Doing Research Projects on Dream Content
It is very difficult to do experimental studies relating to dream content and the meaning of dreams for a variety of reasons, all of which are in complete contrast with what is possible in the waking state. It is not possible to make dreams happen, and it is very difficult to influence them with presleep or concurrent stimuli, so there is no independent variable to manipulate. What past experimental studies show most of all is that the results are trivial and that experiments are hard to do (Arkin & Antrobus, 1991; Foulkes, 1996).
In addition, and rather obviously, you can't watch dreams while they are happening, and the dreamer can't report them while they are happening, so there is no way to have any objective evidence concerning the dependent variable (i.e., the dream). (The study of sleep, we should add, is a completely different story; experimental studies are the method of choice in sleep research.)
The net result is that the best way to study dream meaning is through content analysis. Two of the papers in the research library on this site (Domhoff, 1999; Domhoff, 2000) provide up-to-date information on this method, as do various pages written especially for the site. For a good idea of what kind of results can be obtained, see the pages on this site on Interesting Findings and also the papers in the research library on drugs and dreaming (Kirschner, 1999), children and dreaming (Avila-White, 1999), and the dreams of the blind (Hurovitz, 1999). The site provides everything that is needed to do a high-quality, publishable study, including a spreadsheet that provides the results in table and graph form, along with p values, confidence intervals, and effect sizes. In addition, access to various randomization statistics can be provided once you have entered your codings into the spreadsheet.
The following paragraphs walk through the coding system in very simple language.
Can we be scientific about dreams?
Yes. We can do a "content analysis" of the dream reports that are written down for us, or that you wrote into a dream journal. That is, we have to classify the people, actions, and objects in dreams into categories -- categories like "characters" (friends, family, animals), or "social interactions" (friendliness and aggressions), "striving" (success and failure), and "emotions" (happy, angry, sad, confused, apprehensive).
Once we've sorted all the elements that appear in the dream reports into our various categories -- which takes a great deal of time and patience -- then many interesting analyses can be done.
Here are just a few examples:
By applying the methods described on this web site, you can answer any of these questions, and probably most of the questions that have popped into your mind as you read the examples.
- How do the dreams of teenagers differ from those of college students?
- Is there more apprehension and confusion in dreams of students right before exam week than there is right before a week of vacation?
- How do the dreams that you wrote down in high school differ from the ones you wrote down (or could write down) as an adult?
- How do the 100 dreams that someone you know wrote down in a diary differ from the typical dream content for people his or her age?
But first, how do you obtain the dreams for making the analysis?
The four ways you can obtain usable dream reports
- You can use our Most Recent Dream form to collect the last dream students in a classroom remember having, whether it was this morning, last week, or last month.
- You can advertise in a student paper, local newspaper, or on bulletin boards in bookstores for dream journals that already have been kept for some months or years.
- If you are a prolific dream recaller, or have a friend who is, you can keep a dream diary for a couple of months, or have him or her keep a dream journal.
- You can go to dreambank.net, which has dreams from invididuals and groups, from childhood to old age.
Each method has its advantages and disadvantages. Next, we will say a little more about the first three methods.
Most Recent Dreams
If you have access to homerooms, classrooms, or a student assembly, and have permission to take 20 to 30 minutes of precious class time, then the Most Recent Dream technique is a powerful and efficient method because it allows you to obtain large samples that include all kinds of people (good recallers of dreams and poor recallers alike) in a relatively short period of time.
With this method, you don't have to wait around hoping that at least some of the students you've asked will keep a dream diary at home for a week or two, and then give it to you. Experience shows that most don't. Those who do often write short and hasty reports. Worse, critics can say that those who turned in diaries may differ from those who did not on some important personality or memory dimensions. That is, they would raise questions about the "representativeness" of your sample. The beauty of the Most Recent Dream method is that just about everyone can remember at least one dream from the past six months, so it gives you a "representative" sample of dreams from a "representative" sample of subjects.
