There now exists a large body of firmly established descriptive empirical findings on dream content (Domhoff, 1996; Foulkes, 1982, 1985, 1993). The purpose of this paper is to assess the implications of these findings for Freudian, Jungian, activation-synthesis, and cognitive theories of dreams. The paper concludes that these established findings contradict Freudian, Jungian, and activation-synthesis theory in one way or another, but are compatible with the cognitive theory of dreaming developed by Hall (1953b) and Foulkes (1985, 1999).
Although the focus in this paper is on studies of dream content, it should be noted that there are other kinds of evidence that do not support the Freudian, Jungian, and activation-synthesis views. For example, two different sympathetic examinations of Freudian dream theory 20 years apart by Fisher and Greenberg (1977, 1996), drawing on a wide range of dream studies, conclude that there is no support for its most specific claims. Also, the repetitive dreams of post-traumatic stress disorder have been recognized by several dream researchers as contradictory to Freud's theory of dreams as a disguised attempt at wish fulfillment, including Freud himself (e.g., Fisher & Greenberg, 1977, 1996; Freud, 1920, 1933; Hartmann, 1998).
In the case of Jung, the main critique has focused on the unscientific nature of his key concept for understanding dreams, the collective unconscious. For example, Neher's (1996) comprehensive analysis and synthesis suggests that the concept is unscientific because (1) it is based on the discredited notion of the inheritance of the repeated experiences of the human species and its ancestors; (2) it does not allow for the variation in specific archetypes that is a basic aspect of genetics; (3) it is not grounded in a convincing elimination of the possible influence of socialization and culture in the personal, therapeutic, and cross-cultural examples on which it is based; and (4) it is not able to escape the charge of circularity because its origins are said to be in repeated human experience, which is the phenomenon it is supposed to explain. If experience is the basis for the collective unconscious, there is no need to invoke it to explain experiences that are more likely based on common human conditions.
As for the activation-synthesis theory, it is called into question by the following findings that have nothing to do with dream content. First, the presence of dreams at sleep onset and in NREM sleep does not fit with the theory's emphasis on the REM-sleep origins of dreams (Foulkes, 1962, 1996a; Herman, Ellman, & Roffwarg, 1978; Vogel, 1991; Vogel, Barrowclough, & Giesler, 1972). Second, the lack of correlations between phasic events in REM sleep and types of dream content does not support its emphasis on neurological events during REM sleep as the causes for the abrupt scene changes, unusual juxtapositions, and blended characters that sometimes can be found (Foulkes, 1996a; Pivik, 1978, 1986).
In this paper, however, the focus is strictly on the problems created for Freudian, Jungian, and activation-synthesis theory by four very solid findings concerning dream content:
Since all of these descriptive empirical findings are compatible with a cognitive theory of dreams, the paper concludes that dream researchers should place more emphasis on testing and developing that theory.
Dreaming as a gradual cognitive achievement
There is good longitudinal and cross-sectional evidence from the sleep laboratory that the ability to dream develops gradually between the ages of 3 and 8 (Foulkes, 1982, 1993, 1999; Foulkes, Hollifield, Sullivan, Bradley, & Terry, 1990). For example, in the cross-sectional study by Foulkes et al. (1990), only 20% of REM awakenings at any age between 5 and 8 led to any kind of dream report, and most of those few reports were relatively undeveloped in content, containing static images or observations of simple events. Moreover, children rarely inserted themselves into dreams as active participants until they were 8 years old. Foulkes (1982) concludes that dreaming as a cognitive process does not approach adult levels until ages 8-9 and that dream content does not approximate that of adults in length and content until ages 12-13. Such findings are very different from what is claimed in the Freudian and Jungian dream literatures about the symbolic contents of children's dreams. As for the anecdotal examples of alleged dream sleeptalking by toddlers who have been overheard by adults, such as Freud's (1900, p. 130) account of his daughter dreaming at 19 months about the "stwawbewwies, wild stwawbewwies, omblet, pudden" she could not have during the day due to an upset stomach, they are more likely the product of the numerous mini-awakenings that are known from sleep laboratory studies to occur throughout the sleep period (e.g., Arkin, 1981; Boselli, Parrino, Smerieri, & Terzano, 1998).
