In 2016, I gave a conference presentation in which I analyzed a long series of dreams from an unknown person and made predictions about their waking life.  After I was done, the individual stood up in the audience and graded the accuracy of my inferences.  This was data-driven dream interpretation as a high-wire act without a net. 

The setting was the 33rd annual conference of the International Association for the Study of Dreams, in Kerkrade, the Netherlands. Because I knew the dreamer (pseudonym “Brianna”) was present somewhere in the audience and was going to respond as soon as I finished, I anxiously tried to prepare myself for a very public belly-flop if my analysis turned out to be completely wrong. The text of the presentation is below, which I somehow delivered despite being in a fugue-like state of jet lag.  

A Blind Text Analysis of a Woman’s 2,500 Dreams from 32 Years of Journaling

Deirdre Barrett, Ph.D., Chair

Harvard Medical School

Kelly Bulkeley, Ph.D., Panelist

The Sleep and Dream Database

Presented on June 25, 2016, at the annual conference of the International Association for the Study of Dreams, held at Rolduc Abbey in Kerkrade, the Netherlands.

Introduction:

We are going to share with you today a real-time experiment in dream research, and ask for your input and feedback about the methods being used. 

The experiment involves the “blind” interpretation of an individual’s dream series using a digital method of analysis that focuses on word usage frequencies. This method is blind because it brackets out all external knowledge about the dreamer and his or her personal life. Only the word usage frequencies are factored into the interpretation.

The essence of this approach is its minimalism: it aims to get the most meaning from the least amount of data. It relies on very simple data, just a few objective numbers—objective in the sense that anyone who analyzes the same set of dreams will count the same numbers of words used in the reports.

The method is still in the earliest stages of development, so I can’t promise any particular results. In previous experiments, the mistakes I’ve made have been as illuminating as the things I’ve gotten right, so I’m not seeking perfection. On the contrary, I’m still looking for new ways of applying this approach, so if I’m not making any mistakes I’m probably not pushing it far enough.

To be very clear, this approach does not provide all relevant answers to all questions about dreams. It is very good at answering some questions, and utterly useless at answering others. 

The results of today’s experiment will hopefully take us another baby step forward in developing an approach that has enormous future potentials for bringing dream interpretation more fully into the practice of psychotherapy, medicine, arts education, athletic training, spiritual counseling, and cultural analysis. On Monday there will be a symposium on “mediated dreaming,” and I’ll talk then in more detail about the future potentials of digitally-enhanced dream interpretation, both the pros and the cons. Today, like I said, I’ll talk about the kind of experiment that can lay an empirical foundation for those not-too-distant possibilities.

Procedure:

I was initially contacted by Deirdre Barrett about the possibility of pursuing a new blind analysis of long-term series. All that I knew was the dreamer was female, an avid journal-keeper since 1985, and planning to attend the 2016 IASD conference.  I agreed, and Deirdre sent me the dream reports, which I forwarded to a research assistant for formatting. Given the limitations of time and technology, I decided to look at three subsets of dreams from three separate time periods, not the whole series of 2,500, which would have taken too long to process in time for the conference. This was not really a disadvantage, however, since the three subsets gave us a chance to look for meaningful patterns across three different times in her life. This is what Bill Domhoff and I have done with the Bea series and the Jasmine series, looking at dreams from different periods of time, and the results have been encouraging.

I want to pause for a moment and thank the dreamer, somewhere out in the audience, for her profound dedication to her own dreaming imagination, which is very impressive, and for her willingness to share the treasure of her dream journal with us, which is incredibly generous and so valuable for everyone interested in the empirical study of dreams. It’s an amazing gift, so thanks to the dreamer!

[Note: the following numerical results are contained in an accompanying excel spreadsheet which can be downloaded, titled “Brianna Journals IASD 2016.”]

The first set I looked at was Brianna Journal 2. A total of 100 dream reports comprising the last few from 1998, the whole of 1999, and all of 2000 up until October 11, 2000. This is the start of a 14-month period of no dream reports in her journal, until February 13, 2002. Maybe that gap is significant, maybe not. It makes it a conveniently bounded time period, in any case. It is more or less in the middle of the series, and thus provides a way of testing the method of drawing a subsample from the center of a long series; how much can that tell about the whole? In terms of number, 100 dreams are the minimum Bill Domhoff recommends. I think it can go lower for several categories, but 100 is a good round number, makes percentages very easy. To be honest, I was also curious at the possibility there might be something in the dreams reflecting the cultural atmosphere leading up to the new millennium, the shift from 20th to 21st centuries.  Maybe yes, maybe no.

