM dream recall
could not be attributed to a prior REM dream. Since then, many researchers have confirmed that
dream reports can be obtained from all NREM sleep stages.
Further, there has been an increase in the percentage of NREM dream recall found in
studies over time, though this data remains inconsistent (Nielsen, 2011). Inconsistencies are likely
due to the varying definitions of dreaming used by experimenters. Some use a more inclusive
dreaming definition, such as ‘sleep mentation’, which is characterized by any cognitive activity,
whereas the original concept of dreaming was more qualified as hallucinatory, predominantly
visual, imagery (see also chapter 5). The questions used by an experimenter can alter a participant’s
interpretation of dreaming. For example, the question, “what was going through your mind just
before I called you”, will encourage more frequent dream reports than, “what were you dreaming
about?” In fact, in the aforementioned 1962 study, Foulkes discovered that by modifying the
question posed to participants on awakening, by asking whether they were thinking of something
before being signaled rather than if they recalled a dream, he was able to obtain 70% recall rates
from NREM sleep; in other words, the majority of participants reported having had some thoughts
in mind prior to awakening from NREM sleep. Although initially criticized, these observations
have since been confirmed by other researchers, proving again that dream reports can be obtained
from all stages of sleep if sampled appropriately.
As another methodological consideration, the circumstances around an awakening also
play a significant role in frequency of dream recall. For example, if a participant has a task to
complete immediately upon awakening, this distraction will interfere with and reduce dream recall
compared to participants who are allowed to lie in bed and collect their memory of a dream. Sudden
awakenings are also shown to induce better recall than gradual awakenings. In effect, anything
that distracts an individual upon awakening will interfere with their ability to compose and gather
their memory of a dream. Lastly, not recalling a dream is not a proof of absence of an experience
during sleep. There may be many factors that contribute to remembering and forgetting of an
experience, including individual introspective and motivational factors.
Variations in dream content by sleep stage
In modern day dream research, mentation can be sampled from 4 major sleep stages, with
NREM sleep being split into stage 1 sleep which occurs at the transition between wakefulness and
sleep, to stage 2 light sleep, and finally to stage 3 deep sleep. The majority of researchers agree
that some quality of mentation can occur in any of these three stages of NREM sleep, and of course
is most frequent in REM sleep. However, while dreams may be recalled from any stage, there are
differences in the type and quality of mental activity occurring in these different stages. For
example, dream reports are longer after REM than after NREM awakenings, and REM dream
reports are consistently more elaborate and immersive than are NREM dream reports, which are
more thought-like, and even mundane.
Several studies in the past decade have explored dreams occurring in the brief transition
between sleep and wakefulness which can occur both at sleep onset, and upon awakening, termed

Dream Recall and Content Carr & Solomonova
3
the hypnagogic and hypnopompic states, respectively. The dream reports obtained at the moment
of sleep onset can surpass the frequency and length of dream recall obtained from other NREM
sleep stages. A comprehensive review by Nielsen (2011) found that anywhere between 31-76% of
stage N1 sleep reports contain dream imagery. The content of stage N1 (sleep onset) dream reports
is often brief but with vivid visual and auditory experiences and even sensations of movement in
the body, as in the common sensation of ‘falling’ or ‘dropping’ off to sleep. Another interesting
phenomenon that occurs in hypnagogic and hypnopompic states is the incorporation of external
sensory detail into imagery, such as the alarm clock going off and creating a dream of a dog
barking. This sort of imagery where a real world stimulus seems to give rise to an image was noted
in the famous guillotine dream of Maury, where the headboard of the bed fell and Maury dreamt
of a guillotine dropping on his neck (see also chapter 7). Numerous examples in the literature gave
rise to the theory, known as the “cassette theory of dreaming” that posits that dreaming in fact
occurs only at the moment of awakening, as the conscious mind interprets sensory stimuli through
imagery, and mentally “rewinds” the memory to create a longer and a more coherent narrative. For
many scholars, surprisingly, this theory was perceived as more rational than supposing that
cognitive activity could possibly occur in the sleeping state. With the advent of neuroimaging and
more precise methods of assessing cognitive activity during sleep, including work in lucid
dreaming, researchers now concede that dreaming and cognitive activity do occur within the sleep
state itself.
