Due to school, I’ve been given access to an enormous database of scientific papers dating back as early to the 1960s. So far I’ve found information about 130 papers about experiments concerning LDs, SP, HI, DCs, Old Hag experiences, etc… (including the initial papers published by LaBerge). In a series of articles, I’ll try to give a general description of some of the more interesting papers (purpose, experiments used, results, discussion, conclusion). It may be a boring read to some, but imo it’s good to notice how much precautions and criteria scientists must take into account before one’s able to obtain valuable data. But rest assured, in the other articles, I won’t write all the procedures down in so much detail as I did here 
First paper:
Tyson, P.D., Ogilvie, R.D. & Hunt, H.T. (1984). Lucid, prelucid, and nonlucid dreams related to the amount of EEG alpha activity during REM sleep. Psychophysiology, 21, 442–451.
Purpose
To set up an experiment to determine whether or not alpha activity during REM sleep (without waking up as a result, because alpha brainwaves are often correlated with waking life brain activity) can discriminate between the classification of prelucid, lucid and nonlucid dreams.
One particular early study (Ogilvie et al. (1978)) already suggested there’s an association between LDs and high amplitude EEG alpha activity. This is remarkable actually, because it’s paradoxical with the regular standardized sleep stage scoring systems: these define a minimum of 10s of alpha activity within a 20s epoch as wakefulness. Furthermore, the amount of alpha activity within REM has been correlated with waking life alpha, although REM alpha responds differently to stimulation than waking alpha. The study which I’d like to present focused on whether the amount of alpha activity during REM sleep, without signs of waking up, was associated with reports of lucidity within the dream.
In prelucid dreams, the dreamer develops a critical attitude towards the dream by questioning the apparent reality of that dream. This critical attitude may not develop into a LD after all, and LDs may not all evolve from prelucid experiences. Furthermore, the content of LDs emerging from prelucid dreams may be different from deliberate training to produce LDs. The general features of LDs emerging from LD induction training are active control of imagery, emotional detachment and critical rationality - these are all typical of left brain hemispheric dominance. On the other hand, LDs emerging from prelucid dreams with untrained dreamers exhibited less control, more emotionality and considerably more bizarre elements. Overall, prelucid dreams appeared to be the most bizarre (clouded and hallucinatory), nonlucid dreams were mundane and realistic, and LDs were closer to the realism of nonLDs.
Experiment
There are at least two convergent approaches which can be used to examine the alpha-lucidity hypothesis:
-
Systematic observations of dream content were done by arousing subjects when their REM alpha levels were particularly high or low. The determination of whether a signal is high or low is based upon subjective estimates of the relative amount of filtered alpha 20 or 30s before the arousal. This can be refined by actually calculating the actual mean alpha amplitude preceding each arousal.
-
Computer analysis of the EEG activity preceding each arousal. The computer analysis data from the previous experiment by Ogilvie et al. can be used to refine the testing of the alpha-lucidity hypothesis by quantifying the spectral and temporal aspects of the EEG, thereby standardizing the individual’s alpha amplitude criterion for an arousal, and analyzing the degree to which REM alpha and content variables discriminate between nonlucid, prelucid and nonlucid dreams.
For this particular experiment, five males and five females (aged 19-31) were selected on the bases of recalling five or more dreams a week and having good alpha activity. All of them have reported of having LDs at least occasionally.
All the subjects were required to spend one adaptation night and one or two arousal nights in the sleep laboratory. During the adaptation night, no arousals were made, but it was used to set a within-subject criterion for high and low alpha REM arousals. On the arousal nights, subjects were told that they would be awakened during REM sleep four times and were made familiar with the questionnaire, but were completely blind to the type of arousal, to control for suggestibility and demand characteristics. To ensure that LD reports were not artifacts of awakening from REM sleep, the arousal criterion required at least five minutes in REM stage, a minimum of 20s above or below the alpha criterion, and no movement artifacts, elevations of muscular activity (EMG), eyeblinks, abrupt changes in heart rate or respiratory activity, or other discontinuities during REM. A very careful examination of the sleep records after the experiment was used to verify the arousal criterion for each arousal and eventually led to the rejection of two arousals as ambiguous.
