Fast Content Analysis Using the Internet

G. William Domhoff

University of California, Santa Cruz


Content analysis is the method of choice when it comes to studying dream reports. The most frequently used system, created by Hall and Van de Castle (1966), has yielded impressive results showing that dream content correlates with age, gender, mental health, personal concerns, and culture. It also has shown that dream content is consistent over time and generally continuous with past or present waking concerns. These findings provide a solid foundation for future research studies (Domhoff, 1996; Domhoff, 1999).

However, as impressive as the Hall/Van de Castle system is, it has not been widely used because it is labor intensive compared to the rating scales that are more frequently employed despite their low reliabilities and unproven validity. It takes time to learn the Hall/Van de Castle coding categories and still more time to apply them to large samples of dream reports. Part of this problem has been solved by DreamSAT, a spreadsheet developed by Adam Schneider that provides fast and accurate results once codings are entered, but this timesaver has not been enough to move researchers away from much less useful rating scales (Schneider & Domhoff, 1995).

The other half of the problem can be mitigated through the development of Internet-based resources that are accessible to all dream researchers. Such on-line resources have several advantages that stem from centralization: they can be used from any location without the need for extra software, only one database needs to be maintained, and any improvements to the program or additions to the database can benefit everyone.

This paper discusses one such Internet resource, DreamBank.net, that is available to everyone (Schneider & Domhoff, 1999). This site contains both a large collection of dream reports and a search engine, written in the Perl programming language, that can be used in at least three ways. First, it can be used to identify subsamples in long dream series. Second, it can be used to do consistency studies for specific characters and other specific elements in a dream series. Third, it can be used to explore possible new categories for doing content analysis with the search engine.

The search engine locates individual words, long strings of words, or phrases by means of Unix regular expressions, which are codes used for pattern matching in computer programming (see Schneider & Domhoff, 1999, for a detailed discussion of regular expressions). Once the pattern is located, the search engine reports the identification number for each dream that contains the requested query, along with the percentage of dreams containing the words. Contingencies between patterns also can be calculated, such as the degree of relationship between the words "house" and "father" (Osgood, 1959).

The dream reports retrieved by the search engine can be viewed in full on the screen, or in an abbreviated form that displays only the sentences with the requested words in them. The requested words are highlighted and in boldface when the dreams are brought up on the screen for viewing. When scrolling through the dreams on the screen, those that are not relevant can be eliminated before the dreams are printed or analyzed on the screen. It is also possible to draw random samples from DreamBank search results. The minimum and maximum number of words per dream in the random sample can be specified, along with the desired sample size.

The site contains over 13,000 dream reports from groups and individuals of all ages. These reports can be used for a wide range of studies, including those relating to figurative thought in dreams. They can be drawn upon to create new sets of dreams on specific topics, such as dreams that contain weddings, bridges, or murders. For those who want to use the search engine to study the dreams reports they have collected, but do not want these reports to be part of the DreamBank, it is possible to arrange for a confidential site. Although every effort has been made to insure that the dream reports on DreamBank.net are accurate and authentic ones, some of the dreams from children and teenagers may be the product of poetic license. The best defense against this possibility is to use the largest sample size that is feasible. Table 1 presents an overview of the dream reports available on DreamBank.net.