Stage N2 dream reports have been the most thoroughly collected and studied compared to
other NREM sleep stages, and in the past 15 years the percentage of awakenings that are associated
with some report of mental activity from stage N2 sleep has hovered around 60-70%. Despite this
high recall frequency, comparative analysis between stage N2 and REM sleep dream reports has
consistently found stage N2 dreams to be less frequent and shorter than REM dreams, to have less
perceptual content, characters, places, actions, and emotional content, and to be less personally
involving. Stage N2 dream reports also seem to be less bizarre, to have logical thought similar to
that in wake, and to incorporate fragments of recent waking life episodes. Several of these
differences have been verified in multiple studies, and it has become common practice to refer to
the brief dreams characteristic of stage N2 sleep as more ‘thoughtlike’, while the narrative REM
dream reports are often described as more ‘dreamlike’. Some authors have contested whether stage
N2 mentation can be categorically qualified as ‘dreaming’ per se, due to these qualitative
differences. Nevertheless, these are not absolute qualities, as there is a broad spectrum of
individual differences in quality of stage N2 dream reports, and in fact, a small minority of dreams
reported from stage N2 sleep are actually quite ‘dreamlike’ and even indistinguishable from typical
REM sleep dream reports.
Finally, deep slow wave sleep (stage N3 NREM sleep) has also been associated with dream
activity, particularly by Cavallero and colleagues (e.g., 1992). In this study, 65% of slow wave
sleep awakenings were accompanied by some mental activity, and although many reports were
quite short, they still showed clear signs of dream imagery of self, perceptual, and emotional
qualities. Early researchers considered it physiologically impossible for cognitive processes to
occur during the relative depth and neural quiescence of deep sleep. More recently, neuroimaging
and behavioral studies have found that slow wave sleep is in fact far from being a period absent of
activity, and indeed may be involved in processes of memory consolidation and cognitive

Dream Recall and Content Carr & Solomonova
4
enhancement. In general, the existence of dream generation and veridically recalled dreams from
all stages of NREM sleep is now admitted by the majority of researchers.
It is well accepted that REM sleep dreams are the most ‘dream-like’ in their immersive and
narrative, yet bizarre nature. And given that REM sleep is much more predominant in the early
morning, these are the types of dreams that people most often awaken from and remember in their
daily life. Recent research suggests that certain REM dream qualities, such as emotional intensity,
are driven by the neurophysiology of the REM sleep state. For instance, activity in the limbic
system, which is associated with emotional processing, is heightened in REM sleep and is
somewhat inhibited in NREM sleep stages. The emotional intensity of REM sleep dreams is part
of what makes them so engaging and memorable. Another common attribute of REM sleep dreams
is their reference to recent waking life concerns and incorporations of recent waking life memories.
This pattern of incorporating memories for experiences from the previous day, termed the day
residue, as well as memories for those experiences from about a week ago, termed the dream-lag
effect. The combination of recent personally relevant and emotional memory traces woven into
the long narrative nature of REM dreams, suggests that dreams may contribute to some mnemonic
process attempting to integrate and make sense of those experiences, which are of most import in
waking life.
Nevertheless, there is some contention over whether stage-related differences in content
may be simply due to the duration of a dream, with longer dreams occurring in REM sleep, which
then allow for longer and greater descriptive detail in dream reports. Dream report length can be
assessed either by counting the number of content-bearing words in a report, or by counting the
number of temporal units in a report, e.g., each action or interaction is a distinct temporal unit.
Some researchers have suggested that the qualitative differences mentioned above are only
secondary to a simple quantitative difference in dream length, and indeed, certain REM/NREM
dream content differences tend to disappear when word length is controlled, although not all: in
particular, REM dreams remain more emotionally salient than NREM dreams.
Further, support for claims that REM dreams differ qualitatively from NREM dreams exists
in the various forms of intensified dreams which are more frequently recalled from REM sleep.
