There are serious obstacles to systematic studies of dream content. First, as noted in passing in the first chapter, it is not possible to introduce stimuli that regularly produce predictable variability in dream content, so an experimental approach is not very useful. Second, not all subjects are willing or able to report dreams, raising questions about the representativeness of those subjects who do report dreams. Third, even high dream recallers do not report dreams every morning at home or every time they are awakened in a laboratory setting, raising questions about the representativeness of the dreams recalled by subjects. Fourth, variability in how dreams are collected may affect the content of the reports. Finally, there are no independent checks on the accuracy of the reports provided by subjects; elements of the dream could be omitted or changed, or the entire "dream" could be a made-up story. Given these problems, it is not surprising many psychologists raise questions about the quality of the data used in studies of dream content.
It will be the purpose of this chapter to argue on the basis of several different types of evidence that the quality of our data is very good despite the difficulties and uncertainties involved in obtaining them. It can be shown that the subjects who provide dreams are a surprisingly representative sample of the general population in terms of their personality structures and cognitive styles, and that the dream reports they provide are very likely a fairly representative sample of their overall dream life. The dream reports we analyze are based on a standardized form, and there are reasons for believing the reports are honest ones.
Individual Differences in Dream Recall
The representativeness of the subjects in studies of dream content can be approached through the large number of studies attempting to find personality, cognitive, and other differences between those who are frequent and less frequent recallers (see Cohen, 1979, and Goodenough, 1991, for complete overviews of this literature). Most of these studies are correlational in nature. Recall frequency is determined by either brief questionnaires or daily diaries, which have been found to correlate highly with each other (Cohen, 1979:159-160).
First, and most importantly for our purposes, the great preponderance of studies conclude that personality factors play little or no role in determining who is and is not a frequent recaller in everyday life (e.g., Berrien, 1933; Stickel, 1956; Domhoff and Gerson, 1967; Farley et al, 1971; Robbins and Tanck, 1971; Trinder and Kramer, 1971; Cohen, 1970, 1973, 1979; Redfering and Keller, 1974; Hartmann, Elkin, and Garg, 1991; Tonay, 1993). Cohen (1979:161) speaks for most investigators in this field when he says "correlations between dream recall frequency and specific personality measures have been weak, trivial, or inconsistent," as does Kramer (1982:89) when he concludes that "for the general population personality traits are not a significant factor in determining dream recall." Thus, if the "meaning" of dream content inheres in good part in its relationship to "personality" broadly defined, then for our purposes we have a representative sample of subjects for studying this relationship.
Second, cognitive variables have an influence on differences in everyday dream recall, but the influence is not much larger than personality variables (Cory et al., 1975; Cohen, 1971, 1979; Hiscock and Cohen, 1973; Martinetti; 1983; Fitch and Armitage, 1989). Third, physiological factors unrelated to personality and cognitive variables seem to have a part in making some people less able to recall their dreams. This possibility was first raised in sleep laboratory studies of recall which found gradual arousals substantially decreased recall from REM periods (Goodenough et al., 1965). This led to a study showing some low recallers had high waking thresholds and were difficult to arouse in the sleep lab (Zimmerman, 1970).
However, none of these factors seems as important in adults as interest, motivation, or a belief dreams are meaningful or important (e.g., Strauch, 1969; Cohen and Wolfe, 1973; Tonay, 1993). Emphasis on these more mundane and changeable factors is given added support by the fact there is very little variation in recall by subjects who are studied in the sleep laboratory. Interest and motivation are the factors, of course, likely to lead a person to volunteer to take part in a dream content study, so it is important to note that an interest in dreams does not correlate with personality variables. It may correlate with an openness to inner experience, but that factor does not correlate with personality tests either (Hartmann, Elkin, and Garg, 1991).
Recall frequencies also vary with mood and stress, which suggests, along with the findings on interest level, that dream recall is largely independent of personality and cognitive variables. For example, in one study people instructed to keep dream diaries reported more dreams when they were in a bad mood, especially if they tended to be low recallers (Cohen, 1974). Moreover, there may be gender differences in the effects of such variables. Armitage (1992), in a diary study of 15 male and 15 female young adults over the space of one month, found women recalled more dreams when they were under high stress and males when they were under low stress.
When we add up all these studies, we conclude that we do not have any one or two strong predictors of frequency of dream recall. We agree with Goodenough (1991:157) when he says "people apparently may be nonreporters for a variety of distinctly different reasons." It thus follows that a representative sample of people are able to contribute to studies of dream content, meaning there will be the usual individual differences on personality and cognitive factors in any large-scale study of dream content.
