A word search analysis of a five-year selection from my own dream journal reveals the same consistent patterns of content in both shorter and longer reports.
I’ve been wondering about this question for a long time now. Are shorter dreams different in any fundamental way from longer dreams? Some people naturally remember only brief dream fragments and images, while other people can remember extremely elaborate and detailed dream scenarios. Most researchers prefer to analyze reports in the “Goldilocks zone,” not too short or too long, just right in the middle. That is a reasonable methodological choice, but it still leaves unanswered the question I’ve been pondering.
To continue developing the word search tools of the Sleep and Dream Database (SDDb), I really need to get some clarity on this point. The frequencies of word usage identified by the SDDb tools vary a great deal depending on whether the dreams have a smaller or larger number of total words. Is this a problem, or not?
I also have a personal reason for wanting to explore the question. In early 2015 I began a new approach to my own dream journaling practice, which has led to at least one remembered dream every night for more than two years. This is approximately double my recall rate for the previous several years. The 2015 dreams were also shorter on average (74 words) than the dreams from previous years (all averaging 100+ words). This made me wonder about possible changes in the content patterns of my dreams before and after 2015.
With all of this in mind, I started by tabulating the distribution of my dreams over a five-year period of time (2012-2016), separating them into four categories of word length (less than 50 words, 50-99 words, 100-149 words, and 150 words or more). Here are the totals for each year in the four categories, from shortest to longest:
2012: 41, 67, 50, 43 (201 total)
2013: 68, 83, 51, 50 (252 total)
2014: 51, 54, 40, 41 (186 total)
2015: 145, 120, 56, 31 (352 total)
2016: 91, 134, 64, 77 (366 total)
As I already knew, the increased recall in 2015 happened at the shorter end of the word length spectrum. My new approach to recall seemed to yield a lot of short dreams that I might not have remembered or recorded in previous years. Then the 2016 dreams shifted again, with a more even distribution of word lengths, closer to the previous years but with higher total numbers.
Dividing the dreams into these subsets makes it possible to address the main question: what are the content differences between dreams of different lengths?
For each of the 20 subsets of dreams I used the SDDb 2.0 word search template to determine the frequencies for 40 categories of word usage, organized into 8 classes (Perceptions, Emotions, Characters, Cognitions, Social Interactions, Movement, Culture, and Elements).
The results of this analysis suggest that shorter dreams are not dramatically different from longer dreams in terms of the relative proportions of their word usage. The raw percentages of word usage do rise from shorter to longer dreams, of course, but the relative proportions generally do not.
Consider the following excerpt from the analysis, which shows the four subsets of dreams from 2013, and the results of searching these reports for references to “Perception” words, from shortest to longest reports. The numbers are percentages of the dreams that have at least one reference to the words in the category.
Vision: 24, 54, 69, 76
Hearing: 1, 6, 18, 22
Touch: 1, 6, 16, 26
Smell/Taste: 1, 2, 0, 6
Color: 21, 42, 31, 56
The longer dreams have more references to “Touch” than do the shorter dreams, but the longer dreams also have many more references to “Vision” and “Color” than to “Touch,” which is the same pattern found in the shorter dreams. It’s this kind of pattern—the relative proportions between the various word categories—that remains consistent regardless of the length of the dreams.
This finding suggests the proportions among the word categories do not, for the most part, dramatically change across word lengths. These proportions can be found in short, medium, and long dreams. Even very short dreams preserve the basic architecture of typical dream content.
I need to do a more precise mathematical analysis of these patterns, to illuminate subtler variations that may alter my conclusions. But I’m reassured by these initial results indicating that shorter dreams are just as legitimate as longer dreams for data-driven research and theorizing.
That’s the big picture. Within this portrait of broad consistency, there are a few instances where the longer dreams do have an unusually high frequency of a particular word category. The most prominent are Fear, Speech, Walking/Running, and Transportation. These are the word categories that seem to be over-represented in longer dreams. They are significant contributors to what makes long dreams so long.
Here is an example from the 2016 dreams to illustrate what I mean, using the Emotions class. The numbers are percentages of the dreams that have at least one reference to the words in the category, from shortest to longest reports. Note the dramatic rise in Fear words across the four subsets.