For instance, lucid dreams, which are dreams where the dreamer is aware that they are dreaming,
are specifically recalled from REM sleep. In experiments led by LaBerge (1986), frequent lucid
dreamers were recruited to sleep in the laboratory, where they reported emotionally and
perceptually vivid dreams marked by a mental capacity strikingly similar to that of waking life,
including access to logic, voluntary control of thoughts and actions, and memory, and were able
to recall these experiences quite clearly on awakening. These experiences were in large part
occurring during REM sleep, and only rarely during Stage N2 sleep. In another example,
nightmares are a form of emotionally intense and arousing dream that often result in a sudden
awakening. These dreams likewise occur most frequently in REM sleep, corresponding with
claims that REM dreams are specifically marked by emotional intensity.
In general, certain dream features remain significantly different between REM and NREM
sleep even when word length is controlled, and consistent reports suggest that REM dreams are
more self-reflective, bizarre, visual, emotional, have more self-involvement, and more narrative

Dream Recall and Content Carr & Solomonova
5
quality than do NREM dreams, whereas NREM dreams are more brief, ‘thought-like’ and
mundane than REM dreams.
Variation in recall and content across the ultradian sleep
cycle
Because stages of NREM and REM sleep are defined categorically, a hypnogram of a
normal night of sleep seems to follow a pattern of abruptly transitioning from one type of sleep to
the other (see chapter 2). However, many studies indicate that transitions between sleep stages are
in fact much more fluid than previously thought, following a 90-minute cycle that continuously
ebbs and flows from deep NREM to light NREM to REM sleep and back. Collecting dream reports
at multiple time points within a given sleep stage conforms with this cyclical pattern, with both
frequency and content of dreaming showing oscillatory changes throughout a sleep cycle, as
opposed to sudden changes in dream quality occurring in concert with categorically determined
sleep stage boundaries.
For instance, when relationships between dream report length and time elapsed in REM or
NREM sleep are assessed, dream length fluctuates in a cyclical, sinusoidal pattern over time.
Report length is thought to measure the overall quantity of dream content, and, as mentioned
earlier, REM dream reports are consistently longer than are NREM dream reports. However, more
detailed empirical work suggests that dream length in fact oscillates over consecutive REM and
NREM episodes in an ultradian pattern. A study by Stickgold and colleagues (1994), found that,
for REM sleep dream reports, overall report length is lowest at the beginning (0–15 minutes) and
at the end (45–60 minutes) of a REM period, and length is highest in the middle of the REM period
(15-45 minutes). NREM sleep dream reports show the opposite pattern, with dream length
decreasing into the middle of the period, before increasing again when nearing an oncoming REM
period.
Similar results have been found in other studies that assessed dream report length in concert
with duration of time spent in REM or NREM sleep. For instance, several studies have found that
NREM dream reports are longer and more frequent when they are sampled in close proximity to a
REM period, whereas those sampled further from REM sleep are shorter and less frequent. Finally,
the most reliable indicator of dream recall to date is the underlying patterns of brain activity.
Studies show that in both REM and NREM sleep certain brain rhythms (in particular, decrease in
low frequency delta activity) predicts whether or not the sleeper will be able to recall a dream upon
awakening at that moment (Esposito et al., 2004; Scarpelli, 2017). In other words, dream recall is
better predicted by transient brain rhythms than simply by a sleep stage.
As is the case for measures of dream recall and length, evidence indicates that various
features of dream imagery progressively amplify with increasing time in REM sleep, and may
decrease with increasing time in NREM sleep. For instance, dreams are considered more ‘dream-
like’ when collected from shorter as opposed to longer NREM sleep periods, as well as being more
‘dream-like’ collected in close proximity to a neighboring REM period, e.g., when sampled from
a NREM sleep period just 5 minutes after a previous REM period ended. For REM sleep, dream
reports from longer REM sleep episodes become more dreamlike in several respects than those

Dream Recall and Content Carr & Solomonova
6
from shorter REM periods; longer REM sleep reports have been described as more active,
emotional, and vivid, with more visual features and narrative immersion than shorter REM sleep
reports. In general, these attributes are thought to show linear increases over time in REM sleep.