However, this does not mean all the samples used in dream content studies are equally good. The same issues making for a representative or unrepresentative sample in other types of psychological studies remain relevant. For our purposes, the most important of these issues is sample size. A handful of dreams from each of a few subjects is not going to make for a study with replicable or generalizable results. As will be seen, our studies are usually based on dozens of subjects contributing hundreds of dream reports, and in a few cases we have thousands of dream narratives from thousands of subjects, or several hundred dreams from a single subject. Studies with inadequate samples should not be given the same weight as ours in making generalizations about meaning in dream content.
In any case, the representativeness of the subjects in normative dream content studies should not be a concern. They are typical people with a wide range of individual differences.
The Representativeness of Recalled Dreams
It seems likely that every person is having at least four to six dreams per night. It is estimated that less than 5-10% of all dreams are remembered. This is a very small sample, and thus a sample that may be very biased. Even the highest of recallers, two or three dreams every day, are only remembering from one-third to one-half of their dreams.
For some purposes, it is not relevant if the few dreams we recall are representative or not in their content. Harold Sampson (1969:222) argues that, for cross-cultural and personality studies, the important point is whether the dreams lead to useful and reliable conclusions about group differences or a given subject's personality. Rather than comparing samples of dreams, such as those recalled at home with those collected in the sleep laboratory, Sampson suggests comparing the conclusions about dream meaning and the relationship of dream content to personality derived from studies of the two types of samples. While we agree with Sampson up to a point, the fact remains no one has compared conclusions drawn from the two different types of report samples. Moreover, it is possible to present an argument based on systematic evidence that the dreams we recall are a fairly representative sample of our total dream life, making it unnecessary to compare conclusions drawn from home and laboratory dream samples. We say "fairly representative" because there is evidence of a tendency to remember the most dramatic dreams, but even here there are other factors leading to recall of many prosaic dreams.
We begin our analysis of the representativeness of dream reports by focusing on sleep laboratory studies because they make it possible to sample throughout the night from both REM and NREM stages of sleep. Furthermore, there is anywhere from 70% to 90% recall from REM periods, making it easy to collect a large number of reports fairly quickly. Most dream content studies in the laboratory have focused on REM periods, but this is no problem for our purposes because research has shown there is little or no difference between REM and NREM reports. For example, using data collected by Foulkes and Rechtschaffen (1964), Hall found in an unpublished study with the Hall/Van de Castle coding system that many NREM reports from late in the sleep period scored as high on the dramatic intensity index as REM reports. Antrobus (1983). Foulkes and Schmidt (1983) found few or no differences when they controlled for report length.
The first relevant sleep laboratory finding is that there are no differences on the Hall/Van de Castle coding scales between early and late REM periods except for some minor exceptions due to the shortness of the first REM period of the night (Domhoff and Kamiya, 1964b; Hall, 1966a; Strauch and Meier, 1995). Thus, if there is a tendency for people to recall only their last dream of the night, as studies we will present in a moment do suggest, then this tendency does not introduce a bias into our sample of dream content. In fact, the greater the importance of "recency" in everyday dream recall, the more we are assured of a representative sample.
The second relevant finding from sleep laboratory studies is that there are only slight differences between the home and laboratory dream reports in four different studies using the Hall/Van de Castle coding system (Domhoff and Kamiya, 1964a; Hall, 1966a; Bose, 1983; Strauch and Meier, 1995). Then too, there were no differences in a comparison of 51 home reports with 51 laboratory reports from nine female subjects (Domhoff, 1962), but 51 reports per condition is a very small sample. The primary difference in two of the studies was a greater dramatic intensity in the home reports; in particular, there were more aggressive and sexual acts. While Hall (1966a) and Domhoff (1969) originally argued that inhibitory effects in the laboratory setting were probably as much of a factor in explaining these differences as selective recall for more dramatic dreams at home, subsequent research by Foulkes (1979; with Weisz, 1970) has shown the differences can be accounted for by selective recall of everyday dreams.
But what factors seem to be involved in which dreams are recalled at home, and how important is selection for dramatic intensity? In our view the most interesting evidence on this issue comes from laboratory studies comparing what subjects report from REM awakenings with what they remember in the morning. In the first of these studies, Meier, Ruef, Ziegler, and Hall (1968) studied the night and morning recall of an adult male subject who was awakened at the end of every REM period for 45 nights. The subject recalled 138 dreams when awakened, 88 of which he still remembered in the morning. When the dreams recalled in the morning were compared with those not recalled, it was first of all found that dreams late in the sleep period were recalled more frequently (83% for the last 105 minutes versus 63%, 55%, and 52% for the three preceding 105-minute periods). The intensity of the dream, as judged independently by the dreamer and his Jungian analyst, also predicted recall, with high-intensity dreams being recalled in the morning more frequently than low intensity dreams (83% versus 56%). Longer dreams were recalled better than shorter dreams (87% versus 48%). The number of awakenings per night also had an effect, with more awakenings decreasing morning recall, but that variable will not be considered further here because it has less relevance outside the laboratory.