The developmental findings on low dream recall from REM awakenings before age 8 also raise questions about the activation-synthesis theory of dreaming because of the strong relationship it posits between REM sleep and dreaming. The theory does not lead to the expectation that there could be REM sleep without dreaming in children any more than that there could be dreaming in adults without REM sleep. While the theory could be adjusted to claim that REM sleep is necessary but not sufficient for dreaming in young children, Foulkes's (1982) findings were of enough concern for Resnick, Stickhold, Rittenhouse, and Hobson (1994) to counter with their own sample of home-collected dreams from 8 children ages 4-5 and 6 children ages 8-10, and to suggest that his results were due to the inhibitory effects of the laboratory and/or an unwillingness to report dreams to relative strangers. This critique led to a spirited rebuttal by Foulkes (1996b), who emphasized that he had the same findings with home-collected dreams from the children who had slept in his laboratory.
If the work from the sleep laboratory is taken as the necessary starting point, then dreaming has to be viewed as a cognitive achievement that only gradually approaches adult levels later in childhood, an achievement that correlates with the development of visuospatial abilities (Foulkes, 1993). This unexpected finding not only undercuts Freudian, Jungian, and activation-synthesis theories, but elevates a cognitive theory because it suggests that (1) the ability to imagine without environmental stimulation and context, and (2) the development of a cognitive conception of a "self" are necessary to the kind of dreaming taken for granted by the other theories (Foulkes, 1999).
The similarities between dreaming and waking
There are now several different types of evidence suggesting that there is considerable overlap in dreaming and waking cognition. First, there is evidence that deficits in waking thought are also present in dreaming. In one of the first studies of this kind, the loss of the ability to visualize within waking thought corresponded with a loss of visual imagery in dreams, leading to dreams in which the person had the subjective experience of dreaming, interacted with dream characters, and developed a story line, but did not "see" the dream (Kerr, Foulkes, and Jurkovic, 1978). Other cases of this type are reported in Kerr (1993) and Solms (1997).
Second, there is good evidence that speech patterns in dreams are as grammatically correct as they are in waking life (e.g., Heynick, 1993; Meier, 1993). Third, studies of relaxed waking thought, whether monitored by the EEG in the darkened rooms of a sleep laboratory (Foulkes & Fleisher, 1975; Foulkes & Scott, 1973), or sampled periodically during the day by signaling the participants with a pager (e.g., Klinger, 1990; Klinger & Cox, 1988), suggest that waking thought can sometimes be as disconnected and "dreamlike" as night dreaming. Conversely, studies of the reality of dream reports suggest that a large percentage of them are not as bizarre and disjointed as activation-synthesis theory seems to suggest (Foulkes, 1985, 1996a).
Fourth, there is evidence that waking cognition is more "symbolic," in terms of the pervasiveness of metaphoric thinking, than previously has been realized (Gibbs, 1995; Lakoff, 1987, 1993; Lakoff & Johnson, 1980). If the symbols that Freud and Jung claim to be so prevalent in dreams are thought of as a form of metaphoric thinking, a claim supported by empirical evidence in a study of slang by Hall (1953a), then the overlap between thinking and dreaming may be extensive. However, much work needs to be done on the degree to which dream content is metaphoric before too much can be made of this argument (Domhoff, 1996). For example, Hall's unpublished search for metaphors in one set of 100 dreams, using very generous criteria, led to the estimate that only 10-15% of dreams are metaphoric.
Taken together, these several different types of evidence for similarities between dreaming and waking cognition cast doubt on the Freudian, Jungian, and activation-synthesis theories because all three emphasize the differences between waking cognition and dreaming. Thus, there is little or no evidence for Freud's claims about a unique set of processes called "the dreamwork" that supposedly create dreams (Fisher & Greenberg, 1977, 1996). As for Jungian claims that there are universal symbols that express an inherited collective unconscious based in pre-human and human evolutionary history, these observations are more parsimoniously and plausibly encompassed by the evidence for gradual linguistic socialization into the vast storehouse of conceptual metaphors that is part of our cultural heritage (Lakoff, 1987; Lakoff & Johnson, 1981). Within this cultural/linguistic framework, it is possible to do careful studies that reveal the extent of cross-cultural similarities. For example, the conceptual metaphor "Anger is Heat in a Container" is present in both English and Chinese, which is why the speakers of both languages can understand metaphors like "flipped his lid," "letting off steam," "steam coming out her ears," and "boiling over" (Yu, 1999).
