The categories A5 through A8 can be summed into a "physical aggressions" category. Similarly, categories A1 through A4 can be added up to create a "nonphysical aggressions" category. Finally, the sum total of the aggressions in all eight categories can be treated as a general category of aggressions.
There are many do's and don'ts in the coding of aggression, or any of the other social interactions, for that matter. These specifics are reserved for Appendix A. There is also a set of coding notations that will be saved for the appendix except for a few brief examples. D A2 > 1MFA says that the dreamer (D) swore at or was critical of (A2) his father (1MFA). By the same token, 2MSA A5 > D says a group of males unknown to the dreamer destroyed or stole one of the dreamer's possessions. We use these shorthand summaries of all the aggressions in a dream set or series to answer such questions as the frequency of the dreamer's involvement in aggressions, the frequencies of different types of aggressions, and the frequency of victimization and reciprocation.
When the scoring of a dream report is done on the kind of scorecard shown in Appendix B, which has separate columns for entering aggressive, friendly, and sexual interactions, there is no need to include the A in the scoring of an aggressive interaction. However, if the scoring is being done in the margins of the dream report, which is more feasible in a day when extra copies can be made easily through photocopying, then the interaction should be scored as D A2 > 1MFA so that it can be readily distinguished from the F2's and S2's to be introduced shortly. We now do most of our coding on photocopies of dream reports to improve ease of checking.
The coding system for friendliness closely parallels that for aggressions. First, friendliness is defined as a deliberate purposeful act, only this time involving support, help, kindness, gift-giving, or any other type of friendly act toward another character. Second, there is a distinction between involvement and mere witnessing of the friendly interaction. Third, there are categories for the initiator (the befriender) and recipient (befriended) of the friendly interaction, as well as for reciprocated, mutual, and self-directed friendliness. Fourth, there are several subclasses of friendliness that are defined in Appendix A.
The subclasses of friendship usually are brought together as one overall category of friendliness for purposes of analysis. The notational system is similar to that for aggressions. For example, D F4 > 1ANI means the dreamer helped an animal. D F6 = 1MKA means the dreamer and a known male exchanged slaps on the back or maybe an embrace -- the friendliness was mutual. 1FMA F6 > 1FTA means the dreamer witnessed her mother hugging her sister. The details of how to code for all the intricacies of friendly interactions are once again left for the appendix.
Friendly interactions are usually not as frequent as aggressive ones in dream reports, but they are nonetheless a very useful coding category, especially when they are joined with aggressions for various kinds of comparisons.
Sexual interactions involve everything from sexual fantasies about a character to sexual intercourse. As with aggressive and friendly interactions, a distinction is made between initiators and recipients, and there are categories for reciprocated and mutual sexual interactions. The dreamer can be involved or a witness, and there is a category for self-directed sexuality. There are five subclasses of sexual interactions.
The five subclasses of sexual interaction can be added together for one overall "sexual interaction" score, which often is a necessity because there are relatively few sexual interactions in most series or sets of dream reports. Given this tendency to use one overall score, any difficulties in distinguishing among subclasses become a minor matter. The notational system for sexual interactions is the same as for aggressive and friendly interactions. The following is a steamy sexual scene that also shows how we code a sequence of social interactions:
The Classification and Scoring of Activities
Activities are defined as anything characters do in dreams, such as run, walk, talk, or think. Activities can be done by one character acting alone (e.g., thinking), in conjunction with other characters (e.g., laughing, jogging), or in interactions between characters (e.g., talking). Although most activities are not social interactions by the definitions used in this coding system, activities and social interactions are not mutually exclusive. For example, an aggression can also be a physical activity (e.g., hitting) or a verbal activity (e.g., scolding). There are eight subclasses of activities. The activities categories can be reduced to a physical activities category and a nonphysical activities category. One overall activities score can be derived by adding the frequencies for all eight categories.