The interactions among the three main variables influencing morning recall were then analyzed. It was found recency can compensate for low intensity and shortness of reports, and length can compensate for low intensity. If these results can be generalized outside the laboratory, they suggest recency and length are giving us many dreams not unusually high in dramatic intensity.
Three similar studies using a larger number of subjects also found recency and length were major factors in the morning recall of reports from REM awakenings in the sleep laboratory (Baekeland and Lasky, 1968; Strauch, 1969; Trinder and Kramer, 1971). Two of the three (Strauch, 1969; Trinder and Kramer, 1971) also agreed dramatic intensity correlated with recall of REM reports in the morning.
Domhoff's (1969) naturalistic study of the factors involved in everyday recall over a two-week period suggests recency and mundane memory cues figure in a large portion of our total recall. Six male and six female students at California State University, Los Angeles, were recruited from the student employment office to record their dreams and the events surrounding recall as soon as the dreams were recalled. The subjects were not selected in terms of interest in dreams or frequency of recall, but all said they recalled dreams at least occasionally.
In a counterbalanced design, subjects wrote their dream reports one week and telephoned them to an answering machine the other. After recording the dream, subjects were instructed to answer a series of questions concerning the correlates of the dream recall. The questions concerned the time, location, mood, thoughts, activities, and number of companions at the time of recall. Subjects were also asked to give their opinion as to why they recalled the dream, and to search for links between the dream and any specific events that may have led to the recall.
The 12 subjects made 30 telephone reports and 27 written reports. There were no differences in content or recall factors between the two types of reports. There were wide individual differences in frequency of reporting. Two subjects reported no dreams at all, one coming by regularly with great apologies, the other uncommunicative. Two others reported only one. Two subjects provided two and three telephone reports respectively before dropping out. Six subjects accounted for 23 of the 30 telephone reports and all 27 written reports.
The overwhelming majority of reports (72%) were made immediately upon awakening. Thirty-seven of those awakening reports came in the morning and four came right after afternoon naps. Contrary to what we expected, only two reports (4%) came from night awakenings, and both of those were from the same subject. The rest of the recall came from late morning (6%), afternoon (12%), or evening (6%). Non-morning recall seemed to be related to one or both of two factors, neither of which concerns dramatic intensity. First, minimal external cues seem to suddenly bring back dreams in which the external cue triggering the dream memory may play only a small part. Second, non-morning dream recall seems to occur when the dream memories are in some way linked to thoughts and associations the person is having in a relaxed, sometimes daydreamy mood; often, as might be expected, he or she is alone.
Here are some examples of each type. A male subject, talking one evening with his girlfriend about the movie they were about to see, suddenly remembered a dream from the night before in which he had been talking to Peter Sellers. The whole dream was then available to him in considerable detail, including the fact he had told Peter Sellers he was engaged to a friend of Sellers. Another male subject, telling a friend about his car, suddenly remembered a dream in which the paint was peeling off the car. In this case the recall was very fragmentary, consisting of two short images. A female subject about to sit down in chemistry class, spotted a friend and suddenly recalled in detail a dream about a person's supposed failure in the chemistry laboratory. Another time this same subject was watching a television ad about skiing that brought back a skiing dream.
Examples of internally triggered recall are as follows: A female subject was sitting alone studying when her mind wandered to a recent date and the movie they had seen--suddenly she remembered a dream about a monster trying to get her, without, we might add, suggesting any connection with the male escort. A female subject daydreaming about getting married vividly recalled a complicated dream in which she was a housewife.
A female subject from whom we have not drawn any previous examples presented a case where internal and external cues combined to produce an afternoon recall. Her report, phoned in immediately, began as follows:
My code letters are GG. It's 1:15 Sunday afternoon and I was doing my homework. I've been sitting in the sunshine, but the sun is moving and so it's getting to be half sun and half shade in my chair, and I started getting cold, and when I got cold I started thinking about the trip I'm going on in two weeks with the Photo Club, and while I was thinking about the trip, then I happened to remember the dream I had last night ...