I uploaded these reports into the Sleep and Dream Database, a digital archive and search engine designed to promote the empirical study of dreams. Once in the SDDb, I applied the 2.0 version of the word search template I’ve been developing, which has 40-word categories covering several areas of dream content: perceptions, characters, emotions, social interactions, and others. I then compared the results with the SDDb female baselines, a macro set of 3,095 “most recent” dream reports from a variety of sources (including the Hall and Van de Castle Norm dreams plus dream collections gathered by Calvin Hall, Stanley Krippner, Tracey Kahan, and myself). The baselines provide a kind of measuring stick for estimating the content frequencies of average, ordinary dreams. 

I never read any of Brianna’s dreams, and never looked at their narrative content. At this stage, I’m trying to focus my initial analysis only on the word usage frequencies, a minimal amount of information about the series

I compared Brianna’s word usage frequencies to the SDDb baselines and to other theories from dream research, especially the continuity principle: the frequency of appearance in dream content reflects emotional importance and concern in waking life. The basic idea with this principle is that the more often something appears in a person’s dreams, the more likely it is something of real concern and emotional importance in the dreamer’s waking life. Continuities between dream content and waking life concerns have been identified especially in these areas:

                  Relationships with family and others

                  Daily activities

                  Emotional temperament

                  Sexual behavior

                  Cultural interests

                  Religious background

There are many methodological challenges in this kind of research, of course. We have to be careful about false positives and false negatives, how to work with metaphors, and several other challenging issues. But the good thing is, this approach can be improved by continued use, testing, and experimentation. We can learn from our mistakes, and do better in future studies.

Brianna 2

I’m going to present the analysis as it unfolded, step by step through the process I used with the tools of the SDDb. 

Immediately I see these are very long dreams for the most part, with an average word length of almost 400 words. The SDDb female baselines have an average word length of 101 words, which means the dreams in the Brianna series are likely to have elevated levels of all word usage categories. That adds an extra level of complexity to the word search analysis.

However, one benefit of working with unusually long dreams like these is that it becomes especially interesting to look at areas of dream content that appear only as often, or even less often, than in the SDDb baselines. This is a different way of identifying a continuity: what does the dreamer have relatively little concern about?

Starting with the comparisons between Brianna’s dreams and the baselines, the perception categories are much higher for Brianna than the baselines, especially touch. Longer dreams seem to have more to perceive. The frequencies for the color category are the same between the baselines and Brianna’s dreams, which means the latter is relatively low given their greater length. 

The emotion categories are all higher for Brianna’s dreams. Fear is especially high—almost half her dreams have some reference to feelings of anxiety.

The character categories are relatively low compared to the baselines, although they follow the same distribution pattern from one category to the other. Brianna’s dreams have more references to males and females than in the baselines, but in the same balanced proportion.

Brianna’s dreams have lots of social interactions, with a “normal” predominance of friendliness, and high levels of physical aggression and sexuality.

The movement word categories are all lower for Brianna, with the exception of death references, which appear in every fifth dream.

The cognition categories are very high in Brianna’s dreams. These words seem to account for a large proportion of her dream reports. The references to reading and writing are especially high.

Several of the culture categories have only modest frequencies in comparison to the baselines. Brianna’s word usage frequencies for architecture, food and drink, clothing, weapons, and sports are even with or below the baseline frequencies. 

Other culture categories show higher proportions in Brianna’s dreams. School and transportation are slightly elevated in Brianna’s dreams. Much higher for Brianna are work and money, technology and science, art, and religion. 

Brianna’s references to natural elements are about even with the baselines, except for an increase in the air category. 

Inferences, round 1

These are my initial predictions about Brianna’s waking life. I tried to derive each inference from some specific detail of word usage in her dreams. I don’t feel equally confident about all of them; some seem almost certain, while others are more of a hunch. In two places I note the possibility that the unusually high frequencies might result from false positives in the use of certain words.