The Classification and Scoring of Striving: Success and Failure
Some social interactions and activities involve striving to succeed. The character can "try" to think of a solution to a problem, "try" to fix a fence, or "try" to run from a dangerous character. That is, once a character is described as expending energy and showing perseverance in pursuit of a goal, no matter how trivial the goal is, then the character is involved in striving. If the goal is reached, a "success" is coded. If the goal is not reached, a "failure" is coded. Striving can involve an individual or a group, it can be physical or nonphysical, and it can or cannot be part of a social interaction.
Provision is also made in the coding system for including any aftermath or "consequences" of success or failure. For example, a failure might be followed by a helping hand from another character, which is a friendly interaction as a "consequence" of failure. In practice, we find very few consequences of success or failure.
The Classification and Scoring of Misfortunes and Good Fortunes
In the preceding three sections, emphasis has been placed upon the various interactions and activities of the characters. They may fight, dance, make love, converse, walk around, look, listen, or struggle to accomplish something. All of these acts involve some deliberate voluntary choice on the part of the character engaging in them. As the result of these acts, characters may be killed, hurt, or defeated, or they may become engaged, popular, or prosperous. These bad and good outcomes are, therefore, the consequences of what the characters have done or attempted to do.
Sometimes bad or good outcomes happen to a character independent of anything he or she may have done. Fate, in a sense, has stepped in and produced certain results over which no character has any control. In the Hall/Van de Castle coding system, these impersonal "fatalistic" events are called "misfortunes" when bad things happen to a character and "good fortunes" when good things happen to a character. A misfortune is any mishap, adversity, danger or threat happening to characters through no fault of their own and without intent of aggression on the part of some other character. The six subclasses of misfortune range from death, accidents, and illnesses to being lost or late. A good fortune is something positive, like finding money or winning the lottery, that happens "out of the blue," by luck. No one "caused" the happy event, nor is it the result of striving by the character. Since good fortunes are relatively rare in dream reports, no attempt was made to create any subclasses for them.
The Classification and Scoring of Emotions
Emotions are defined as any feeling states explicitly stated in the report as experienced by a dream character. The emphasis has to be on "explicitly stated" because there is a great temptation to infer unexpressed feelings when dreamers describe emotion-arousing events such as falling, being chased, or facing great danger. The only exception to the explicitness rule is if the dreamer describes the kind of autonomic nervous system activity accompanying an emotion in that situation, such as tears upon hearing news of a death (sadness), or sweating and trembling when cornered by a dangerous animal (apprehension).
Developing a coding system for emotions was an extremely difficult task because there are so many different words for the different affective states, and these affective states can shade one into the other. There also was the problem of whether or not to code for varying intensity in emotions. After trying various coding classifications, Hall and Van de Castle learned that they could only achieve good intercoder reliability if they limited the emotional states to five in number and made no attempt to discriminate levels of intensity. However, there is a separate coding system for all types of intensity that is part of the descriptive elements category. The five categories for the emotions are anger, apprehension, sadness, confusion, and happiness.
It is surprising how seldom emotional states are mentioned in dream reports unless they are asked for explicitly, and even then they are not always present. Moreover, most of the emotions fall into one of the four categories of "negative" emotions (anger, apprehension, sadness, and confusion). Pleasant emotional states are so few that there is little need to distinguish between types of happiness. Thus, the five categories turn out to be quite adequate for the limited role they are called upon to play in the quantitative study of dream content.
The Classification and Scoring of Physical Surroundings: Settings and Objects
The characters in a dream report do not act, interact, emote, strive, and meet their fate in a vacuum. The dream report usually contains physical surroundings that are divided into two very general categories in the Hall/Van de Castle system: settings and objects.
Almost all dream reports include some form of recognizable setting, and people frequently begin their reports by saying something about the setting. Just as there are often several acts and scenes to a play, so, too, is it common for the setting to change during the course of a dream narrative, sometimes quite abruptly.