Table 1. The dream series available on DreamBank.net as of December 2003
 Gender  Years  N
Alta: a detailed dreamer  female1985-1997422
Angie #1: age 18  female199622
Angie #2: age 20  female199826
Arlie: a middle-aged woman  female1992-1998212
Barb Sanders  female1960-19973116
Barb Sanders #2  female1997-20011138
Barb Sanders #3  female2001-20030
Bay Area girls: Grades 4-6  female1996-1997234
Bay Area girls: Grades 7-9  female1996-1997154
Betty: a college student  female199328
Blind dreamers (F)  femalemid-1990s238
Blind dreamers (M)  malemid-1990s143
Chuck: a physical scientist  male1991-199375
College women, late 1940s  female1946-1950681
Dahlia: Body image issues  female?24
David: teenage dreams  male1990-1999166
Dorothea: 53 years of dreams  female1912-1965900
Edna: a blind woman  female1948-194919
Elizabeth: a woman in her 40s  female1999-2002564
Emma's Husband  male1940-199872
Emma: 48 years of dreams  female1949-19971221
Esther: an adolescent girl  female1998110
German dreams (F)  female1990s397
German dreams (M)  male1990s140
Hall/VdC Norms: Female  female1940s-1950s490
Hall/VdC Norms: Male  male1940s-1950s491
 Gender  Years  N
Jeff: a lucid dreamer  male200087
Joan: a lesbian  femalemid-1980s42
Mack: A poor recaller  malelate 1990s38
Madeline  female1988-20030
Mark: a young boy  male1997-199923
Melissa: a young girl  female1998-200067
Merri: an artist  female1999-2000316
Miami Home/Lab  male1963-1965445
Midwest teenagers (F)  female1998111
Midwest teenagers (M)  male199883
Nancy: Caring & headstrong  female199744
Prospero: a blind man  male1970202
Ringo: from the 1960s  male196416
Robert Bosnak: A dream analyst  male?53
Samantha: in her 20s  female1992-199963
Seventh grade girls  female199669
Swiss children (F)  female1989-1995164
Swiss children (M)  male1989-1995135
The Natural Scientist  male1939234
Tom: An outgoing man  male1990s27
Topwater: A violent man  male1965-1980301
UCSC women, 1996  female199681
Vickie: a 10-year-old girl  female199535
West Coast teenage girls  femalemid-1990s89
TOTAL  13907

Facilitating a Hall/Van de Castle content analysis

The coding of Hall/Van de Castle categories with large dream samples can be aided in several ways that save many hours of reading through dream reports. Simple word searches are most valuable in facilitating the coding of social interactions with specific dream characters. For example, in the 3,116 dreams in the "Barb Sanders" series, there are 99 appearances by one of her brothers and 45 by the other. Her aggressive and friendly social interactions with them were coded at the rate of one or two per minute by entering their names into the search program and scrolling through the dream reports on line.

Word strings are most useful when they are based on phrases or expressions that are found to be frequent in an individual dream series. For example, some dreamers use only a few phrases for sexual interactions, such as "made love" or "had sex." These phrases then can be combined with a standard string of sexual interaction terms created by using "pipes," which are the means for signifying "or" in the language of regular expressions. For example, a string like kiss|hug|intercourse|ma(de|ke|king)_love|ha(d|s|ve)_sex might capture a representative sample of sex dreams in a given dream series. Word strings also can be useful for creating two subsamples with different defining elements that can be compared to each other. For example, this approach is used in the analysis of the Barb Sanders series to compare dreams highlighted by the references to music in them with dreams that feature theatrical performances.

DreamBank also makes it possible in some cases to save time by entering large strings of words that are likely to locate most of the elements that might be coded for a particular Hall/Van de Castle category. Such strings seem practical for codings for food and eating references, emotions, and animals. However, even the longest of strings is likely to miss some dream reports that contain instances of the coding category, so there is a trade-off between efficiency and accuracy that must be taken into account. The loss of accuracy can be gauged in some cases by reading through random samples of the dream reports to determine what percentage of the codeable elements were missed. Moreover, it is still necessary to look at each instance brought to the screen in order to eliminate those that do not meet the coding rules for one reason or another. This latter point can be demonstrated with a brief analysis of the animals that appear in the normative male and female dream sample.

For this demonstration, the dream reports coded by Hall and Van de Castle as containing at least one animal were located by entering the code for animals -- ANI -- into a SearchCodings program available through dreamresearch.net for use with already coded sets of dreams. Once the actual animals in the dreams were written down, this string of animal terms was then entered into the query box for a search of all the reports in the two normative samples. The search produced 8 more male reports and 10 more female reports than were coded for an animal character. This overshot is due to instances such as "soldiers in a fox hole," a discussion of "small words like 'cat,'" and the presence of a "goose-necked lamp." Moreover, there were figurative expressions such as "flying like a bird," "making pigs of themselves," and "acting like maddened animals." Once these cases are removed, the findings are similar to the normative Hall/Van de Castle figures for the percentage of dream reports that have at least one animal. For future studies of animals in dreams, long strings of animal words could be used as a starting point, but they would have to be augmented by a reading of the dream series to be studied.