I predict the following about Brianna:

Female (even though I already knew, I definitely would have made this prediction based on elevated frequencies for fear, family, friendliness)

Fear and anxiety are a concern

Closer to mother than father

No husband

No children

Not a dog or cat owner

Not a video game player

A friendly, socially engaged person

Frequently has nightmares of violence and physical aggression

Sexually active

Not a lot of outdoor activities

Death a concern

Cognitively alert, active

Verbal, talkative

Books are important, writing too

Has gone to school, or is in school at university level

Works a lot, has a job at an office (or, false positives with work?)

Likes art, is actively engaged with it

The Jewish religion plays a significant role in her life (or, false positives with god oaths?)

Not interested in sports

Brianna 1

Next, just by virtue of getting them uploaded second, I looked at the first set of dreams, from the beginning of Brianna’s journal on September 9, 1984, to the end of 1988, a total of 292 dreams.

The word search results on these dreams should help to refine the portrait of Brianna that began with the set of 100 dreams. Having more dreams to analyze presumably leads to more precise and accurate inferences.

Evaluating dreams from the same person from different times allows for a new kind of comparison, and a new sphere of inferences: what seems to have changed in the person’s waking life over this time, and what seems to have stayed the same?

To be clear, this involves a different kind of application of the continuity hypothesis: do changes in the frequency of dream content reflect changes in waking life?

To some extent, yes, as Domhoff and I found in our studies of Bea and Jasmine. But some of the chronological inferences in those studies were incorrect or too ambiguous to judge. This is a more tentative area of dream analysis, requiring an extra degree of caution.

The dreams in Brianna’s set 1 are long, but not as long as set 2. Brianna’s dreams lengthened over time.

This prompts the question: are there any categories where set 1 has higher frequencies than set 2, despite having shorter average word lengths?

Perceptions: Higher than the baselines, but lower on vision, touch, and smell & taste compared to set 2.

Emotions: Higher than the baselines, but lower on fear compared to set 2.

Characters: Very consistent with set 2, and comparable to baselines.

Social Interactions: Very consistent with set 2 on friendliness and physical aggression, and much higher on sexuality compared to set 2 and the baselines.

Movement: Mostly consistent with set 2, except for fewer death references in set 1.

Cognition: Consistent with set 2, and much higher than the baselines.

Culture: More references to architecture, weapons, and sports than in set 2, and fewer references to school and religion, although the religion frequency is still high compared to the baselines.

Elements: Very consistent with set 2 (except for air) and with the baselines. 

Now let’s compare the two sets, starting with the biggest differences. 

         The first set has more references to sexuality and architecture, and somewhat more to weapons and sports.

         The second set has more references to vision, touch, smell & taste, fear, wonder & confusion, death, speech, school, and religion.

Along with the differences are numerous similarities between the two sets, suggesting a basic level of consistent content over time. 

Taken together, the comparison of the two sets seems to strengthen most of the inferences, with a few refinements regarding possible changes over time.

Inferences, round 2

Female (frequencies for fear, family, friendliness)

Fear and anxiety are a concern

         especially in set 2

Closer to mother than father

Unlikely to be married

Unlikely to have a child

Not a dog or cat owner

         More contact with cats in set 1

Not a video game player

A friendly, socially engaged person

Frequently has nightmares of violence and physical aggression

Sexually active

         More so in set 1

not a lot of outdoor activities

death a concern

         More so in set 2

cognitively alert, active

verbal, talkative

books are important, writing too

has gone to school, or is in school at university level

         More school later in life?

Works a lot, has a job at an office?  (or, false positives with work?)

likes art, is actively engaged with it

         More music in set 1

The Jewish religion plays a significant role in her life (or, false positives with oaths)?

         More so in set 2

Not interested in sports

         Maybe some tennis in set 1

Brianna 3

Then I uploaded the third set, from 2015 to the present, 145 dream reports total.  Still long, longer than set 1, but shorter than set 2.

Perceptions are consistent across all three sets, with more vision and touch in set 2

Emotions are mostly lower in set 3, especially fear

Character proportions remain consistent, though there’s a drop-off in family references—mother is still top family character—somewhat more references to females?

Social interactions are the same for friendliness but big drop-off in physical aggression, and less sexuality

Movement goes even lower than previous sets—less concern about death in older age?

Cognition remains very high in all three sets

Culture categories down in set 3: architecture, food/drink, school, transportation—but higher in set 3 for technology, and still high for all sets in religion and art, and work

Elements are very low, lowest I’ve ever seen—what to make of that?