Establishing the categories for settings was the most difficult aspect of the entire coding system. The initial efforts to classify settings included a rather extensive number of possible settings. However, it proved impossible to obtain adequate intercoder reliability when such a large number were involved, so Hall and Van de Castle eventually collapsed all settings into two broad groupings--indoor and outdoor settings. Indoor settings are ones in which the dreamer is in a building or in an area attached to or part of the exterior of a building. Outdoor settings are those where the dreamer is described as being out-of-doors or outside a building, even if in a vehicle or a cave. If the setting cannot be determined with certainly, it is coded as ambiguous. Settings also are coded for their familiarity to the dreamer. There are five levels of familiarity described in the appendix.
In order to provide a more detailed picture of the physical environments dreamers create for the enactment of their nightly dramatic productions, attention must also be paid to the various "props" that are on the initial stage or introduced as the play proceeds. These "props" are classified under the heading of objects.
An object is a "thing" having tangibility, palpability, dimensionality, and definite physical boundaries or limits. Intangibles such as air, wind, fog, and sky are excluded by such considerations, as are songs or sounds, which have temporal boundaries but not physical ones. Locations such as cities, streets, rooms, and lakes have physical boundaries and are consequently classified as objects. Persons and animals are not coded as objects because they are handled separately under the classification of characters, but parts of persons and animals are treated as objects.
Since any object we encounter in waking life can be represented in dreams, along with some we would be startled to see with our eyes open, the problem of formulating a system for the classification of objects was a tedious one. The number of possible groupings could be very large if one chose to categorize by reference to size, shape, color, weight, age, composition, ownership, location, function, and other qualities that could readily be suggested. After several arrangements had been tried, Hall and Van de Castle finally settled on a system that includes twelve broad classes, three of which are further subdivided, plus a miscellaneous class. All objects appearing in dreams are therefore classifiable under one of these headings. They are as wide as "nature," "region," and "travel," and as narrow as "food," "body parts," "clothing," and "money." These highly detailed categories are not regularly used in every investigation, but some of them have proven to be helpful in specific instances.
The Classification and Scoring of Descriptive Elements
In addition to characters, actions, interactions, and emotions in a physical surrounding of settings and objects, dream reports also contain a very wide range of descriptive terms characterizing people and things. Dreamers are in "fast" cars, see "young" men, feel "intense" anger, and perform "gracefully." Dream reports also may refer to a particular time of the day or note that a certain amount of time has passed. They may use negative terms in order to describe what people or things are not (e.g., "not a big car, but a small one"). All of these are called "descriptive elements" in the Hall/Van de Castle coding system, and they are broken into three general categories--the modifier scale, the temporal scale, and the negative scale.
The Modifier Scale
The modifier scale consists of adjectives, adverbs, and phrases used for descriptive elaboration. There are nine subclasses. Each of the nine subclasses ranges along a bipolar dimension, so can be coded with a plus or minus sign to indicate which pole of the modifier is present in the dream report. "Size," "velocity," and "intensity" are examples of the modifier subclasses.
This scale is for (1) specific units of time, such as "a few minutes," "a week," "a long time," or even "the night shift," or (2) references to a particular time for the purpose of dating an event, such as "tomorrow," "early in the morning," "6 p.m.," or "Christmas day." The trick with this scale is not to code age, sequences of events (e.g., after, next), or greetings like "good morning" or "good night."
This straightforward scale is used to code all negative descriptions. Negative descriptions are of two types. First, there are common negative words like "no, not, none, never, neither, and nor." Second, there are a set of negative words formed by the prefixes "un-," "im-," "il-," "ir-," and "non-"; here the rule is that a word is coded on the negative scale if the word "not" could be substituted for the prefix without changing the meaning of the word. For example, "illegal" is "not legal," "inexcusable" is "not excusable," and "improper" is "not proper."
Food and Eating Scale
The food and eating scale was created due to the possibility that it might reflect the great concern with these issues in industrial urban societies. The scale consists of eating, drinking, the activities leading to eating and drinking, and the surroundings in which these activities occur. The five subclasses can be added together for one overall code.