Using DreamBank.net to do content analysis

The search engine on DreamBank.net makes it possible to carry out all four steps of a content analysis: (1) create carefully defined categories that have high reliability; (2) determine frequencies for each category; (3) transform the raw frequencies into percentages or rates; and (4) compare the findings with norms or a control group. This approach is completely independent of the Hall/Van de Castle coding system. The phrases or strings of words entered into the search program in effect "define" the new categories. Since the same results are guaranteed each time the sample is searched, the categories have perfect reliability. The search engine also does the second step of a content analysis, which is to provide frequency counts. It also creates some of the necessary percentages, but this step has to be supplemented in most studies by separate analyses of the kind presented below.

The fourth step of a content analysis, a comparison with a normative or control group, can be carried out by drawing on other sets or series in the DreamBank. For many studies, the ideal comparative sets may be the dream reports that Hall and Van de Castle used in creating their male and female norms. Using these dream reports as a baseline gives a study some comparability with Hall/Van de Castle findings. However, in some cases it might make more sense to use another series as the comparison group. For example, one or two dream series from adult women might make good comparison groups for studying a new dream series from a woman of about the same age. The comparison samples can be searched for the phrase or word string at the same time the main sample is being searched.

These points can be demonstrated with a study of pet animals in dreams. The categories first determine the frequency with which people mention the two most common house pets, cats and dogs. These findings are then used to create percentage-based indicators that distinguish between "cat people" and "dog people." To begin, the search program is put on the "OR" mode and the box for "contingency analysis" is marked. Next, two strings of words are entered that include regular expressions of various kinds to insure a focused and quality search:

^(cat|kitten|kitty|kittie|feline)s?^ ^(dog|doggy|doggie|puppy|puppies|canine)s?^

Then the dreams of two cat lovers, Alta and Barb Sanders, are selected for searching, along with those of a young girl, Melissa, who loves all animals, and an older woman, Arlie, who has little or no interest in animals. In addition, the dreams used to create the female and male norms for the Hall/Van de Castle system are selected to provide comparison groups. The male normative dream sample is included to see if there is a gender difference in the preference for cats and dogs. The search produces the results that are presented in Table 2

Table 2. Findings on pets with selected dream series on DreamBank.net
     cats    dogs    "pet animal %"    "cat %"  
Female Norms (n=491)9 (1.8%)11 (2.2%)19 (3.9%)45%
Male Norms (n=491)2 (0.4%)11 (2.2%)13 (2.6%)15%
Alta (n=422)55 (13.0%) 21 (5.0%) 67 (15.9%)72%
Arlie (n=212)1 (0.5%) 3 (1.4%) 4 (1.9%)25%
Barb Sanders (n=250)12 (4.8%) 8 (3.2%)18 (7.2%)60%
Melissa (n=67)9 (13.4%)11 (16.4%)18 (26.9%)45%

As can be seen in the table, the results reveal the number and percentage of dreams that contain a mention of at least one cat, at least dog, and at least one dog or cat. The third column from the left, which contains the summary information for the presence of a cat OR dog, provides an indication of a general interest in house pets. It can be tentatively called the "pet animal percent." As expected, Arlie, with only 1.9 percent of her dreams containing a cat or a dog, is below the female norm of 3.9 percent. The other three are above the female norm, suggesting that they are interested in pet animals. There is a small but trivial difference between the male and female normative dreams on the pet percent.

An indication of the degree to which a person dreams about cats as compared to dogs is obtained by dividing the total number of dreams with at least one cat by the total number of dreams with at least one cat or at least one dog. Recalling the assumption that frequency reveals the intensity of "interest" or "concern," this "cat percent" reveals that Alta and Barb Sanders are more concerned with cats than dogs, whereas young Melissa seems to be interested in both cats and dogs. The normative sample for women suggests that Alta and Barb Sanders are well above the average woman on cat percent. The comparison of the female and male norms shows a large difference between women (45%) and men (15%). The results of this simple study, in the context of the continuity principle derived from previous findings with the Hall/Van de Castle system, lead to a substantive prediction: the pet percent and the cat percent will be continuous with waking interests and concerns in future studies using these scales.

