Comparing Dream Content and Zeo Sleep Data

An advanced feature of the Sleep and Dream Database is the ability to analyze dream content using sleep stage measurements from the Zeo Sleep Manager as search constraints. So far, the SDDb has only one series with both dream reports and Zeo sleep data from the same nights (KB DJ 2009-2010). In coming months I will be pursuing new studies with other participants using a combination of dream journaling and the Zeo device. (If you’re interested in contributing to this research, please let me know!)

Using the word search template of the SDDb, I analyzed 135 dream reports with Zeo data in terms of total REM sleep, total light sleep, total deep sleep, total time awake during the night, and total ZQ (an aggregate number measuring overall sleep quality). For each of these five Zeo variables I divided the 135 reports into three or four subgroups of roughly equal number and average word length, then searched each subgroup to determine its frequency of using the seven word classes and forty word categories available in the SDDb.

At this very early stage of working with dream and Zeo data, my goal is to learn enough to be able to ask more refined questions in future research. The small size of these subgroups (28 the smallest, 52 the largest) means that the statistics are not definitive and surely include a fair amount of noise. The variation in average word length of the reports in each subset (105.53 the shortest, 142.49 the longest) is another reason to view these results cautiously. Some of the reports provide only a brief mention of sexual activity, omitting additional details for privacy reasons.  The KB DJ 2009-2010 series has 182 total dreams, but 47 of the reports do not have corresponding Zeo data.

If patterns in the sleep data do correlate with patterns in dream content, I suspect the effects are likely to appear at the extremes, at the high and low ends of each measurement scale. Unusual frequencies may be nothing more than random noise, but they may also be genuine signals of interaction between sleep physiology and dream content. I’m hoping to identify where these signals might be appearing in data.

The spreadsheet with all the results can be found on Google docs.

Over the next few weeks I’ll post some comments about these data and what I think they mean. For anyone who repeats the SDDb word searches I did on the KB DJ 2009-2010 series and finds an error in my spreadsheet, I’ll send you a free book!



Children’s Dreams: A Word Search Analysis (part 5)

From numbers to narrative: The SDDb makes it easier than ever to combine quantitative and qualitative modes of dream research.  It’s possible to look only at numbers when studying dreams, just as it’s possible to look only at their narrative qualities.  But now that digital archives provide the ability to do both in a variety of creatively coordinated ways, there’s no reason you have to choose one method or the other.  

In fact, the burden is now on single-method researchers to explain why their investigations would not be enriched by the easy integration of other methods.  For those of us who have long struggled to explain and defend the advantages of multidisciplinary research, this is a satisfying turn of the tables.   

Back to the children’s dreams: After using the word searches to highlight some large-scale patterns in this set of 622 dreams, I’m ready to look into the dream narratives themselves. 

Depending on your original question, you may want to start reading a set of dream narratives at the very outset, or you may want to extend the statistical analysis even further than I have up to this point.  Given my initial interests, I have enough statistical information by now to feel comfortable going ahead and reading selected dream reports with a focus on details that relate to special themes I’m studying in terms of “big dreams.”

I start with death, in part because I’m curious what kids are thinking, feeling, and imagining about the end of life. I’ve also found in past studies that dreams relating to death are often connected to bigger religious/spiritual beliefs in the individual’s life.  Guided by that, I often begin reading my way into a set of dreams through the reports using death-related words.

Here are some of the children’s dreams about death that illustrate recurrent themes found elsewhere in the set, along with my initial notes about what might be going on.

“I remember having a dream that my mom died. I couldn’t recall where I was. All I could think about was who was going to take care of us. I felt scared. I don’t remember how my dream ended.” (boy, 17)

Many of the death-related dreams involved a mortal threat to parents or family members.  This surely reflects a primal fear in child psychology.   

“I had a dream a couple of nights ago about my mom dying and I couldn’t save her. It was very hard to understand why I had a dream like that about my mom.” (girl, 9) 

Here is the same theme, with an extra emphasis on the child’s futile efforts to stop death.  The dream pushes her waking mind to consider something it does not understand but can’t help wondering about.

“My mom and dad were in the house with me and there were ghost versions of my mom and dad. The ghost versions of my parents let me play computer games and do whatever I wanted and they were yelling a lot. They shot the real versions of my parents and then my parents died. I cried but then Jesus showed up with me. This was a vivid nightmare i had when I was 8.” (girl, 15)

There’s more bizarreness in this dream, which may reflect metaphorical dimensions of meaning (hard to explore without the ability to dialogue with the dreamer).  She is scared of the death not of her parents but of their disciplined care for her (their superego function?), which is then replaced by the companionship of Jesus.  What does this say about the adolescent psychology of religion?