We also know how many Most Recent Dream reports you need for a good study -- at least 100, but the more the better. We know this from drawing "subsamples" of different sizes from very large samples of dream reports where we know the overall findings, then determining how close their results approximated the overall results. Twenty-five or 50 or even 75 dreams from a group is not good enough. The results aren't reliable. To be serious, you need at least 100 to 125 from each group you are studying (Domhoff, 1996). But the truth is that several hundred from each group is even more powerful, and there are ways to do analyses of that many dreams that don't take as much time as you might think.
In collecting Most Recent Dreams, you should emphasize that it is important to follow the instructions closely. The instructions ask for the participants to write down the date they think they had the dream and the time of day they remembered it. Asking for the date and time has two rationales:
- It focuses the participants' attention on the fact you want the "last dream they can remember." If they don't pay attention and write down their most memorable or frightening dream, then the sample is biased toward unusual dreams.
- It allows you to discard dream reports where the date is more than six months prior to the date you collect the dreams. That helps to assure a representative sample of typical dreams.
Before you hand out the forms, you should emphasize that participation is voluntary, and that they should write that they "don't recall any dreams" if that is the case or that "they don't want to participate" if that is the case. If participants feel they are forced to participate, they may make up dreams.
Once you've collected the reports, there are several things you can do to improve the quality of the data for your study even before you start making analyses:
First of all, you can eliminate those reports where the person says something like "this is a dream I have over and over" or "this is my worst nightmare." Second, you can eliminate any reports that are just one or two sentences. We often use a 50-word minimum as a cut-off point. Third, you can eliminate any reports where the person didn't really write down a dream, which you will be fairly good at spotting after you have read through all the dream reports. If you have concerns that very many are made up, then you can contact us and we will send you a copy of our experimental "Made-Up Dream Detection Scale." We developed it by comparing real dreams from college students with dreams we asked them to make up.
For the most part, though, almost all participants will be cooperative and honest and helpful. Consider it a successful session if 80-90% give you usable reports.
Advertising for a dream journal
You'd be surprised how many people have written down their dreams for at least a month or two at one time or another. Not millions, but maybe two or three percent of any group, which is more than enough to obtain some very useful dream data if you put up signs around schools, colleges, or bookstores saying you are looking for a dream journal to study, or if you place a little ad in a campus or local newspaper. There may even be groups of people in your area who meet to discuss dreams and use them as inspiration for art or creative writing. They usually aren't very scientifically oriented themselves, but they might be willing to let you use their dream diaries for scientific purposes.
For these kinds of studies -- where you are looking at a single person's dreams -- you also need to have at least 100 dream reports. Once again, we know this from drawing subsamples of varying sizes from much longer dream diaries where we know the overall percentages and rates for our various categories.
People write their dreams into a journal for many different reasons -- to use as artistic raw material; for personal enrichment; because they think dreams predict the future; or as part of their therapy with a psychologist influenced by the famous Swiss psychiatrist, Carl Jung, who put a great emphasis on the usefulness of dreams in his approach to helping people. That may not sound like a very "representative" sample of people, but dream journals are legitimate data in the social sciences even though they are not from a representative sample of people. Basically, that's because people write their dreams down for many different reasons and the we analyze those dreams for our own reasons. That means the journals may have all different kinds of "biases," but are not influenced by our research purposes. They are therefore called "nonreactive archival data." For useful information on why such archival data is legitimate for scientific studies, here are three good sources: Allport, 1942; Baldwin, 1942; Webb, Campbell, Schwartz, Sechrest, & Grove, 1981.
For the most interesting kind of study with someone else's dream journal, it is best that you know little or nothing about the person before you study the dream reports. We call that a "blind analysis." It is the closest we can come to approximating an "experiment" with archival data. If you know nothing about the dreamer, then you can formulate several inferences about him or her on the basis of your analysis, and then have the person tell you if he/she thinks the inferences are right or wrong. Can you tell if the person loves animals? Likes his older brother better than his younger brother? Is afraid of authority figures? Feels a lot of sadness? Several studies of this kind have shown that we can indeed make some powerful inferences.
If you already know a great deal about the person who gave you the dream journal, all is not lost. In this case you simply formulate some hypotheses about what you are likely to find in the dreams. You formulate these hypotheses on the basis of what we call the "continuity hypothesis." That is, previous studies suggest that people dream about what they think about and worry about and like in waking life. If you know they talk all the time about basketball or dancing, then it is likely that they dream about basketball or dancing. If you know they still miss their cat or dog that died last year, they probably are still dreaming about it. If they like the outdoors, then the settings for their dreams will more often be outdoors. And so on.