More generally, these two scales show the possibility for creating scales for parents, family members, and friends. They would by no means be perfect, but it is easy enough to figure out just how accurate they are by making comparisons with hand tallies. Their justification is that they make possible overview studies of lengthy dream series that would not otherwise be feasible. However, it remains essential to tailor each scale on the basis of a careful examination of the dream series to which it is to be applied.

The development of useful scales that are independent of the Hall/Van de Castle system also can be demonstrated through a "sensory references" coding scale that was created to study 372 dream reports from 15 blind men and women (Hurovitz, 1999). This example also provides an entry point into the issue of identifying possible instances of figurative thought in dreams, because many sensory terms are used metaphorically (e.g, "I see what you mean;" "the taste of victory is sweet;" "that deal smells fishy to me"). In this study, the researchers compared the dreams of the congenitally and adventitiously blind for differences in the percentage of visual, auditory, olfactory, gustatory, and tactile references. All forms of the words "see," "saw," "watch," "look," and "notice" were used to provide a starting point for coding visual imagery. All forms of "hear" and "listen" were used to locate possible auditory imagery. All forms of "taste," "smell," "aroma," "scent," "feel," "felt," and "touch" were used as a starting point for the other three senses. This list of words is based on a careful reading of the dream reports for all sensory terms.

After the dream reports with possible sensory references were located by the search program, two coders studied each boldface word and its context to determine which sensory mentions seemed to be literal and which might be figurative. The distinction between the literal and the figurative use of terms is important in most studies, but it was especially crucial in a study looking for visual imagery in the dreams of blind participants. The sensory references were divided into three categories: visual, auditory, and taste/smell/touch, and the percentage of each type was determined. The findings were analyzed in terms of a "taste/smell/touch percent" because earlier studies showed it to be the distinguishing sensory feature in the dreams of blind people. As expected, the taste/smell/touch percent was very high in the dream reports of the congenitally blind participants. On the other hand, of 36 sensory references in the dreams of two women who lost their sight after age 8, only one was in the taste/smell/touch category. Five of their sensory references were auditory, and the rest (86%) were visual. The findings with the adventitiously blind participants are similar to those for sighted adults in large-scale studies (Snyder, 1970; Zadra, Nielsen, & Donderi, 1998).

The seeming visual exceptions in the dreams of two congenitally blind people turned out to be metaphoric in nature, employing a frequently used conceptual metaphor in which "Knowing" or "Experiencing" is "Seeing" (Matlock, 1988; Matlock & Sweetser, 1989; Sweetser, 1990). For example, , a 52-year-old woman reported a dream in which she and her husband (also blind since birth) visited with Thomas Jefferson himself on a trip to Monticello. She first said that Jefferson "was glad to meet us and he didn't care if we couldn't see," which confirms her lack of vision. However, she then reports that Jefferson took them to "see" the plants in his garden, which seems to be an instance of "Experiencing is Seeing."

Scales aside, DreamBank.net provides a quick and powerful way to study consistency in dreams. Such studies are most clear-cut for the consistency with which a specific character appears, but they can be done for any word or string of words that is entered. The frequencies that are found can be turned into percentages in one of two ways, by dividing the frequency for a given year by the number of dreams for that year, or by using the number identifications for each dream to determine frequency per 50 or 100 dream reports. A example of such consistency studies is provided in the final section.

Studying figurative thought in dreams

The DreamBank search engine, especially when used in conjunction with Hall/Van de Castle categories, provides an entree into the difficult problem of studying figurative thought in dreams. Three brief examples are provided here that are only meant to suggest possibilities. They concern dreams about weddings and dreams about bridges.