“One night when I was about 15 years old I had a dream that I was at home and everyone was sleeping when we got a phone call saying that my grandfather had died. The next morning after I had woken up from the dream I went downstairs and my mother was crying. My grandfather had been put into the hospital after a heart attack but luckily he made it through.” (girl, 15)

Strange things happen in families during times of grave illness and death.  Perhaps the girl subliminally heard the phone call while sleeping and incorporated it into her dream, or perhaps her dreaming mind picked up on the emotional stress of her family through means we do not yet understand and wove it into an adaptive preparation for the crisis in waking life. 

“I played with my dog Lita that died 2 years ago. i took her on a walk to our favorite rock and she licked my face. I was so happy. I wished it was real.” (girl, 10)

There are several visitation dreams in the set, some with family members and some with animals.  It’s a dream of happiness and mourning that spurs waking reflection on the relation of wishes and realities.

“It was about my cat Nick he died over a year ago and I dreamt that he came back to life to hang with me and my family this happened about a month ago it felt so real that when I woke up and saw he was not there I was so sad.” (girl, 11)

A similar kind of visitation dream prompts waking feelings in relation to loss of a pet.  Is this type of dream a cruel reminder that would be better ignored, or is it part of the lifelong psychological process of coping when loved ones die?  Could it also be a dawning insight into the existential fact of mortality for all of us, animals and humans alike?

“I had a dream a few days ago that I was in Japan in the 40’s. I was there when one of the bombs dropped from either Heroshima or Nagasoki. I don’t know which one. But I remember seeing the huge mushroom cloud engulf the city the cloud was right in front of me. I didn’t feel afraid. I felt accepting of whatever death was about to come. I was 15 when I had this dream.” (girl, 15)

A spiritually precocious dream in which the girl imagines herself into the iconic scene of nuclear horror that defined the nightmares of the 20th century.  She describes an unusual emotional calm as she accepts the inevitability of death.  It would be very interesting to know more about this girl!  Her experience resonates with the mystical dream traditions of many cultures, where apocalyptic imagery can herald moments of existential insight and self-transcendence.

Next: What can be learned from these findings

Children’s Dreams: A Word Search Analysis (part 4)

To summarize the results of the word search analysis so far:

The dreams of this group of 622 children ages 8-18 have more references to family, animals, and fantastic beings, more happiness and sadness, and more flying as compared to the Hall and Van de Castle “norm dreams” of young adults.

As mentioned earlier, it took less than half an hour to search all the dreams for the 40 word categories.  It would have taken perhaps 100 hours to get equally precise results using what used to be a standard method of dream content analysis (i.e., a team of two people reading the dreams, coding them, then comparing results and determining intercoder reliability).

A word search approach doesn’t eliminate the need for people.  It frees them to devote those extra 99 ½ hours to higher-level analyses and intuitive explorations of patterns in the data.

Now that I’ve got this initial orientation to the set of dreams as a whole, I can look in more detail at particular groups.  First up is gender—what are the male-female patterns in this set of dreams?

Of the 622 participants, 228 were boys and 394 are girls.  It’s a girl-biased sample, which for various reasons is fairly common in dream research.

The boys and girls have similar frequencies on perceptions, with the boys using somewhat more intensity words and the girls having more references to color.  The girls have somewhat more fear, and the boys more happiness.  They are mostly the same on cognition and nature words.  The girls have more family and animal references and more friendliness, while the boys have more physical aggression.  The boys have more references to Christianity, the girls more to death.  (see the table below)

In adulthood, women tend to have higher frequencies than men on many dimensions of dream content and recall, which was mostly the case with this group of children, i.e. the girls were higher than the boys on many categories.

The biggest exception is the higher physical aggression among the boys, which fits with previous studies of gender differences in dream content.  If confirmed by other studies of children’s dreams, this finding would indicate that a gender disparity in aggressive dreaming appears very early in psychological development.

More unexpectedly, the boys had a higher frequency of happiness in their dreams.  If the continuity hypothesis applies here, does it mean boys are happier in general than girls?  I’ll have to look at the dream narratives in more detail to see what that might be about.