Looking now at all three sets of dreams, 537 in total over 30 years, the first thing to note: lots of consistency over time. The range of variation across the three periods of time is quite narrow for many categories. I don’t know how to formulate it in more precise terms, but it does seem clear these three sets of dreams came from the same dreamer.

The second thing to note is some big variations in the three sets, which further qualify some of the inferences, although nothing in the third set seems to require a major change in the inferences initially made based only on the second set of 100 dreams.

Here are the final inferences, round 3

Female (frequencies for fear, family, friendliness)

Fear and anxiety are a concern

         especially in set 2, much less in set 3

Closer to mother than father

Unlikely to be married

Unlikely to have a child

Not a dog or cat owner

         Some contact with cats in set 1

Not a video game player

A friendly, socially engaged person

Frequently has nightmares of violence and physical aggression

         Much less in set 3

Sexually active

         More so in set 1

Not a lot of outdoor activities (movement plus elements)

         Least in set 3

Death a concern

         More so in set 2, less in set 3

Cognitively alert, active

Verbal, talkative

Books are important, writing too

Has gone to school, or is in school at university level

         More school later in set 2

Works a lot  --  has a job at an office?  (or, false positives with work?)

Likes art, is actively engaged with it

         More music in set 1, more poetry in sets 2 and 3

The Jewish religion plays a significant role in her life (or, false positives with god oaths)?

         More so in set 2

Not interested in sports

         Maybe some tennis in set 1

Final comments before discussion

The dreamer also provided a list of characters who appear most often in her dreams. I did not look at this list until after I had done the initial WS 2.0 analysis of the three sets of dreams. Had I done so, I would have seen that Brianna used the pseudonym “Janie” for her daughter. Right away, we find a mistake in my inferences! Had I known to include the name Janie along with the family words category, I would have found almost as many references to her in set 3 as to Brianna’s mother, clearly indicating a high level of personal importance. Also, the seemingly low family frequency in set 3 (32 percent) would rise to the same level as the other sets (41 percent) when adding “Janie” to the search.

The characters list enables at least two further inferences:

Lila became a very important person during the time of set 2

Mark was an important person in set 1 and remained so throughout her life

To go any further with the characters will require having the full set of Brianna’s dreams available. That way the long-term patterns in the relationships can be highlighted more clearly. 

Brianna also told me which dreams she considered her most significant ever, and her worst nightmare ever. Both of the dreams she chose came from different time periods than the ones studied here, so I’m going to leave that aside for now and wait until the entire collection is available before looking at those dreams for central themes in the series as a whole.

Before we open up the discussion and you tell me what you think of all this, I want to thank Deirdre for mediating between me and the dreamer. And once again, all my gratitude and appreciation goes to the dreamer for her willingness to let us explore these fascinating dreams, made possible by her lifelong commitment to heeding the call of her dreaming imagination, which is where it all starts. 

Now I look forward to meeting her, wherever she is!

Post-discussion note

When the dreamer had her chance to respond to my inferences, it turned out that most of them were correct. The one about not having a child was wrong; she does have an adopted daughter in waking life.  The one about being “closer” to her mother than her father was too vaguely phrased for her to accept, so I’d count that one as incorrect, too. She judged the other eighteen inferences as accurate, and she provided some fascinating additional biographical information that I’ve quoted below, in edited form.

Phew, no belly-flop!

Here are the final inferences, round 3:

Female (frequencies for fear, family, friendliness)  Yes

Fear and anxiety are a concern

         especially in set 2, much less in set 3    Yes

Closer to mother than father: here, you may want to consider a word other than "closer." My father died when I was young, so he is not a part of my day-to-day life.  However, he was much, much more of a kindred spirit to me than my mother is, and he is a big part of my inner life. You may not have accurately identified all dreams about him, because I don't always use the word "father", "Daddy," or "Dad”…I wouldn't think of myself as "close" with my mother in the emotional sense.  On the contrary, I tend to dream about her a lot because our relationship tends to be pretty conflictual and laden with unfinished business and pain. At the emotional and spiritual level, I was (and still feel) much closer with my father. So, you might want to conceptualize this not as "closer" with mother than father, but as something like, "mother is part of her daily life more than father" or "mother is on her mind more often than father" or something that's not suggestive of emotional closeness.

Unlikely to be married:  Yes. I've never been married.   