Wedding dreams are of potential interest for two reasons. Women have more of them than men do, and experience more mishaps and mistakes in them than men do. In terms of the Hall/Van de Castle coding system, women's wedding dreams have an unusually high number of misfortunes -- the wrong groom, the wrong church, or the wrong dress, for example (Domhoff, 1996). These misfortunes may be a useful starting point for the question of figurative thought in dreams. This is because many of the phrases that are characterized as misfortunes in the Hall/Van de Castle system can be construed as metaphoric -- e.g., "I was lost," "I was overwhelmed by a tidal wave." The possibility therefore arises that wedding dreams might provide a collective metaphoric portrayal of how American women think about weddings.

The first step in such a study would be to search of all the adult dreams on DreamBank.net for the word string wedding|marry|marriage|groom|bride. This provides a frequency for each individual series and set. After irrelevant instances are removed, such a search in fact confirms that women dream more often about weddings than men do. As a second step, random samples of wedding dreams could be created and coded with the Misfortunes scale. Then the findings for men and women could be compared to each other, and to the male and female norms. If the previous finding of more misfortunes in women's wedding dreams is upheld, more detailed and qualitative studies of the specific misfortunes in these dreams could be undertaken.

Dreams about bridges are of potential use because bridges are central to some waking metaphors for personal transitions, such as "cross that bridge when you come to it," "don't burn your bridges," and "that's water under the bridge." Boss (1958), the phenomenological dream theorist who professes to be adamantly opposed to the idea of symbolism, has a detailed discussion of "bridgeness" in rendering what he claims is a non-symbolic interpretation of a dream. If bridges have a metaphoric meaning in dreams, then perhaps they would be more likely to be associated with misfortunes and emotions than they are in waking life, where they are matter-of-fact, easily crossed, and seldom in danger of breaking. So, after eliminating instances of the word "bridge" that designate a card game or dental work, the contingency between bridges and misfortunes, and bridges and apprehension, could be determined. If an earlier study using a small sample is any indication, it is likely that these relationships will be found (Hall & Nordby, 1972, pp. 136-137).

DreamBank's program for contingency analysis also might prove very useful for studying figurative thought in dreams. For example, contingencies could be determined for the word string wedding|marry|marriage|bride|groom and strings of emotion words, or for the word bridge and strings of emotion words.

Using the DreamBank search engine to study the "Emma" series

Many of the points made in this brief overview can be demonstrated with several analyses of the "Emma" series. Emma is a 77-year-old woman who first wrote down her dreams when she was in a Jungian analysis as a young woman. Several years later, now married and the mother of young children, she began writing down her dreams again, with the intent of studying them for symbolism and archetypes at some later point. She eventually lost interest in analyzing the dreams, but continued the journal out of habit. There are 1,221 dreams in all, most of them written down between the ages of 40 and 77. There are no dreams for 1981, l988, and 1989, and only 47 dreams for 1993 to 1997, when he series ends. Her dreams came to the DreamBank because she asked a Jungian author familiar with content analysis if anyone might want them. She wrote down dreams for another three years after she gave her dream journals to the DreamBank.

Reading through the part of the journal that is more or less continuous over 37 years, the most striking impression is the frequency with which her husband and the minister at her church appear. This impression led to a search of the dreams for the number of appearances by various friends and relatives. It confirms that her husband and minister are by far the most frequent characters in the series. Her husband appears in 30 percent of the 1,137 dreams during this time period and her minister in 23 percent. Taken together, they appear in 48.3 percent of those continuous dreams. If frequency is an indicator of "intensity,", as all past studies indicate, then these two men are her greatest concern. Since Emma knew both of them over the entire time period, the question arises as to how consistently she dreamt about them. This study shows that she was more likely to dream about her minister in the 1960s, and more likely to dream about her husband after that time.

A reading of the dreams about the husband and the minister suggests that they do not appear very often in the same dream. This possibility can be studied by using the contingency program that is part of the search engine. Since her husband appears in 30 percent of the dreams and the minister in 23 percent, the expected value for their joint appearances is 6.9 percent (.30 X .23), or 78 dreams. In fact, they appear together only 51 times. This difference has a p value of .02, suggesting that the two men occupy somewhat separate spheres in the dreamer's mind. Although the effect size is not large, this finding fits with the fact that Emma's husband is not religious and does not go to church with her.