Elements from the Past
It is sometimes claimed that the elderly and the psychologically disturbed tend to live more in the past than other people. Some theorists believe that dreams are often about the past or are "regressive" in nature. The Elements from the Past Scale was created so these kinds of claims could be addressed in a quantitative way. There are seven straightforward subclasses that encompass such events as being younger, in locales not recently visited, and seeing someone the dreamer has not visited for over a year.
The Elements from the Past Scale is the final empirical scale in the original Hall/Van de Castle coding system. We now turn to some new developments since the scales were created.
Dramatic Dream Reports and Unrealistic Dream Elements
Due to common beliefs about dreams as packed with drama and bizarreness, many scales have been developed to study these aspects of dream reports. Some of these scales will be mentioned briefly in a later section of this chapter. Most of them suffer from an attempt to make a global rating of a dream report on a continuum from highly dramatic to mundane or from bizarre to realistic. Such judgments, as we pointed out at the beginning of this chapter, are difficult to make with a high degree of intercoder reliability.
In this section, we present two very different scales to deal with what may be two different dimensions of dreams--their degree of excitement, "jazziness," or "dramatic intensity" on the one hand and their degree of unreality or unusualness on the other.
We believe that the dramatic intensity of a dream report is best indexed by a simple summation of the number of scores entered in the following seven content categories: aggression, friendliness, sex, success, failure, misfortune, and good fortune. For example, a dream report with three aggressions, two friendly interactions, two successes, and two misfortunes would receive a Dramatic Intensity (D.I.) score of nine; a dream report with one occurrence in each of the seven categories would receive a D.I. score of seven; and a dream report with no entries in any of these seven categories would receive a D.I. of zero. This scale was first used by Hall (1966a: chap. 3) in a study comparing dreams collected at home with dreams collected in the sleep laboratory. He found home dream reports tended to be more dramatic than laboratory dream reports. The dramatic intensity scale also was put to use in a study of dream recall (Trinder and Kramer, 1971).
We also have developed a set of coding categories for unrealistic elements in dreams. These unrealistic elements range from the improbable to the impossible. They include unusual activities and occurrences, distortions, and metamorphoses. We agree with Bonato et al. (1991) that "unrealistic" is a better designation for these kinds of elements than the term "bizarre" now common in the literature. A detailed account of the Unrealistic Elements Scale is presented at the end of Appendix A.
Contrary to what might be expected on the basis of popular stereotypes about dreams, unrealistic elements are relatively rare in the dream reports we have analyzed. For example, Hall and Van de Castle (1966:168) found only 24 metamorphoses in 1000 dream reports from college men and women. They found only 4% of the settings in the same dream reports were distorted in any way. When Hall (1966a:40-41) applied the Unrealistic Elements Scale to the same home and laboratory dream reports that he coded for Dramatic Intensity, he found only 10.4% of 815 dreams which had even one unrealistic element. Unlike what he found with the scale for dramatic quality, there were no more unrealistic elements in the home dream reports than in the laboratory dream reports. Several other investigators using very similar scales (Snyder, 1970; Dorcus et al., 1971; Bonato et al., 1991) also have reported very low frequencies of unusual, magical, or impossible elements in dream reports.
The findings with these two scales suggest there may be two different dimensions underlying our popular conception of dreams as wild or otherworldly. It is also likely that there are individual differences in the dramatic or unrealistic quality of people's dreams, and in the categories of unreality emphasized from dreamer to dreamer. One dreamer may be high on metamorphoses, for example, another on unusual occurrences. Such findings would be another argument against global ratings of dreams as ranging from highly unusual to mundane. Not only are such ratings usually unreliable, but they might throw away information about individual differences in dream content.
Psychopathology and Dream Content
The preceding sections of this chapter presented a wide range of empirical coding categories. We believe these categories encompass virtually all elements to be found in dreams, a point we will demonstrate more concretely in a later section of the chapter discussing coding systems developed by others. For many readers, however, the coding system may appear to be lacking in one major component, a scale for inferring psychopathology from dream reports. This omission might seem especially striking because dreams are considered by many psychotherapists and dream theorists to be a rich source of insights into a person's psychopathology.