Perhaps the factors of external cue, relaxed state of mind, and being alone are best summarized in the following early-evening recollection, the only dream reported by this male subject:
I'd come home and I was relaxing and sitting around. I lit a cigarette and all of a sudden I remembered that this is what I'd dreamed about [burned his fingers lighting a cigarette, then it burned down and burned his fingers]. I wasn't thinking about anything in particular at the time. I was kind of relaxed and sitting alone ... I was relaxed I think for the first time today.
Much of these recall data are anecdotal in nature, but they are worthwhile for the point they make: recency, length, and intensity are not the only factors involved in everyday dream recall. Simple everyday external cues and internal associations also may be important (cf Foulkes, 1985; 81-87 for a discussion of "retrieval cues" in explaining dream recall).
Putting together the various findings from different studies of what factors relate to dream recall, we can say there is good reason to believe the dreams recalled by our subjects are a fairly representative sample of their dream lives on the elements we quantify. True, there is a strong tendency to remember dreams occuring just before we awaken, but that is no challenge to the representativeness of our data because of the laboratory evidence there are no systematic differences in dream content throughout the night. True, too, there is a tendency to remember longer dreams, but that is an aid in terms of representative dream content because many long dreams have low-intensity content. True, also, there is a tendency to remember more intense or dramatic dreams, but that tendency is counterbalanced to some extent by the fact external stimuli and drifting thoughts can trigger what seem to be very mundane, not intense, dream reports.
The argument that we obtain a large number of relatively mundane dream reports from everyday dream recall will be strengthened by the normative findings in the next chapter. Only about 10% of dream reports from college students contain any sexual content, for example. Misfortunes occur in one of every three dreams. Most important of all, perhaps, only 44% of female reports and 47% of male reports contain any aggression. If we look at dream reports from a psychological standpoint, we might argue that the amount and intensity of aggression in dream reports is high, but in terms of an argument that only dramatic or intense dreams are recalled, the important point is how many dreams do not contain dramatic or aggressive content.
The Standardization of Dream Reports
As with any personal information, the nature and quality of dream reports can be shaped by experimenter effects and the demand characteristics of the experiment. For example, Kremsdorf, Palladino, and Polenz (1978) found in a study of five male and five female subjects that opposite-sex interviewers received dream reports with less sexual content and more conflict elements than did same-sex interviewers. Stern, Saayman, and Touyz (1983), in a study using both home and laboratory dream reports from 12 college students, varied the instructions on the report forms, requesting subjects to pay special attention to either the outdoor/nature or urban settings in their dreams. They found statistically significant differences in the relative frequencies with which the two types of settings were reported.
To deal with these kinds of problems, the dream reports utilized in most of our studies have been collected in written form with complete anonymity for the subjects through the use of a code number or a self-chosen code name. The completed forms are placed in a large envelope or box with a minimum of interaction with the person in charge (usually an instructor in a classroom). Sometimes the dream reports are mailed to the investigator. As will be seen in the next chapter, we have found no differences in dream content in studies conducted by men or women. The instructions on our form, reported in Appendix C, are similar to those used in most dream content studies. Some studies by other investigators have used our form exactly.
We said in the previous paragraph "most" of our studies have used this form. This qualification was necessary for five reasons:
Many of the young children's dream reports in our studies were told to teachers or parents.
Some of our dream reports from psychiatric patients were told to psychotherapists.
Most of our long dream series were written down by their authors for their own purposes and without knowledge of our work on dreams.
Some of the reports used in our studies were tape-recorded in sleep laboratories and then transcribed.
Most of the dreams in our cross-cultural studies of preliterate societies were written down by ethnographers.
The fact that young children are telling their dreams to adults in an interpersonal situation may well affect the nature of their reports, but there is little other choice until children are old enough to write full reports. The number of such reports in our studies are relatively few in any case, and the differences in the content of children's reports to be presented in chapter 5 are very likely due to developmental factors discovered by Foulkes (1982) in a five-year longitudinal study.
With psychiatric patients, there may well be a selective factor in which of their dreams they tell their therapists. This point was demonstrated empirically in the case of two subjects by Whitman et al. (1963), who found that the dreams these subjects reported after REM awakenings in the sleep laboratory were not always the same ones they told their therapists the next day.
When it comes to the tape-recorded reports in laboratory studies, they are usually longer than written reports. They often contain more characters and more descriptive details. These factors sometimes need to be corrected for in comparing tape-recorded reports to written reports. In terms of the percentages and indexes we use in our analyses, the greater length of tape-recorded reports is not a problem.
Dream content analysts are not unmindful of the many possible methodological issues involved in the collection of adequate dream reports (Urbina, 1981). However, these problems are minimal if we collect reports in an anonymous and written fashion with written instructions of the kind we and other investigators commonly utilize.