The question naturally arises as to the nature of Emma's interaction with the two men, which can be explored by using the Hall/Van de Castle categories for friendly and aggressive interactions. To answer this question, Thomas Van Rompay first used the search engine to draw two random samples of 100 dreams that contained one of these two dream characters. Then the dreams were coded for friendly and aggressive interactions between Emma and either of the two men. The contrast is striking. Her interactions with her husband are aggressive, but her interactions with her minister are friendly. Furthermore, Emma usually initiates the aggressions with her husband, which are mostly angry thoughts, critical comments, and yelling. She and the minister both initiate friendly interactions when they interact. Some of the specific findings from this analysis are presented in Table 3.

Table 3. Emma's social interactions with her husband and minister
(n=100 dream reports for each sample)
     Husband    Minister  
% with one interaction26 52
% with one friendly interaction11 93
% with one aggressive interaction89 7

The findings with the two men are all the more interesting in the context of the findings from a random sample of 100 of her dreams, which Van Rompay had coded earlier for another project (Van Rompay, 2000). The most atypical aspect of her dream life in comparison to the female norms is the high rate of friendliness and the low rate of aggressions, which makes her pattern with her husband even more striking.

Three inferences about Emma's waking thoughts can be drawn from these various analyses of her dream series. First, the dreamer is intensely "concerned" about both her husband and the minister in waking life, as shown by the large number of times they both appear in her dreams. Second, her interest in the minister has declined since a high point when she was in her 30s and 40s. Third, she has more positive interactions with and feelings about her minister than her husband.


This paper has demonstrated some of the possibilities for studying dream reports by using the Hall/Van de Castle system and DreamBank.net in conjunction with each other. It also has shown that DreamBank.net can be used to make analyses that are independent of the Hall/Van de Castle system. In addition, the large number of dreams that are available for searching makes it possible for a wider range of researchers to study dreams without having to develop samples of their own. At the least, the dream reports on DreamBank.net can serve as control groups in studies by other researchers.


Boss, M. (1958). The Analysis of Dreams. New York: Philosophical Library.

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, 115-137.

Hall, C., & Nordby, V. (1972). The individual and his dreams. New York: New American Library.

Hall, C., & Van de Castle, R. (1966). The Content Analysis of Dreams. New York: Appleton-Century-Crofts.

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, 183-193.

Matlock, T. (1988). The metaphorical extension of "see". Paper presented at the Proceedings of the Western Conference on Linguistics.

Matlock, T., & Sweetser, E. (1989). Semantic change of perception verbs to evidentials and mental state verbs. Psychology Department, University of California. Santa Cruz.

Osgood, C. (1959). The representation model and relevant research methods. In I. de Sola Pool (Ed.), Trends in Content Analysis (pp. 54-78). Urbana, IL: University of Illinois Press.

Schneider, A., & Domhoff, G. W. (1995, updated 2004). The Quantitative Study of Dreams [Web site]. Available: http://www.dreamresearch.net/.

Schneider, A., & Domhoff, G. W. (1999, updated 2004). DreamBank [Web site]. Available: http://www.dreambank.net/.

Snyder, F. (1970). The phenomenology of dreaming. In L. Madow & L. Snow (Eds.), The psychodynamic implications of the physiological studies on dreams (pp. 124-151). Springfield, IL: C.C. Thomas.

Sweetser, E. (1990). From etymology to pragmatics. New York: Cambridge University Press.

Van Rompay, R. (2000). Emma's Dreams. Undergraduate Senior Thesis, Psychology, Leiden University, Leiden.

Zadra, A. L., Nielsen, T. A., & Donderi, D. C. (1998). Prevalence of auditory, olfactory and gustatory experiences in home dreams. Perceptual & Motor Skills, 87(3, Pt. 1), 819-826.

Go back to the Dream Library index.
dreamresearch.net home page dreamresearch.net contact info