The Continuity Hypothesis
From the beginning of modern-day dream research by psychologists who wrote down their own dreams in the late 19th century, systematic studies have supported the idea that there is considerable continuity between dream content and waking cognition (Calkins, 1893; Weed & Hallam, 1896). This finding was repeated in the few laboratory studies of the 1960s that analyzed correlations between dream content and objective personality measures. For example, Foulkes and Rechtschaffen (1964) and Foulkes, Larson, Swanson, and Rardin (1969) reported that young males with the most unpleasant dreams also tended to have the highest scores on MMPI psychopathology indicators.
The continuity between dream content and waking thought and/or behavior is one of the most striking findings from Hall's content analysis studies. For example, people dream most often about the individuals, pets, and interests that preoccupy them in waking life. They show the most aggression in dreams toward the people with whom they have the most conflict in waking life. The results are so consistent for these kinds of continuities that Hall adopted the term "continuity hypothesis." His blind analyses of the dream of Kafka, a child molester incarcerated in a prison hospital, a neurotic patient in psychotherapy, and numerous average people who kept dream journals provided the major evidence for this hypothesis (Bell & Hall, 1971; Domhoff, 1996, chap. 8; Hall & Lind, 1970; Hall & Nordby, 1972; Schneider & Domhoff, 1999).
Several of the empirical papers in this special issue add support for the continuity hypothesis. The findings on gender differences by Barrett, Land, and Schneider (1999) once again show that dream content corresponds with gender roles in American society. Saline (1999) notes that the gender differences she found in the Most Recent Dreams of children between the ages of 8 and 11 fit with findings on the emergence of gender differences in this age group. Strauch and Lederbogen (1999) find that the number of same-sexed peers increases in dreams at about the same time that young adolescents become more peer oriented and socialize in same-sex peer groups.
There is experimental evidence for the influence of waking concerns on dream content that supports the continuity hypothesis. In a study of 10 participants in a sleep laboratory, suggestions related to the participants' current concerns influenced dream content more than other suggestions (Nikles, Brecht, Klinger, and Bursell, 1998). However, a continuity between waking thought and dreams is not always immediately present, as Roussy, Camirand, Foulkes, De Koninck, Loftis, and Kerr (1996) demonstrated in a laboratory study that attempted to match presleep thoughts and significant daytime concerns with dream content from the first four minutes of the first REM period of the night. Independent judges were not able to match presleep thought samples or significant daytime concerns with the brief dream content beyond a chance level. These findings show the limits of the continuity hypothesis, but they are not necessarily incompatible with the Nikles et al. (1998) study or studies based on many dreams from a single individual, such as those summarized in the three previous paragraphs.
The Hall/Van de Castle findings on continuity present a challenge to Jung's "compensation hypothesis," which is second only to the idea of a "collective unconscious" as a unique contribution of Jungian theory. Whereas Jungian theory would predict that extraverted, outgoing people who put their energies into physical activities would have more introspective and thoughtful dreams, they in fact have dream content that is continuous with their waking behavior and interests. Jung (1963, 1974) often claimed that modern human beings are too externally oriented and neglect their spiritual needs, and demonstrated his point with dramatic dreams reported by patients, but no systematic studies support this claim.
The Consistency Hypothesis
There is considerable evidence that adults, unlike children as they grow up, are consistent in what they dream about over months, years, or decades, as defined by three types of consistency (Domhoff, 1996, chap. 7). "Absolute constancy" means that the score on a content indicator does not change in any consistent way over a given time period. "Relative consistency" means that two or more indicators remain in the same relationship with each other even though they all may rise or fall. "Developmental regularity" means that there is a consistent rise or decline on an indicator over a given time period.
The evidence for one or more of these consistencies is of two types, cross-sectional and longitudinal. Several cross-sectional studies, most in the United States, but one in Canada, suggest that dream content shows absolute constancy as people grow older, with the possible exception of declines in aggression and negative emotions (e.g., Brenneis, 1975; Hall & Domhoff, 1963, 1964; Howe & Blick, 1983; Lortie-Lussier, 1995; Zepelin, 1980-1981, 1981). While the longitudinal studies are not large in number, they are very similar in their results. For example, the few deviations from absolute constancy in a journal kept by a woman for 50 years were in such minor matters as fewer mentions of travel and ill health as she grew older, both of which were in keeping with her later life, when she traveled less and enjoyed better health (Domhoff, 1996; Smith & Hall, 1964). Several unpublished analyses by Hall with similar results are summarized in Domhoff (1993; 1996, chap. 7).