Several scales have been constructed to assess "ego strength" or "adjustment" (e.g., Polster, 1951; Sheppard, 1964). They are usually ordinal scales requiring the coder to judge the degree to which settings are realistic, themes are logical in structure, reactions to other dream characters are appropriate, and solutions to problems are reasonable. In some cases, implicit value judgments are built into the scale. In the Polster scale, for example, an aggressive response to aggression is classified as "appropriate" and a nonaggressive response to aggression is "inappropriate" (Hall and Van de Castle, 1966:208). Thus, if a person escapes from an aggressive character, this is supposedly a sign of low ego strength. Given the problems of using rating scales, and especially ones based in value judgments, very little work has been done with ego-strength scales.
We therefore think that a very different approach to the issue of possible psychopathological indicators in dream content should be taken, one based on the empirical categories already presented in this chapter. In our view, it is not productive, at least at this stage of development in dream content studies, to search for something as general as "psychopathology" because dream content may reveal many different kinds of specific problems. For one person, the problem may be the absence of a capacity for friendly interactions, for another, feelings of victimization, and for another, gender confusion or feelings of sexual inadequacy.
We thus believe that several different percentages and indexes derived from our empirical categories hold out the most promise for revealing various kinds of psychopathologies and maladjustments through the study of dream content. As will be seen in later chapters, several of these indicators have been of use in one or more of our studies. However, very little work has been done as yet on this issue. We therefore put forth the following indexes as hypotheses to be tested. The normative expectations for each of them are presented in Table E.1 in Appendix E.
In addition, we believe a summary profile using some or all of the following indicators will prove useful. This summary profile, to be introduced and explained in chapter 4, is based on the magnitude of a person or group's deviation from the norms on each of our indicators.
We stress again: these potential indications of psychopathology in dream content are based on a very few cases. They must be tested on a wide range of individuals and groups. Moreover, some of the indices may prove to be more useful than others. Then, too, it might be necessary to make small adjustments in some of the indices to make them more useful. For example, it is an empirical question as to whether "witnessed" elements are as useful as "dreamer-involved" elements in such indices as the Torso-Anatomy Percent and Bodily Misfortunes Percent. Finally, as noted earlier in this section, we think it highly likely that a pattern of scores on several indicators is more likely to be useful in distinguishing types of psychopathology than any single indicator alone. A method of displaying such patterns using the h statistic will be introduced in chapter 4.
The Issue of Intercoder Reliability
Interjudge (intercoder) reliability, as we have stated several times, is a critical issue in constructing and assessing a content analysis system. However, it is not a straightforward task to decide what constitutes agreement between two coders because there are several ways comparisons can be made. For example, we can compare the total number of codings that two coders make for several samples of dream reports. While such an approach has its uses, it does not determine whether or not the coders are actually coding the same elements in arriving at what may be very similar totals. Thus, the most meaningful and stringent measure for our purposes is to determine the percentage of times that the coders agree on the number and types of specific elements occurring in each dream (see Hall and Van de Castle, 1966:144-149 for a discussion of the various ways of assessing intercoder reliability). We call this preferred method the "percentage of perfect agreement" approach. It is always based upon just those elements coded as present by at least one coder. The numerator contains the number of coding agreements and the denominator consists of the number of agreements plus the number of disagreements.
However, it is not enough to ascertain intercoder reliability. It is necessary for the investigator and coders to discuss and resolve any coding differences. The final results should be based on the consensus results of the coders. This approach corrects for the blind spots that particular coders may have, meaning that some coders may overlook one category, for example, while others may confuse or overlook other categories. This procedure not only improves the quality of findings, but increases reliability of coding for future studies.
In the best of all possible worlds, the issue of reliable coding would transcend individual investigators or investigative teams. Dream content researchers would periodically compare their coding of the same dream series or dream set in order to ensure that everyone is using the Hall/Van de Castle system in the same way. Such comparisons would help to increase confidence that different findings from research group to research group are real findings, not artifacts of differences in coding. In this regard, it should not be forgotten that sleep researchers in different laboratories had to work very hard for two years to achieve reliability in their scoring of sleep EEG records (see Rechtschaffen and Kales, 1968).