Nevertheless, these measures are not always consistent between studies, and some have found that
certain features of REM dreams do not change as a function of stage duration.
In general, the findings for recall and length show clear patterns in sync with the 90-minute
ultradian sleep cycle, whereas the specific quality of dream content shows some, but less
consistent, patterns of ultradian variation.
Variation in recall and content across a night of sleep
A growing body of research suggests that the frequency and content of dreaming evolves
over the course of a night, with dreams becoming more realistic and immersive in later sleep
cycles. In general, the dreams collected from the first or second 90-minute cycle of sleep are
markedly different from those in later cycles, and especially from the dreams occurring at the end
of a sleep period in the morning.
There is a great deal of evidence that dreams late in the sleep period are longer than early
dreams. Pivik and Foulkes (1968) were the first to empirically demonstrate that dream length
changes across the night, with the finding that NREM dream reports became longer with each
successive sleep cycle. Since then, several other researchers have followed suit with varying sleep
paradigms, for instance, showing that dreams collected in the first half of the night are shorter than
those collected in the second half of the night, and that longer sleep durations are associated with
longer dream reports. Other qualities of dream content also seem to reliably amplify across the
night. In general, studies have found that perceptual details, including visual features, are
augmented in late morning dreams, and laboratory studies, have shown there is a positive
correlation between the time of night and the vividness, bizarreness and emotional intensity of
dreams - for both REM and NREM dreams across the night. In other words, dreams from all stages
of sleep become more ‘dreamlike’ as the night goes on. There is also some evidence that dreaming
across the night exhibits some continuity in thematic content, for example, dreams sampled from
different stages and across a single night may all carry an overarching theme related to work or
occupational concerns, even if the perceptual or emotional quality and specific details of the dream
report change. This suggests there is some continuum of dreaming processes occurring across the
night.
It is thought that the circadian peak, which occurs at the end of the sleep cycle in the early
morning, stimulates higher dream recall and a greater predominance of attributes such as character
interactions, emotional contents, and dream vividness. However, some researchers have suggested
that increasing dream vividness is tied to the duration of the sleep period, as opposed to the 24-
hour circadian rhythm. Evidence for a specific circadian influence on dreaming has been shown in
studies that alter the timing of a sleep period without altering sleep duration. For instance, delaying
sleep by 3 hours was associated with more vivid dreaming in one study (Antrobus, 1995). Others
have used nap protocols to assess dream content outside of a normal sleep period. For example,
one experiment used an ultra-short nap protocol where participants took 20-minute naps every
hour for 78 hours and reported dreams after each nap. Consistent with a circadian explanation, it

Dream Recall and Content Carr & Solomonova
7
was found that NREM dream intensity peaked in the morning, even though NREM sleep duration
is shortest in the morning (Suzuki et al., 2004). Finally, a recent study used a morning nap
paradigm, and found dream recall rates for NREM (89%) and REM (96%) sleep that were even
higher than those seen for a typical night of sleep (Carr & Nielsen, 2015). This suggests that a
circadian peak in dream generation may occur in the morning, regardless of the duration of the
sleep period.
While dreams sampled from both NREM and REM sleep seem to intensify across the
course of the night, the two types of reports continue to differ in some qualitative respects. In
particular, REM dream reports continue to exhibit elevated levels of bizarreness and emotional
content, and more vivid perceptual experiences, which suggests that certain stage-related content
differences may persist across the night. Additional research comparing dreams sampled at
different time-points throughout the day and night is needed to determine the qualities of REM
and NREM dream reports that are more influenced by circadian or stage-related mechanisms.
In general, it seems that both circadian- and ultradian-rhythms exert some influence on
dream content, and while late-REM dream reports continue to be more emotionally intense and
vivid than late-NREM dream reports, both REM and NREM dream reports are more bizarre,
dreamlike, and longer later in the night.