Hall (1984) also demonstrated absolute constancy in a study of the male/female percent in 11 dream series. The sample included 8 men and 3 women. The series spanned months, years, or decades. The subsets ranged from 50 to 122 dreams, and were taken from the beginning and end of the series when there were several hundred dreams. Only one person had a statistically significant change. The largest non-significant difference was 9 percentage points. Seven of the 11 dreamers showed differences of 4 percentage points or less. The one statistically significant exception was the woman whose consistency was discussed in the previous paragraph. Her male/female percent fell from 53/47 in the years 1912-1933 to 39/61 in 1960-1962, a drop of 14 percentage points, when she was 73-75 years old. Perhaps this change makes sense in terms of the "continuity hypothesis" because most of the male peers with whom she associated had died and she was living in a women's retirement home during the last years of her life. In short, there was a "developmental regularity" that was continuous with her waking life.
A dream series from the summer of 1939 used by Hobson (1988) to demonstrate the bizarreness, distortions, and scene changes emphasized in the activation-synthesis theory also shows great absolute constancy on Hall/Van de Castle indicators. Hobson recognizes the consistency of at least some content by calling the dreamer the "Engine Man" because he dreamed so frequently about the train locomotives that were a dramatic part of the landscape for the first half of the 20th century. However, Hobson then turns his attention to the unusual structural features of the dreams. Nonetheless, a content analysis of the 187 dreams in the journal with 50 or more words revealed that the first half hardly differed from the second half, a demonstration of consistency within the time frame of three months. Moreover, the atypical high and low indicators in his h-profile corresponded to his waking interests and personality, as revealed in a four-page obituary in a scientific publication (Domhoff, 1996, pp. 147-150; Schneider & Domhoff, 1999).
Within this context of consistency, a study of dreams from women before, during, and after menopause, who ranged in age from 30 to 68, is of special interest because it was designed to test ideas derived from Jungian theory concerning changes in dream content during and after menopause (Abel, 1994). The investigator created seven theoretical scales derived from Jungian theory that were used by three coders who were naive about Jungian theory and blind as to the age of the dreamer. An expert on Jungian theory coded all dreams on a scale for the degree of archetypality.
There were 32 women who were still menstruating, average age 39.6; 24 women who were perimenopausal, average age 49.2; and 20 women were postmenopausal, average age 58.6. Women in the first group averaged 32 dream reports over a one-month period; women in the second group averaged 21; and women in the third averaged 16. The only significant content difference concerned an "initiation" scale within an overall "transition" scale, but the difference was not for the predicted group. The codings by the Jungian expert did not show any difference in the degree of archetypality in the three samples. Since several comparisons were made, it would not be surprising if the one difference was due to chance. Moreover, the low reliability of theoretical scales (Domhoff, 1996, 1999) strongly argues that the reliability of the initiation scale would have to be demonstrated with new coders and the findings replicated before this one positive outcome could be considered a reliable finding. In short, this is yet another study showing little or no difference in dream content with age.
The absolute constancy of dream content in mature adults, especially as revealed in longitudinal studies, casts doubt on the important Jungian claim that individuation, integration, and the increasing role of the self archetype in adult life can be revealed through the study of dream content (Jung, 1963, 1974). The lack of change in Abel's (1994) test of Jungian theory with women before, during, and after menopause is especially interesting in this regard. Nor is consistency in dreams over time something that would be predicted by activation-synthesis theory, although the theory does not rule out the possibility.
When this brief overview of well-established descriptive empirical findings is matched with the Freudian and Jungian theories of dreams, neither theory fares very well. After nearly 100 years of case studies, along with a few positive results with theoretical rating scales of questionable value (Domhoff, 1999), there is little reason to believe specific hypotheses put forward by either Freud or Jung. Foulkes's (1982, 1999) developmental findings may be enough in themselves to force the abandonment of their theories, but the other content-based findings presented in this paper, along with the lack of support from most other types of systematic dream studies, make the case for abandonment much stronger.