The Honesty of Dream Reports
Some investigators may be concerned with whether or not dream reports are invented by subjects. Those concerns are addressed by studies comparing "real" and "made-up" reports on content dimensions or asking judges to try to distinguish the two types of narratives. Brenneis (1967:86) and Carswell, Melody, and Webb (1985), the latter using 180 dream reports and 278 artificial reports, found content differences, but Cavarello and Natale (1988-89), using 24 dream reports and 12 artificial reports, did not. As far as judges, Darbes (1952) found a group of people who had little or no experience with dream reports were able to make the distinction, but Carswell, Melody, and Webb (1985) and Cavarello and Natale (1988-89) found judges couldn't tell the two types of report apart.
While we realize it is possible that some subjects invent or purposely alter details in the dream reports they give to investigators, we believe this problem is very rare as long as subjects are anonymous and participation is voluntary. Most people simply do not feel enough personal "responsibility" for the content of their dreams to have any motivation to censor them (Foulkes, 1979:249). We believe further that the large number of subjects and dream reports in our studies make the impact of any invented reports extremely minor.
Some indication of the frequency of invented reports and the crucial importance of voluntary participation is demonstrated in a normative study of women's dreams by Veronica Tonay (1990-91) discussed in the next chapter. After collecting dream reports on a voluntary basis from some of the students in a large psychology class, Tonay administered a post-collection informational questionnaire to all students in the class. It included two questions concerning artificial reports. Four of her 109 subjects (3.7%) said they made up one or more of the five dream reports they turned in. Forty-three percent of all students in the class, which means both participants and non-participants in Tonay's study, said they probably would have made up dream reports if participation in the study had been required. Tonay's findings give us confidence in the quality of our data, but they also warn us it would be a serious mistake to require the reporting of dreams.
In all, then, we think there is every reason to believe that our findings are based on representative subjects, a representative sample of their dream life, and honest dream reports. The consistency of the findings reported in later chapters, including correlations with cultural patterns and personal preoccupations, will be further evidence for this point.
Studies of dream content from REM awakenings in the laboratory assure us that any differences from everyday home-recall reports are matters of degree, not of kind. There are no types of dreams or dream experiences that were unknown before laboratory studies of dream content began in the latter half of the 1950's. However, some differences in dramatic intensity may need to be corrected for in comparing laboratory and home samples. In particular, laboratory studies using the norms to be presented in the next chapter should not make too much out of lower frequencies on the categories making up the dramatic intensity scale presented in the previous chapter. This is particularly the case with sexual and aggressive elements.
To say that the quality of our data is generally good does not mean all studies of dream content are good. There are very large variations in the quality of such studies. The problems of uneven quality are then compounded by the fact that some authors of summary papers and popular books take every study at face value, turning poor studies into conventional wisdom and giving the false impression there is more disagreement and confusion in the literature than there really is. Then too, theorists with axes to grind are often uncritical of the quality of studies reaching conclusions favorable to their viewpoints.
Dream content studies can be of poor quality for a number of reasons. Sometimes the number of subjects or reports in a study is very small, making any conclusions dubious at best. Sometimes studies are based on brief dream reports that are clearly hasty or half-hearted productions. Sometimes the dream narratives have been collected by asking for only a memorable dream, making comparisons with everyday dream reports collected over the space of days or weeks a highly questionable enterprise. Sometimes the studies lack adequate comparison groups. Some studies draw conclusions without a knowledge of previous research that should be taken into consideration. Others misinterpret previous findings or use inappropriate statistical tests. Still others do not make clear that the findings are based on questionnaire surveys asking for the subject's opinion on his or her dream content; conclusions from such studies cannot be compared directly with quantitative studies of dream content. For all these reasons, it is an unpleasant but necessary task to point out the weaknesses and misinterpretations in some of the studies cited in later chapters.
If our criticisms in subsequent chapters of inadequate studies are correct, then we will have strengthened our case for the quality of the data used in good quantitative studies of dream content. In other words, the problem with the research literature on dream content is not the quality of the data, but the varying quality of studies based on that data, along with the uncritical acceptance of poor studies by partisans and popularizers.
We are now ready to consider the most systematic findings we have using the Hall/Van de Castle coding system.
Before serving as a subject in the laboratory, this middle-aged man kept a dream diary for about 1 1/2 years (560 nights), during which time he recorded 105 dreams, or about one every five or six nights. Prior to keeping the dream journal as part of a Jungian analysis, he remembered only a few dreams in his entire life. The subject held an executive position in a large pharmaceutical company at the time of the study.