Keeping your own journal
Your own dream journal can be a very valid source of dream reports because our method is objective enough that your personal biases will not cloud your analysis in the way they might if you used a subjective method. Furthermore, you could have someone else do an analysis of some of the dream reports to see if you are using the content analysis system well with your own dreams.
A study of your own dreams can be especially good if (1) you wrote down dreams once before in your life, or (2) you have been writing them down for a long time and have hundreds.
If you decide to keep a dream journal, just write them down each morning and don't even look at them until you have finished the collection period. If you can type them or put them on a computer disk, all the better, because then you can analyze them by doing word searches with the program we have for such a task. That is, we can provide you with your own private space on dreambank.net so that you can make a detailed study of your dreams.
Remember, though, you need at least 100 dreams in your dream journal for your study to be based on what we are fairly sure is a "representative" sample of your dream life. Be sure to write down every dream. If you are selective for seemingly "interesting" ones, then you won't have a sample that can be compared with our normative findings.
What will you find if you study an individual dream journal, whether yours or someone else's? Well, we have some results in our "Findings" section. For example: Lucile, an older woman who wrote down her dreams for several years. Or "Mark," the college student. Or the "Engine Man," so named because he loved train locomotives; he wrote down his dreams faithfully for three months back in 1939, and then never did another thing with them -- but a psychiatrist who studies dreams found them at an antiquarian bookstore in the 1980s and saved them for posterity. His dreams can be found in dreambank.net under the title "The Natural Scientist."
How to make sense out of the dreams you collect
Once you have collected a number of dreams, your next step is to do a data analysis. First, you read the dreams and mark certain people, things and events which are "codeable." Second, you add up frequencies in the content categories you are using. Third, you do various arithmetic and statistical procedures that are fairly simple. Fourth and finally, you compare the results with our typical findings for either young men or young women, called "norms."
All of the counting and statistics can be performed by the DreamSAT spreadsheet that we have available for download in our site; thanks to DreamSAT, the whole process is automated except for the original coding of the dream narratives.
Most of the statistics we use are based on percentages and ratios. As we explain more fully in our page entitled "Our statistical approach", we use percentages and rates because they are the best and simplest way to correct for the fact that the dream reports you will receive will be of differing lengths. Since longer reports are more likely to have more of everything, we need to take that into account, and percentages and rates do so in a clear and explainable way.
To see what we mean, take a look at the table and figure to the right. They both compare findings with 12- and 13-year-old girls to the norms we've developed on the dreams of young adult women. We show here only a few of the categories you can use, but this keeps it simple for now.
Start with the "Friends percent" under the general heading "Characters." Note that we expect the friends percent for women to be 37%, which means that 37% of all the human characters in women's dreams are people they know. Now look at the bar graph, or "h-profile." We arbitrarily call it the "h-profile" because the length of the bars is determined by "h" points, which are about twice as large as a percentage difference. We have to make this mathematical transformation of percentages into "h's" for complicated statistical reasons, which are explained briefly in our statistics page; you can look at if you are mathematically inclined, but for now it is enough to know that "h" is about twice as large as a simple percentage difference.
When you look at the h-profile by "Friends percent," you notice that the bar goes out to the left, which means that girls have a lower friends percent than women. How much lower? Well, only six "h" points, which you recall means about three percentage points. It's not a very big difference. We are not impressed with a difference until it is at least .20 to .30, and we don't become excited until the difference reaches .40 or higher.
Now take a look at some of the other indicators and findings for "Characters." Whether we look at the percentage and "h" differences under "animal percent," "male/female percent," or "family percent" in the table, or at the length of the bars in the h-profile, we see that girls and women do not differ very much as far as the characters in their dreams.
But look at the various categories under "Social Interactions." For the thoughts and actions that we define as "aggressions," girls are higher than women, especially on the subset that are called "physical aggressions," like stealing, chasing, hitting, and killing.