In order to determine the reliability of their system, Hall and Van de Castle did a new study in which each of them scored from 50 to 100 dream reports for each separate category, then compared their results. They first of all found very high levels of agreement in the overall frequency of elements for each category; their correlation coefficients concerning the total number of elements present, as well as the number of elements within separate categories, were generally above .90. The figures on perfect agreement ranged from about 60% to 90%, depending on the complexity of the scale being coded.
Deciding among the hundreds of possible character combinations in 100 dream reports, Hall and Van de Castle agreed exactly on the number, gender, identity, and age of 76% of the characters. They agreed on the presence of a character 93% of the time, on gender 89% of the time, and on age 92% of the time. In the settings category, where there are 16 different codings if we include "no setting," there was 73% agreement. The percentages were in the 80's for objects, activities, and modifiers.
Generally speaking, the greatest difficulties were in the coding of social interactions. Although Hall and Van de Castle agreed almost perfectly on the total number of social interactions in 50 dream reports, the percentage of perfect agreement ranged from 54% for aggressive interactions to 64% for sexual interactions. However, it must be remembered that coding a social interaction involves a number of components. In order for perfect agreement to occur, the coders had to agree that a social interaction was present, then agree as to the coding for the characters involved, and finally agree as to the appropriate subclass of the interaction and whether it was an initiated, reciprocated, mutual or self-directed interaction. If a more lenient criterion were followed whereby the coders could disagree on some single component, such as the subclass number, then the level of agreement was over 70% for each of the three types of social interactions.
A summary of the findings from Hall and Van de Castle's reliability study can be found in Table 2.1. These findings show that the coding system can be used with great reliability by those willing to take the time to master it. We suggest that the best way to do this is for two or more people to learn it together. This helps to maintain interest in the learning process, provides an incentive to try harder, and makes it easier to pinpoint and overcome blind spots.
To aid in the task of learning to use the system in a reliable fashion, as noted earlier, Appendix B provides Hall and Van de Castle's coding of ten dreams from a young adult male. His series was selected because it had a large number of difficult coding decisions. Once new investigators have learned the rudiments of the system and tried to apply it to 25 or 30 dream reports that they have available to them, we suggest that they then read through and code the ten dreams reports in Appendix B, and compare their coding to that by Hall and Van de Castle.
Other Coding Systems
As noted in the first chapter, many other coding systems for the content analysis of dreams have been developed over the years, and here we can add that new investigators often develop new scales, ignoring what has been done in the past. Most of the older coding systems have been brought together in Hall and Van de Castle (1966: chap. 15) and Winget and Kramer (1979). However, there are drawbacks with these other coding systems, as will be shown in this section. Moreover, their categories can be found in the Hall/Van de Castle system or can be duplicated by bringing together two or more categories from the Hall/Van de Castle system. If the arguments and evidence brought together in this section are convincing, thereby leading new dream researchers to use the Hall/Van de Castle system, then the possibility for an accumulation of findings on dream content will be increased.
Weaknesses of Other Coding Systems
The first and biggest problem with most other coding systems is that the categories are not defined clearly enough to make it possible to develop a high level of interrater reliability. Even when good reliability can be achieved among coders in the same laboratory, there often is difficulty in obtaining the same levels of reliability when investigators from outside the original research group try to use it.
These problems concerning reliability are demonstrated with an Imagination Scale developed by Foulkes and Rechtschaffen (1964) to determine the effects of two different films shown just prior to bedtime on the degree of elaboration and reality in subsequent dream reports. This weighted scale is summarized in Table 2.2, which is adapted from Winget and Kramer (1979:118).
