The study of dreams in scientific research laboratories began with the discovery of REM sleep by Aserinsky and Kleitman (1953). They are credited with having the first sleep research laboratory that discovered the “rapid, jerky, binocularly symmetrical movements” characteristic of REM sleep. They further associated dreams with these eye movements, finding that 74.1 percent of participants recalled dreams from REM sleep, whereas only 17.4 percent recalled dreams from NREM sleep.
Around the same time, Calvin Hall was working on a new cognitive theory of dreaming, and in 1966, Hall and Van de Castle created and published a thorough content scoring system for dreams (Hall & van de Castle, 1966). Together, these papers stimulated the acceleration of dream research publications and allowed for a systematic method of studying dreams.
Since then, researchers have found the correlation of dream recall and REM sleep to be consistently high (around 80 percent on average). However, there has been an increase in the percentage of NREM dream recall found in studies over time, though this data remains inconsistent (Nielsen, 2011). These inconsistencies could be due to the varying definitions of dreaming used by experimenters. Some used 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 imagery.
Further, 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?” This is still a major problem in dream research, since varying methods of collection largely influence the frequency and length of dream reports.
Further inconsistencies in estimates of dream recall can be attributed to participant/experimenter expectancies. Herman et al. (1978) experimentally influenced the collection of dreams by giving a placebo pill to participants who were then told, along with the experimenter, that the placebo would increase dream recall. This created high expectations in both the participant and the experimenter, and in turn, higher dream recall was observed when compared to a control group (Herman et al., 1978).
To further complicate the picture, dream recall can vary even within a particular sleep state. For example, analyzing dream recollection as a function of the amount of time elapsed within NREM or REM sleep before awakening reveals a cyclical pattern. There is an increase in dream recall up to 45 minutes into REM sleep, followed by a decrease. The opposite effect is found in NREM sleep, with a longer time in NREM correlated with less dream recall, up until a slight increase that occurs just before the transition into REM sleep (Stickgold et al., 1994).
In another example, in one experiment using an ultra-short sleep/wake protocol, participants took 20-minute naps every hour for 78 hours, reporting and rating dreams after each nap. The researchers found that NREM dream intensity peaked in the morning hours, despite NREM sleep stage duration being minimal at this time (Suzuki et al., 2004). Rather, REM related processes, which peak in the morning appeared to be influencing the NREM dreams' intensity in a "covert" manner (Nielsen, 2000).
Relationships between REM sleep neurophysiology and dreaming have been demonstrated with a number of other methods. Herman et al. (1984) found that the rapid eye movements participants produced just prior to waking up matched their visual descriptions of dream imagery. For instance, an eye movement "up and to the left" was associated with dream imagery with looking "up and to the left." This evidence supports the scanning hypothesis of dreaming, which posits that eye movements in REM sleep are a result of scanning the dreamscape, although not all studies support the scanning hypothesis (Ogawa, Nittono, & Hori, 2002).
In another example, neuroimaging of lucid dreams produced neural correlates of dream activity that were similar to activity that would be found in wake. For instance, fMRI BOLD responses were observed in the same sensorimotor cortical regions for both wakefulness and lucid dreaming, though activation during dreaming was weaker and more localized (Dresler et al., 2011). Further, in patients with REM sleep behavior disorder, dreamed behaviors are physically enacted during sleep, reinforcing the notion that motor activity in the dreaming brain is similar to that of waking actions (Valli et al., 2015).
Altogether, the study of dreams has evolved substantially since the discovery of REM sleep. REM sleep has been consistently associated with higher dream recall than NREM sleep, although variations in methodology, time of night, and participant expectations can influence or increase the frequency of dream recall from NREM sleep.