Part Five:From numbers to narrative

YQ Males YQ Females
(N=228) (N=394)
Vision 20.6 20.1
Hearing 3.5 4.6
Touch 0.9 3.3
Smell 0 0.5
Taste 0.9 0.5
Intensity 19.7 15.2
Chromatic color 4 5.6
Achromatic color 2.6 5.3
Aesthetic evaluation 11 14.7
Fear 17.1 21.1
Anger 2.6 3.3
Sadness 4.4 4.8
Confusion 1.3 2.8
Happiness 12.3 8.1
Awareness 2.6 4.6
Speech 12.3 16
Imagination 0 0.8
Planning 0.4 1
Choice 4.4 2.8
Effort 1.8 0.5
Reading/writing 0.9 2.3
Weather 2.2 2.3
Fire 4 4.8
Air 5.7 3.3
Water 11 9.4
Earth 6.6 5.3
Flying 6.6 4.1
Falling 5.7 7.4
Family 39 45.4
Animals 18 20.1
Fantastic beings 6.6 5.3
Social Interactions
Friendliness 28.5 37.3
Physical aggression 22.4 16
Sexuality 2.2 1.5
School 16.7 16
Transportation 12.7 11.9
Technology 7 5.1
Money 4.8 2.5
Christianity 5.7 2.3
Death 4.8 7.6

Children’s Dreams: A Word Search Approach (part 3)

In a couple of recent articles (here’s one) I’ve developed and tested a set of 40 word categories to help in studying dreams via word searches. These categories are still a work in progress, but already they’ve proven useful in accurately mapping out some basic content patterns in people’s dreams.

A template with these 40 word categories is built into the SDDb search function, to make it easy to get results that can be compared to other sources.  You can search for other words if you choose, but it’s quicker to use the template if you want to gain an overview of the whole dream series.

It took me 26 minutes to perform the searches and record the results in a separate excel spreadsheet. The searches for “animals,” “friendliness” and “physical aggression” took the longest time to process because these categories include the largest numbers of words.

With more practice the searches could be performed even faster, and eventually the whole process could be automated, but a lot of preliminary work needs to be done before that’s a technological step worth taking.

Now I have a page of statistics showing how often certain categories of words appear in the dreams of this group of 622 children and adolescents.  What can be done with this information?

I’ll start by matching it to the word frequencies of the Hall and Van de Castle “norm dreams,” a set of 490 dreams from 100 college females and 491 dreams from 100 college males, available for study on Bill Domhoff’s website  Although not a perfect sample of human dreaming, the HVDC norm dreams are the most widely used measuring stick for analyzing patterns of dream content.  I’d like to know, at this early stage of the process, how the children’s dreams compare to the HVDC norm dreams on basic features of content.

Ultimately I’d like to develop a better baseline for comparing different dream series.  Until then, I’m using the HVDC norms.

I go into the comparison with a number of expectations:

1. The children’s dreams will have lower frequencies overall, due to the inclusion of shorter dreams (20 minimum vs. 50 minimum in HVDC) and their immature writing skills and cognitive development relative to the HVDC college students.

2. The children’s dreams will have the same “ur-patterns” I’ve found via word searches in many other series of dreams: sight is the most frequent perception, smell and taste the lowest; fear is the most frequent emotion and sadness the least; there’s more aggression than sexuality, more family than animal references, more water than other elements, more falling than flying, and a high overall frequency of speech.  These are over-arching patterns I’ve found in virtually all other series, and I expect they will be present in the children’s dreams, too.

3. Because the HVDC dreams are “most recent” whereas the children’s dreams are “highly memorable,” I anticipate the children’s dreams will have more “primal” qualities such as nightmarish emotions, nature references, physical aggression, family characters, imaginary/fantastic beings, and magical activities (consistent with findings from a study I’ve just done with Ernest Hartmann comparing most recent and most memorable dreams).

I’m going to hold off on any strong conclusions until I’ve divided the children’s dreams by gender and performed another round of word searches.  To start I want a big picture of the whole set, but later I’ll take gender differences into account.

That’s a lot of preamble!  Let’s get to it and look at the first class of word categories in the SDDb, “Perception.”

Harris YQ HVDC Males HVDC Females
(N=622) (N=491) (N=490)
Vision 20.3 37.9 47.1
Hearing 4.2 12.2 12.7
Touch 2.4 6.5 8.4
Smell 0.3 1 0.4
Taste 0.7 1 1.4
Intensity 14.8 34.6 46.3
Chromatic color 5 7.9 17.1
Achromatic color 4.3 7.5 11.4
Aesthetic evaluation 13.3 12.6 20.2

This is pretty much what I expected: lower frequencies overall, but still following the same patterns of relative distribution, with vision highest and smell and taste lowest.  I’m surprised there’s not more color or intensity.