Fear: 3, 11, 36, 55
Anger: 2, 4, 9, 17
Sadness: 4, 2, 8, 5
Wonder/Confusion: 23, 40, 63, 75
Happiness: 11, 22, 20, 26
The shortest dreams have scarcely any references to fear, whereas more than half the longest dreams have a reference to fear.
What I think this means is that when a dream introduces a reference to fear, it heightens my awareness of what’s going on in the dream space. It stimulates an expansion of what I notice and find significant, and after awakening this requires a lengthier report to describe adequately.
What about the unusual increase in Speech references in longer dreams? Perhaps a dream in which people start talking with each other is more likely to deepen the interaction and extend the overall experience.
Same with the increased references to Walking/Running and Transportation: a dream in which people are moving from one place to another is probably going to include additional details about what happens before, during, and after the movement.
So here’s a more refined conclusion: Shorter dreams are mostly similar to longer dreams in their basic content patterns, except that longer dreams tend to be scarier, more mobile, and more conversational.
Looking specifically at the 2015 dreams, I found the word usage frequencies were mostly lower compared to previous years, but they generally stayed the same in terms of their relative proportions to each other. Even though the 2015 dreams were much shorter than the dreams of previous years, they shared with the other dreams a consistent profile of relative frequencies across all the word categories. So my increased recall that year did not significantly alter the content patterns of the dreams.
Finally, I thought it would be fun to try a “blind analysis” of my own dreams. Now that I have identified this remarkably stable profile of my dream content over five years of time, including both short dreams and long dreams, what do the patterns reveal about my life?
If I pretend that these dreams came from a stranger about whom I have no biographical knowledge, I would predict that in waking life this person:
Is visually oriented
Often experiences wonder/confusion
Is sexually active
Cares about his wife
Cares about cats
Has equal relations with men and women
Is not concerned about death
Has lots of interactions with cars and streets
Likes music and movies
Has lots of interactions with water and earth
All of these inferences are grounded in the statistical results of the word searches, and I would have to affirm every one of them as accurate. Indeed, this is a remarkably concise summary of my concerns, interests, and activities in waking life.
Most importantly for the topic of this essay, the content patterns that helped me generate these inferences are observable in the shortest dreams. I would have made most of these same accurate predictions if I had only been looking at the dreams of less than 50 words.
This means the answer to the opening question is no, there is not a significant difference in patterns of content between short and long dreams. Perhaps dreams should be conceived as having a kind of fractal quality: even at a small scale they reflect the same basic structures that shapes things at a larger scale.
I will close by noting the three most striking discontinuities between the word usage frequencies in my dreams and the concerns, interests, and activities of my waking life. These are instances where my blind analysis predictions would have been wrong.
First, I have very few references in my dreams to “Fantastic Beings,” which might lead to the inference that I do not like the cultural genres of science fiction or fantasy. This is not true; I have always loved books, movies, and tv shows in the sci-fi and fantasy realm. Perhaps what I like about these stories are not the odd characters (vampires, zombies, aliens, robots, etc.) but rather the spirit of unpredictable novelty and imaginative adventure. Putting it in those terms, my high frequency of “Wonder/Confusion” words might be a better sign of my cultural interests in this direction.
Second, I have only moderate references to “Reading/Writing,” which might suggest I do not engage much with these activities. This is not true; I am a voracious reader and prolific writer, and have been so for several decades. What strikes me as discontinuous is that my dreams don’t have far more references to reading and writing, given their central importance in my waking life. Ernest Hartmann’s notion that we typically do not dream of the three R’s might be a factor here.
And third, I have very few references to “Religion,” which would prompt the inference that I have little or no concern about religion. At one level this is definitely false; I have a Ph.D. in religious studies and I read and write about religion very frequently. One would never know this about my waking life based only on the patterns of my dreams. And yet, at another level this inference is surely true; I was not raised in a religious household, I do not personally identify with any official religious tradition, and I rarely attend religious worship services. Perhaps this all makes sense in that religion is an important intellectual category for me, but it is not a personal concern. My spiritual pursuits are more likely to be expressed in dreams with references to other word categories like water, art, sexuality, animals, and flying.
That’s as far as I’ve gotten. The next step will be trying this same process of analysis with other sets and series of dreams.
Note: this post was originally published in Psychology Today on May 4, 2017.