Following each arousal, the experimenter reads a list of eight questions to eliminate the possibility of different demand characteristics biasing the interrogation. The taped arousal interviews were coded, transcribed blind and given to HTH (some processing unit I guess…) for scoring. HTH was blind to the code, and initially, HTH was given only the narrative protocols to rate for dream content and bizarreness using the scoring methods developped and reliability tested on a sample of 800 dreams. The lucidity ratings were obtained during a second examination of the recoded, transcribed answers to standardized lucidity and prelucidity. Positive answers to the questions were subdivided in terms of whether a brief momentary awareness was being described, an awareness that lasted through a definite narrative sequence (rated beginning, middle or end), or an awareness roughly continuous througout the entire dream.
I won’t go into detail of how the data was analyzed (not that I have a clue of what it all means ), but here are the results:
Results and discussion
After removing the two ambiguous dreams, analysis was done using the 38 reported dreams, classified as lucid, prelucid or nonlucid. An examination of the EEG 30s before the arousal in terms of alpha amplitude and standardized amplitude showed a clear relationship between high alpha activity and prelucid dreams. From the obtained data, one could directly derive prelucid dreams were the highest in alpha activity and most deviant relative to each subject’s mean REM amplitude. Low amplitude alpha preceding the arousal indicated either a nonlucid or a lucid dream.
Ogilvie et al. (1982) hypothesized a linear relationship between the amount of alpha activity and lucidity, such that low amplitude alpha would coexist with nonlucid reports, moderate alpha with prelucid or brief lucid episodes, and high alpha with lucidity throughout the dream. The above association of prelucidity with high alpha led to reordering the earlier lucidity scale from Ogilvie et al., so that prelucid rating were weighted more heavily than lucid ratings.
So far we have learned how to distinguish prelucid dreams from nonlucid and lucid dreams through high alpha activity after at least five minutes of REM sleep, 30s before arousal. But can we make a distinction between nonlucid and lucid dreams? Therefore we need to have a look at the alpha activity at the beginning of the dream. The experimenters found out that lower alpha deviations at the beginning of the dream might be associated with nonlucidity at the time of the arousal, whereas both lucid and prelucid dreams had higher alpha in the beginning. By weighing out the alpha patterns towards each other, one could derive the three types of dreams were empirically distinct from each other, and the most significant difference was between nonlucid and prelucid dreams. The dreams with the highest probability of being classified as nonlucid had relatively low levels of alpha activity at the end and beginning of the REM period, while the dreams classified as prelucid dreams had relatively high levels at the end and beginning. As REM alpha levels became less extreme, the probability of correct classification was reduced; therefore LDs classified by moderately low levels of alpha at the beginning were the most difficult dream type to discriminate. Overall, the discriminant score for each dream correctly classified 71% of the dreams using the beginning and end of the REM period.
Blind scoring of dream content found the three types of dreams significantly different on both dimensions of bizarreness (clouding and hallucinosis), but not significantly different in control or self-rated emotional involvement in the dream:
-
the amount of clouding in prelucid dreams was significantly greater that both nonlucid and lucid dreams, but
-
both nonlucid and lucid dreams were not significantly different in reported clouding;
-
the prelucid hallucinatory reports were significantly greater than nonlucid dreams, but
-
hallucinosis was unable to significantly discriminate between prelucid and lucid, or lucid and nonlucid.
The revised scale which ordered prelucidity greater than lucidity was significantly correlated with clouding and hallucinosis (as were these both dimensions of bizarreness significantly correlated with each other). Related findings included:
-
bizarre content was significantly correlated with alpha activity at the beginning of the REM period, but
-
not significantly correlated with alpha activity preceding the arousal;
-
bizarre content seems to be the most typical of prelucid dreams and high REM alpha at the beginning of the REM period, but
-
REM alpha was a better discriminator of LD types than any of the dimensions of dream content.