Conclusion
Dreams, while initially associated only with REM sleep, are in fact possible in all stages
of sleep. Though REM sleep continues to be consistently associated with higher dream recall than
NREM sleep, variations in methodology, sleep stage duration, and time of night, can influence the
content or increase the frequency of dream recall from NREM sleep. In general, the quality of
NREM and REM dream reports differs in consistent ways. NREM dreams are typically shorter,
more fragmented and more thought-like whereas REM dreams are longer, more emotional, and
more bizarre. Attributes such as length, bizarreness, and perceptual vividness increase for both
NREM and REM sleep reports across the night, although REM sleep dreams continue to be more
emotionally and perceptually vivid than NREM dreams. While these results require more
clarifying research, they demonstrate that various features of dreaming are driven by both ultradian
and circadian cycles, with the time-of-night increase in dream vividness likely due to circadian
influence, and the REM/NREM differences driven by ultradian influence.
References
Antrobus, J., Kondo, T., Reinsel, R., & Fein, G. (1995). Dreaming in the late morning:
Summation of REM and diurnal cortical activation. Consciousness and Cognition, 4(3),
275–299.
Aserinsky, E., & Kleitman, N. (1953). Regularly occurring periods of eye motility, and
concomitant phenomena, during sleep. Science, 118(3062), 273–274.
Carr, M., & Nielsen, T. (2015). Daydreams and nap dreams: content comparisons.
Consciousness and Cognition, 36, 196–205.

Dream Recall and Content Carr & Solomonova
8
Cavallero, C., Cicogna, P., Natale, V., Occhionero, M., & Zito, A. (1992). Slow wave sleep
dreaming. Sleep, 15(6), 562–566.
Dement, W., & Kleitman, N. (1957). Cyclic variations in EEG during sleep and their relation to
eye movements, body motility, and dreaming. Electroencephalography and Clinical
Neurophysiology, 9(4), 673–690.
Esposito, M. J., Nielsen, T. A., & Paquette, T. (2004). Reduced Alpha power associated with the
recall of mentation from Stage 2 and Stage REM sleep. Psychophysiology, 41(2), 288–
297.
Foulkes, W. D. (1962). Dream reports from different stages of sleep. The Journal of Abnormal
and Social Psychology, 65(1), 14.
Hobson, J. A. (1988). The dreaming brain. New York: Basic Books.
LaBerge, S., Levitan, L., & Dement, W. C. (1986). Lucid dreaming: Physiological correlates of
consciousness during REM sleep. The Journal of Mind and Behavior, 7, 251–258.
McNamara, P., McLaren, D., Smith, D., Brown, A., & Stickgold, R. (2005). A “Jekyll and
Hyde” within: aggressive versus friendly interactions in REM and non-REM dreams.
Psychological Science, 16(2), 130–136.
Nielsen, T. (2010). Ultradian, circadian and sleep-dependent features of dreaming. In M. Kryger,
T. Roth & W. C. Dement (Eds), Principles and Practice of Sleep Medicine (pp. 576-584).
New York: Elsevier.
Nielsen, T. A. (2000). A review of mentation in REM and NREM sleep:“covert” REM sleep as a
possible reconciliation of two opposing models. Behavioral and Brain Sciences, 23(6),
851–866.
Pivik, T., & Foulkes, D. (1968). NREM mentation: relation to personality, orientation time, and
time of night. Journal of Consulting and Clinical Psychology, 32(2), 144.
Scarpelli, S., D’Atri, A., Mangiaruga, A., Marzano, C., Gorgoni, M., Schiappa, C., … De
Gennaro, L. (2017). Predicting Dream Recall: EEG Activation During NREM Sleep or
Shared Mechanisms with Wakefulness? Brain Topography, 30, 1–10.
Stickgold, R., Pace-Schott, E., & Hobson, J. A. (1994). A new paradigm for dream research:
mentation reports following spontaneous arousal from REM and NREM sleep recorded
in a home setting. Consciousness and Cognition, 3(1), 16–29.
Suzuki, H., Uchiyama, M., Tagaya, H., Ozaki, A., Kuriyama, K., Aritake, S., … Kuga, R.
(2004). Dreaming during non-rapid eye movement sleep in the absence of prior rapid eye
movement sleep. Sleep, 27(8), 1486–1490.