Here’s the second SDDb class of word categories, “Emotion.”

  Harris YQ HVDC Males HVDC Females
(N=622) (N=491) (N=490)
Fear 19.6 16.1 27.8
Anger 3.1 6.7 9.8
Sadness 4.7 2.2 4.9
Confusion 2.3 7.5 10.2
Happiness 9.7 6.1 10.8

Here it gets a little more interesting.  The overall frequencies are lower among the children, but not as much as with the perceptions.  Fear is highest as expected, but sadness is not the lowest.  That’s the most intriguing difference discovered so far—the children’s dreams seem to have relatively more references to sadness.

Next, the SDDb class for “Cognition”:

  Harris YQ HVDC Males HVDC Females
(N=622) (N=491) (N=490)
Awareness 3.9 20 18.8
Speech 14.6 37.1 45.1
Imagination 0.5 2 3.5
Planning 0.8 4.1 5.1
Choice 3.4 5.5 11.4
Effort 1 1.8 1.6
Reading/writing 1.8 6.7 6.7

Hmm, very low use of awareness words among the children.  Speech has the highest frequency for this class, but it’s still much lower than the frequency of speech references in the HVDC norms.

Here are the frequencies for “Nature” words:

  Harris YQ HVDC Males HVDC Females
(N=622) (N=491) (N=490)
Weather 2.3 7.1 5.5
Fire 4.5 5.3 2.9
Air 4.2 3.5 2.4
Water 10 13.8 16.9
Earth 5.8 4.7 6.7
Flying 5 4.5 2.4
Falling 6.8 9.6 7.1

The children’s frequencies seem roughly the same as the HVDC norms, contrary to what I expected about higher nature words usage in especially memorable dreams.  Water is the most frequently mentioned element in children’s dreams, and they have somewhat more falling than flying, though the flying frequency is slightly higher than the HVDC norms.  It will be interesting to look in more detail at the children’s flying dreams.

The results for the SDDb class “Characters” were the most dramatic, as measured by the sudden rise of my eyebrows:

  Harris YQ HVDC Males HVDC Females
(N=622) (N=491) (N=490)
Family 43.1 26.7 39.2
Animals 19.8 11.2 11
Fantastic beings 5.8 0.8 0.6

The children’s dreams, though lower on most other frequencies, are much higher on all three categories of characters.  The relative distributions are the same (family>animals>fantastic beings), consistent with the HVDC norms and most other dream series.  But the children’s dreams have more references to all these character types, which probably reflects several factors: the higher proportion of family and fantastic beings in memorable dreams, the prominence of fantasy in children’s literature, and the many roles, both actual and symbolic, of animals in childhood.  There’s clearly a lot here to study further.

The sixth SDDb class covers three types of “Social Interactions.”

  Harris YQ HVDC Males HVDC Females
(N=622) (N=491) (N=490)
Social Interactions
Friendliness 34.1 37.1 50
Physical aggression 18.3 26.5 13.9
Sexuality 1.8 11 3.7

These results seem to make sense in light of our expectations so far.  The children’s dreams have the same “ur-pattern” or relative distribution (friendliness>physical aggression>sexuality) as the HVDC norms.  The low sexuality is consistent with their young age, and the somewhat high physical aggression may relate to the “primal” features of memorable dreams.  This is a place where I’ll be interested to see the gender frequencies in the children’s dreams, since the HVDC norms suggest that males have much more physical aggression in their dreams than do females.

The last of the SDDb classes is “Culture,” which covers several word categories relating to people’s activities and experiences in the world of culture.  It’s kind of a catch-all class for now.

  Harris YQ HVDC Males HVDC Females
(N=622) (N=491) (N=490)
School 16.2 14.5 24.1
Transportation 12.2 26.9 22.9
Technology 5.8 8.4 7.3
Money 3.4 8.6 7.1
Christianity 3.5 3.7 4.5
Death 6.6 4.9 6.7

All the frequencies trend lower for the children’s dreams, with the exception of death.  That might reflect a higher proportion of nightmares in the children’s dreams.

Part Four:Assessing the results

Children’s Dreams: A Word Search Analysis (part 2)

Once you’re ready to perform a word search analysis—once you’ve formulated a question, chosen a dream series, and acknowledged the limits of your approach—you have to decide the length of the dream reports you’re going to study.