The rest of our findings on how 12- and 13-year-old girls differ from young women can be found in a research paper that we've put on this web site as one example of the kind of study you can do.
Where do you go from here?
Whether studying Most Recent Dreams or a Dream Journal, at some point you need to look at the coding system. It can look daunting at first glance, but it really isn't that bad.
For example, with characters there are basically three general categories: people, animals, and creatures (although we don't record many creatures after early childhood).
We break "people" down according to four different categorization schemes:
Then each character has a "code" -- a number or letter for each of the four categories. So your father is 1MFA: 1 for individual, M for male, F for father, A for adult. And your 9-year-old twin sisters are 2FTC: 2 for group, F for female, T for sister, C for child.
- individual (1) or group (2)
- male (M) or female (F)
- role/relation to you (e.g., family, friends, strangers)
- age (B for baby, C for child, T for teenager, A for adult).
Social interactions aren't very difficult to code either. An aggression is one of the following:
When we do an analysis, we count up the total number of aggressions (A1 through A8) and divide by the total number of characters in the dreams to determine a "rate" of aggressions per character. We call it the "A/C index."
- angry thoughts (A1)
- a nasty or critical remark (A2)
- a rejection or refusal (A3)
- a serious verbal threat (A4)
- stealing or destroying a possession (A5)
- chasing or capturing (A6)
- physical attack (A7)
- murder (A8)
We also total up the "physical" aggressions (A5 through A8) and divide them by the total number of aggressions; that gives us the "physical aggression percent."
Those are just examples, but it doesn't become any more difficult. The problem is not how hard it all is, but how much of it there is when you do a good job on five or six general categories. Here's the ones we use most frequently and recommend to you:
- Social interactions (aggression, friendliness, sexuality)
- Misfortunes and good fortunes
- Striving (success and failure)
- Settings (indoor/outdoor, familiar/unfamiliar)
- Emotions (happiness, anger, sadness, confusion, apprehension)
We have examples of already-coded dreams on this site that help you to learn the system and then see how close you come to matching the codings by seasoned coders, including the originators of the system.
For even more details about the topics that were touched on in this page, we highly recommend that you check out our Resources for Scientists page; there, you'll find links to all of the more "advanced" sections of this Web site.
- Allport, G. (1942). The use of personal documents in psychological science. New York: Social Science Research Council.
- Arkin, A., & Antrobus, J. (1991). The effects of external stimuli applied prior to and during sleep on sleep experience. In S. Ellman, & Antrobus, J. (Ed.), The mind in sleep: Psychology and psychophysiology (Second ed., pp. 265-307). New York: Wiley & Sons.
- Avila-White, D., Schneider, A., & Domhoff, G. W. (1999). The most recent dreams of 12-13 year-old boys and girls: A methodological contribution to the study of dream content in teenagers. Dreaming, 9(2/3), 163-171.
- Baldwin, A. (1942). Personal structure analysis: A statistical method for investigating the single personality. Journal of Abnormal and Social Psychology, 37, 163-183.
- Domhoff, G. W. (1996). Finding Meaning in Dreams: A Quantitative Approach. New York: Plenum Publishing Co.
- Domhoff, G. W. (1999). New directions in the study of dream content using the Hall and Van de Castle coding system. Dreaming, 9(2/3), 115-137.
- Domhoff, G. W. (2000). Methods and measures for the study of dream content. In M. Kryger, T. Roth, & W. Dement (Eds.), Principles and Practices of Sleep Medicine (3rd ed., pp. 463-471). Philadelphia: W. B. Saunders.
- Foulkes, D. (1996). Dream research: 1953-1993. Sleep, 19, 609-624.
- Hurovitz, C., Dunn, S., Domhoff, G., & Fiss, H. (1999). The dreams of blind men and women: A replication and extension of previous findings. Dreaming, 9(2/3), 183-193.
- Kirschner, N. (1999). Medication and dreams: Changes in dream content after drug treatment. Dreaming, 9(2/3), 195-200.
- Webb, E., Campbell, D., Schwartz, R., Sechrest, L., & Grove, J. (1981). Nonreactive Measures in the Social Sciences (2nd ed.). Chicago: Rand McNally.