Cohen (1979) hypothesized that dream lucidity would gradually increase during the period asleep reflecting a shift from right to left hemispheric dominance. There was a tendency for the three dream types to differ in terms of time of night and not to differ significantly for REM or arousal number. The average time of night in hours and minutes of prelucid (5:38) and lucid (5:02) dreams was later than nonlucid dreams (3:55), but only prelucid dreams were significantly different from nonlucid dreams. Another interesting point is the apparent significant increase of self-rated emotional involvement with the number of arousals during the night.
In summary, the major finding was a relatively distinct pattern of REM alpha for prelucid, nonlucid and lucid dreams which correctly classified 71% of the dreams. Prelucid dreams with relatively high REM alpha at the beginning of the period and preceding the arousal were most distinct from nonlucid dreams with low REM alpha. Bizarre dream content also discriminated prelucid from nonlucid dreams and was associated with high REM alpha at the beginning of the period. Although empirically distinct, LDs were the most difficult to discriminate with a pattern of relatively high REM alpha at the beginning and moderately low REM alpha preceding the arousal. The amount of clouding discriminated prelucid from lucid dreams, but outside of the lucidity dimension, nonlucid and lucid dreams could not be discriminated with any of the other content dimensions measured in the described study.
In the paper there was also a discussion about the physiological correlates of the alpha activity, but I’ll leave that out here…
So LDs were found to have relatively high REM alpha at the beginning of the period, followed by lower levels near the arousal. In some LDs, the change from relatively high to low alpha was quite abrupt and may cycle over time during long REM periods. The association between high alpha and prelucid, bizarre experiences at the beginning of the REM period supports Celia Green’s (1968) suggestion that lucidity is sometimes preceded by prelucid experiences. Finding LDs in the middle in terms of REM alpha and between the predictable, realistic nonlucid dream content and the unpredictable, bizarre prelucid dreams implies that lucidity might be a combination of more than one process, coexisting simultaneously. This can be compared as the coexistence of two subsystems in the example of driving a car and carrying on a conversation. Lucidity shifts the hierarchy of attention allowing you to drive the car and simultaneously monitor the driving instead of engaging in a variety of other activities such as carrying on a conversation. The allocation of attention to monitoring waxes and wanes as you attend to other activities. Lucid monitoring of driving wanes during extremely predictable and unpredictable driving conditions. Lapses of active monitoring sometimes happen for long periods of time on long trips on well trod routes, and lapses also happen during slippery conditions or stressful emergency situations where total attention capacity is allocated to driving the car (Tyson, 1982). The periodic cycling of lucidity is like a balanced scale and similar to the waxing and waning of alpha activity. In terms of dream content, as the hierarchy of cognitive organizations begins to level, attention becomes more simultaneous, memory more image-like, associations become more random and dreams more bizarre.
Conclusion
This analysis refined the Ogilvie et al. (1978, 1982) alpha-lucidity hypothesis by defining distinct patterns of alpha activity during the course of the REM period for each type of dream. High REM alpha was associated with prelucid dreams and bizarre dream content. LDs were found to have high alpha early in the REM period, followed by a distinct lowering of REM alpha, which supports the hypothesis that lucidity sometimes emerges from prelucid experiences.
@ Jeff (in case you read this ): since you’ve had an EEG for ten years, can you perhaps relate the findings in this study to your own recordings?
Brainwaves don’t cause or control dreams, but they reflect the kind of dreams you’re having. So by closely monitoring the brainwaves, and especially the alpha brainwaves (who normally are not very strong in REM sleep), one can notice whenever a person asleep is having a prelucid, nonlucid or lucid dream. Ofcourse, this is no absolute method and certainly doesn’t work in all cases (in this particular experiment, they succeeded in 71% of the cases), but it’s an intriguing correlation nonetheless.