Instead of supporting the unique claims of either Freud or Jung, these findings suggest that a theory of dreams has to fall back to the common general point in the two theories, which is that dreams tend to be about what variously have been called "emotional preoccupations," "personal concerns," and "unfinished business." These are not ideas to be scorned or downplayed, and Freud and Jung can be honored for putting them on the scientific agenda. Moreover, if it can be shown in the future that at least some dreams are metaphoric in nature, then Freud and Jung also could be credited as interpreters of metaphoric expressions that reveal some of the general problems of human existence. But an emphasis on "emotional preoccupations" is not as dramatic as many of the claims within Freudian and Jungian theory, and dreams as metaphors does not seem as profound as claims about disguised oedipal wishes or wisdom from the collective unconscious.
Generally speaking, the systematic findings on dream content are not as hard on the activation-synthesis theory as they are on Freudian and Jungian theory. Instead, the most difficult findings for it concern the occurrence of dreams at sleep onset and during NREM sleep late in the sleep period, when it is difficult to tell REM reports from NREM reports (Antrobus, Kondo, Reinsel, & Fein, 1995; Domhoff & Schneider, 1999; Foulkes, 1966). The lack of correlation between any of the phasic events of REM sleep and content variables, despite many efforts to find them, also is a difficult objection for the theory to overcome (Foulkes, 1996; Pivik, 1978, 1986). However, the content findings do add to the case against the activation-synthesis theory. The low levels of dreaming in the REM sleep of children up to age 8, the similarities of waking and dreaming cognition, the realistic nature of much dream content, and the consistency of dream content over years and decades in adults are not findings that fit easily into a theory that sees dreams as the best that the cortex can do in reaction to periodic bombardments by random stimuli from the pontine region of the brainstem.
On the other hand, all of these findings are compatible with a cognitive theory of dreaming and dream content. The developmental nature of most cognitive processes has been a basic tenet within cognitive psychology for many decades. Findings on the overlap between thinking in the waking and sleep states come as no great surprise to this theory, and the fact that dream content seems to reflect emotional preoccupations fits nicely with studies of undirected thinking in the waking state (e.g., Klinger, 1990; Klinger & Cox, 1978; Singer, 1988, 1993).
If this paper is convincing in suggesting that current evidence favors greater development of an open-ended cognitive theory based on the idea that dreams express our conceptions and emotional concerns, then the methods and findings in the papers in this special issue provide the foundation for hypothesis-testing studies that could extend and lend nuance to the theory. For example, could large-scale studies of the Most Recent Dreams of children and teenagers between the ages of 10 and 17 tell us anything about the developmental nature of dreaming? Are there new findings on cognitive changes in adolescence that might be extended to the understanding of dreaming? Is the emergence of greater interest in peers regularly reflected in dream content in adolescents, as the findings by Strauch and Lederbogen (1999) suggest, thereby supporting the continuity hypothesis?
Are there types of dream content, such as physical aggression, that are more continuous with waking thought than with waking behavior? Can content analysis lead to criteria for distinguishing metaphoric from nonmetaphoric dreams? Can the theory be advanced by drawing inferences about waking concerns from blind analyses of lengthy dream journals kept by average people (Domhoff, 1996, chapts. 7 and 8)? Can claims about concerns, consistency, and continuity be tested with special populations and individuals, as in the studies in this special issue by Hurovitz, Dunn, Domhoff and Fiss (1999) and Kirschner (1999)? Would the Most Recent Dreams of a large population of obsessional neurotics change after they were on medication for a given number of months, as monitored during their periodic visits to the clinic where they were being treated?
These are only a few of the many modest questions that could be studied with the aim of developing a cognitive theory of dreams using the methods, statistics, and strategies of data collection demonstrated in the seven empirical papers in this special issue on dream content. These questions add up to an opportunity for dream researchers to develop a common focus and communal effort similar to what emerged among sleep researchers after the discovery of sleep stages in the 1950s.
Thanks to David Foulkes, Don Kuiken, Andrew Neher, Tore Nielsen, Adam Schneider, and Richard Zweigenhaft for their helpful comments on this paper.
Go back to the Dream Library index.