Unlikely to have a child: The phrase "unlikely to have a child" sounds future-oriented, as if it's an unlikely future event, which would not have been true since I always wanted a child and actively tried to conceive naturally before adopting my daughter. So, perhaps "probably doesn't have a child" might be a better phrase.  

Not a dog or cat owner:  True.  the only pet I've ever had was a kitten, for about 6 months, in 1975 when I was 12.  It died and I was so crushed that I never felt able to ever own another pet again.  

         Some contact with cats in set 1

Not a video game player:  True.  I've never played a video game in my life!  Also, btw, I don't own a television.  

A friendly, socially engaged person:  Yes.

Frequently has nightmares of violence and physical aggression

         Much less in set 3:  Yes. [here’s how she suggested I phrase it:] "the dreamer later indicated that between the time periods of Sets 1 and 2, and the time period of Set 3, she'd had a fruitful, meaningful and, by her perception, effective psychotherapy treatment, and she attributed her decline in nightmares, and in anxiety-laden dreams more generally, to the positive impact of that treatment.  Thus, it might be that dreams could serve as an indicator of the progress and effectiveness of a psychotherapeutic treatment."

Sexually active

         More so in set 1:  Yes, alas, those youthful days are over.... :-)

Not a lot of outdoor activities (movement plus elements)

         Least in set 3:  Yes.  By 2005 I was already having pretty bad problems with my knees, and even before that, I've never been very into sports, fitness, or outdoor activities, other than travel, which I do have a passion for.  

Death a concern

         More so in set 2, less in set 3:  Yes. The feedback I've received from my daughter, friends, and others, is that I'm much more preoccupied with death than almost anyone they've known.  I think about death quite often.  If it's less evident in set 3, then I would say that's probably attributable to the psychotherapy. [She went into further personal detail about experiences with people dying, being threatened with death, and her professional work in international human rights organizations for many years in countries struggling with horrific violence.]

Cognitively alert, active:   Yes.  At least, I like to think so!

Verbal, talkative:  Yes.  I'm a writer/poet, and so, very oriented toward the verbal, much more so than I am toward the visual.  

Books are important, writing too:  Yes!  I completed my MFA in creative writing a few years ago.  My apartment was lined with books in every wall, nook, and cranny - I even had books hanging from invisible ceiling shelves!  Writing is one of my passions.  I have been writing poetry for most of my adult life.  I've also had many academic articles and chapters in edited books published.  

Has gone to school, or is in school at university level

         More school later in set 2:  Definitely.  I have several academic degrees.  So, I've immersed myself in universities over and over again.  Also, I've done adjunct teaching at universities a few times per year since 2003.  

Works a lot  --  has a job at an office?  (or, false positives with work?)

   19-- I was completing my Ph.D. dissertation.  Not doing "employment" work.  19-- I was working long hours at a one-year, full-time postdoctoral fellowship position at a university School of Medicine.  19--  I worked for nonprofit agencies, in an office.  19-- I was completing studies in psychology and political violence, so I was doing internships rather than employment type of work.  20-- incredibly long, exhausting hours of work, due to the need to work 2 jobs to survive in the city.  

Likes art, is actively engaged with it

         More music in set 1, more poetry in sets 2 and 3:  Yes.  I am much more drawn to literary arts (poetry, theater) and music than to the visual arts (though I do like film).  I'm less visually oriented than most people.

The Jewish religion plays a significant role in her life (or, false positives with god oaths)?   Yes.  It might be as much Jewish culture/ethnicity/identity/history vs religious practice per se, for example, I sometimes dream about the Holocaust or about Israeli culture, but I think you're close enough here.  

         More so in set 2

Not interested in sports:  True, I have never liked sports, neither as a participant nor an observer.  Someone once brought me to a basketball game and I read a novel for the entire game and scarcely looked up....  that's pretty extreme!  I've never been into tennis, but I am interested in REM sleep and EMDR, so maybe if there was tennis in my dream it was related to the back-and-forth eye movements?  

         Maybe some tennis in set 

You are reading

Dreaming in the Digital Age

The Science of Dreaming: 9 Key Points

The essential scientific findings about dreaming everyone should know.

Analyzing a Long Series of Dreams: A Conference Presentation

The "blind analysis" of a dream series illuminates personal concerns in waking.

1,001 Nights of Dream Recall

Exploring the meaningful patterns in a long series of dreams.