In the original reliability study with the scale, the two coders' ratings were quite similar. The Pearson correlation coefficient based on an average score for each subject was .78 for REM dream reports and .86 for NREM reports. However, to his great credit, Foulkes candidly reports that it was difficult to obtain good levels of reliability with new coders when he moved to a new university setting:
The second major problem with most other coding systems is one we noted earlier: they are constructed as rating scales, requiring the coders to make difficult and often subtle judgments about more or less imagination, more or less hostility, or more or less ego strength. Moreover, as we also noted earlier, some rating scales are based on questionable assumptions about which elements in a dream deserve greater weight. Rather than refer readers to the earlier examples, we present in Table 2.3 the weighting system for the first two categories of Sheppard's eight-category Ego Scale of Integrative Mechanisms (1963, 1969), as summarized in Winget and Kramer (1979:100-101).
We do not think there is a strong rationale for such an 8-4-2-1 weighting system. Nor is it obvious, even if a weighting system is accepted, why a mutilation is only half as significant as a deformity, for example.
There is a third problem with most other scales. They have been used in only a small number of investigations, usually one or two. Rarely have any other scales except those developed by Hall and Van de Castle been used by anyone except the original investigators. There is thus a proliferation of scales, none of which has been used enough to establish its general validity or usefulness.
Beyond the three serious weaknesses of other coding systems, most of them duplicate the Hall/Van de Castle system, or can be constructed in a more objective fashion by combining elements of the Hall/Van de Castle system.
To take the most striking example of overlap, Rychlak (1960; with Brams, 1963) breaks the dream report into five categories very similar to those developed by Hall in 1951, and then incorporated into the Hall and Van de Castle system:
There is also considerable overlap with Brenneis's (1967, 1970, 1975) "34 Dimensions in Ego Styles," summarized in Winget and Kramer (1979:96-97). We present the first 11 questions on this scale to give readers a flavor:
Systematic evidence for our claim of overlap between the Hall/Van de Castle system and other scales can be found in a study by Hauri (1975) comparing many different coding systems. Hauri began his study by putting together a set of 100 randomly selected REM dream reports--50 from men, 50 from women-- collected in ten different sleep laboratories. Hauri (1975:272) explained his starting point as follows:
Hauri then sent these 100 dream reports to the authors of many different dream scales for coding by the authors or their close associates. After receiving the various codings of the dreams reports from the different investigators, Hauri did a factor analysis. He found six basic factors or dimensions in the various scales:
Can these factors be duplicated with the Hall/Van de Castle scales? We think they can for all but the final one, compulsivity of reporting. Factor I, dreamlike quality, brings together scales for distortion, imagination, vivid fantasy, and primary process. We saw one of these scales already when the reliability of the Foulkes/Rechtschaffen Imagination Scale was discussed. Other of these scales use dimensions like "plausibility," ranging from plausible and realistic in terms of the subject's everyday life to implausible and bizarre. We believe the dimensions of dream content being tapped by these scales can be duplicated by the combination of the two scales presented in the previous section of this chapter: The Dramatic Intensity scale and the Unusual Elements scale.
Factor II, hostility and anxiety, picks up hostility, anxiety, unhappiness, and lack of human relationships according to Hauri (1975:272). Basically, most of this factor is contained in the Hall/Van de Castle categories for aggression. The rest of it is picked up by the percentage of negative emotions and the aggression/friendliness percent, presented earlier as possible indicators of psychopathology.
Factor III, motivation to self improvement, basically involves three categories from the Hall/ Van de Castle system: the initiation of friendliness, the initiation of aggression, and successful striving. In the Hall/Van de Castle system, a person who befriends dream characters more often than he or she is befriended by them, who is an aggressor more often than a victim, and who has more successes than failures would score high on this factor. Factor IV, sex, essentially overlaps the detailed Hall/Van de Castle coding categories for sexuality. Factor V, activities, essentially overlaps the detailed Hall/Van de Castle coding categories for activities.
Factor VI, labeled "compulsivity of reporting" by Hauri, contains only a category for the number of words in the dream report and a coding scale for "anality" developed by Sheppard (1964). There does not seem to be any obvious relationship between this scale and the number of words in dream reports. Nor is there any Hall/Van de Castle scale that duplicates this factor. In any case, it does not seem to us to be a very coherent or useful scale, and it never has been used by anyone but its author.