 If you search for reports of any length, your results will include lots of short reports saying “none,” “no dream,” etc.  You’ll also get answers like “dreamed of a whale,” or “plane crash,” reports so short that it’s hard to work with them. You might also get super-long reports with elaborately detailed descriptions and additional waking commentary, which are also hard to work with.

 Unless you specifically want to study the shortest or longest reports, my advice is to set minimum and maximum word lengths for your searches. 

 I started by setting the searches for dreams between 20 and 300 words.  That gave me 622 reports to study.  After I learn more about the series I’ll look at the shorter and longer dreams and find a way to integrate them into the analysis.

 One factor I’m always thinking about is how to make my findings commensurable with those of other researchers. For example, the Hall and Van de Castle content analysis system, which has been used as a base of comparison by many researchers, focuses on dream reports between 50 and 300 words in length. Eventually I’ll look at that narrower range of reports, but at the beginning of my analysis I want to cast a wider net and include more reports in my initial assessment, hence the lower minimum length.

 So, where to start the word searching? 

 My immediate, admittedly vain concern was to know whether I was right or wrong about a recent prediction I made about this dream series.

 A couple weeks ago I wrote a post about people’s dreams of Harry Potter, drawing on results from another survey on highly memorable dreams I commissioned from Harris Interactive.  In that survey 1003 American adults 18 years and older reported dreams between 20 and 300 words in length, and I found two reports using at least one of the following words:

 “harry potter” voldemort hogwarts hagrid dumbledore malfoy snape hermione draco

At the end of that post I predicted there would be more HP-related dreams in the Harris children’s survey, which I had not yet studied.  Now that I have the children’s survey uploaded into the SDDB, I can quickly put that prediction to the test. 

 Of the 622 dreams between 20 and 300 words in the children’s survey, 6 of the dreams used at least one of these HP-related words, 1% of the total vs. 0.2% in the adult survey.

 It’s not an epic difference, but it’s statistically significant (p=.036), and it makes sense in terms of the different roles the HP novels have played in the waking lives of children vs. adults.      

 OK, I wasn’t completely wrong.  That’s the first thing I wanted to settle.

Part Three:Using the pre-set template of 40 word categories

Children’s Dreams: A Word Search Analysis (part 1)

I’ve just begun a new project using word search methods to study dream reports from children and adolescents.  I thought that showing in real time the steps of my analytic process might help other people learn how to apply these methods to their own dream studies.

Any research project starts with a question.  In this case my question was about “big dreams” in childhood (the subject of a book-in-progress).  I wanted to know more about recurrent patterns in the dreams that children and adolescents remember most vividly.  Other researchers like David Foulkes have studied normal, average dream patterns in children, but my question focused on the distinctive features of highly memorable dreams in the early stages of life

Earlier this spring I contacted Harris Interactive, an opinion research company, regarding their “YouthQuery” survey, which enables a researcher to ask a single question and receive online answers from @1000 American children ages 8-18, along with a few other demographic data points.  (The cost of this survey, while considerable, was no more than I’ve paid research assistants to help with other projects in the past.)

There are pros and cons to online surveys.  On the downside, it’s impossible to validate a person’s answers, and it favors participants who are educated and affluent enough to use computers.  On the upside, participants can give their answers in a private setting in their own words, which is extra valuable for a word search approach.

I try not to let excessive angst about methodology slow me down.  Every study has its limits.  Once you’ve identified them, you move on and do the work.  I’m more interested in discovering what a method can do rather than dwelling on what it can’t do.

The Harris people and I decided to word the survey question as follows:

“We are interested in hearing about a dream that you had and remember a lot about.  Please try to tell us everything you remember about the dream, including where you were, who else was there, what happened, how you felt, what you were thinking during the dream and how it ended.  Please also tell us about how old you were when you had the dream.”

The other questions asked in the standard YouthQuery survey regarded age, gender, race/ethnicity, current grade in school, school location (urban, suburban, rural), and school type (public, private, parochial).

Harris conducted the survey in early April, and then I uploaded the info (thanks to Kurt Bollacker) into the sleep and dream database (SDDB).  The dream reports and answers to the other questions can be seen at:

Making this information publicly available enables others to check my work and test my claims, always a good thing in empirical research.  More importantly, it allows other researchers to explore facets of the data beyond what I or any single analyst can pursue.

Dream researchers have operated for too long with isolated sources of data that never receive more than one investigator’s systematic attention.  Digital databases can help our field move forward into a more dynamic and collaborative future.

Next: testing my first predictions