"Masochism" and Dream Content
Support for our claim that scales developed by others can be encompassed within the Hall/ Van de Castle system can be found in a study by Clark et al. (1972) comparing the findings from a "masochism" scale with the findings from three Hall/Van de Castle scales focused on the same dream elements as the alleged "masochism" scale, which makes questionable assumptions about what dream elements show a love of suffering and sexual submission, and never has been validated for the theoretical claims in its label. The "masochism" scale was developed by investigators to test their hypothesis that depressives had a greater number of masochistic dreams than non-depressives (Beck and Hurvich, 1959; Beck and Ward, 1961). It is one of the few scales ever used by other than the original investigators (e.g., Kramer, et al., 1965, 1966, 1969; Cartwright, 1992). The scale can be summarized as follows:
(For more detailed statements of this scale, see Hall and Van de Castle, 1966:205-208, or Winget and Kramer, 1979:85-86.)
After comparing the categories in the "masochism" scale with the Hall/Van de Castle system, Clark et al. concluded that all of the alleged "masochism" categories were encompassed by just three categories in the Hall/Van de Castle system: victimization, failure, and misfortune. They then coded 21 morning-collected dream reports from manic-depressive patients and 91 laboratory collected dream reports from depressive patients using both the "masochism" scale and the three Hall/Van de Castle scales. The percentage of agreements between the two systems was 81% for the manic-depressives' dream reports and 91% for the depressives' dream reports. When the investigators compared the points of disagreement between the two coding systems, they found that in 11 of the 12 instances the Hall/Van de Castle system coded a "masochistic" element while the other scale did not. Thus, the Hall/Van de Castle system is more comprehensive than the original scale. The authors conclude: "The results of the present study indicate the feasibility and the potential advantages of deriving new dream content scales from existing reliable and valid coding systems rather than increasing the already large collection of independent scales" (Clark et al., 1972:118).
Additional evidence for the similarity of the "masochism" scale to a combination of Hall/Van de Castle scales can be found in a gender difference in "masochism" reported by Cartwright (1992:82). Analyzing dream reports from men and women going through divorce, Cartwright found that all the women in the study, whether depressed or not, had higher scores on the "masochism" scale than did the men. This theoretically anomalous finding may be explained within the Hall/Van de Castle system with findings on college men and women to be presented in the fourth chapter: college women are slightly more likely to be victims of aggression than are college men (67% vs. 60%), and to have a lower success percent (42% vs. 51%). Moreover, they have a higher bodily misfortunes percent (35% vs. 29%). Thus, the alleged difference in men and women on "masochism" probably is accounted for by victimizations, failures, and bodily misfortunes.
In this chapter, we have presented an overview of the Hall/Van de Castle system and shown that it is empirical, objective, and comprehensive. Furthermore, we have demonstrated that it can be coded with good reliability on a very stringent measure, the percentage of perfect agreement. Throughout the remainder of this book we will show it has been learned and used by investigators all over the world solely on the basis of the coding instructions and examples presented in Appendices A and B. We stress again that everything a person needs in order to learn and utilize this system is included in this book.
This chapter also has shown that the Hall/Van de Castle system encompasses most, if not all, other coding systems. It has argued that the Hall/Van de Castle system is more useful than the others because (1) its coding categories are more explicit; (2) it does not make unwarranted ranking or weighting assumptions; and (3) it can be used to create a set of psychopathology indicators.
But does the system have "validity"? Does it give us results that make sense in terms of what we already know about gender, age, and cultural differences, and does it give us new hypotheses about dream meaning or a dreamer's personality that turn out on further investigation to be accurate? Before we can answer these questions, it is necessary in the next chapter to address the quality of our data. No matter how reliable the coding system may be, we will not be able to obtain useful findings if the quality of the data is poor.
Go back to the Dream Library index.