Contents

1. Method
2. Subjects
3. Procedure
4. Results
5. Discussion
6. Possible Mechanisms of Dream Sharing
7. Implications of Dream Sharing for the Origin and Interpretation of Dreams
8. A New Theoretical Framework for the Interpretation of Dreams
9. APPENDIX
10. Verbatim Transcription of Shared Dream Pair 160/172
11. REFERENCES

ABSTRACT. Dream reports were tape recorded each morning independently for 10 weeks by both members of an adult heterosexual couple who alleged they regularly shared dreams. The transcribed reports were evaluated blind by 12 trained analysts using quantitative dream-content analysis. The couple had identified 13 pairs of dreams (17% of the sample) as shared. The mean percentage content overlap score of these 13 pairs (39.15%) was significantly greater than that of 80 randomly matched pairs of unshared dreams (5.23%). The content overlap scores of each analyst correlated significantly with those of the unblinded male subject (r = .72). The mean probability that the observed overlap in putatively shared dreams was caused by chance was calculated as 1 in 5 billion. Shared dreams appeared to occur in a distinct temporal cycle of 30-35 days. Content overlap between shared dreams was greatest in three of seven categories; objects, themes, and effect.

THE EXPERIENCING OF DREAMS of similar or identical content by two or more persons while sleeping has been reported anecdotally for a long time (Jung, 1974). Lincoln (1935) described shared dreams among tribal societies from the Solomon Islands, the West Solomons, and British New Guinea, and psychoanalysts in developed countries have reported shared dreams between themselves and clients and between clients. Reports of dream sharing are therefore frequent, old, and cross-cultural.

The first systematic attempt to collect and analyze shared dream reports began in the 1880s, when the English Society for Psychical Research collected dozens of accounts of extrasensory dreams to demonstrate various psychic phenomena (Hutchinson, 1901; Mason, 1899). Shared dreams drew little external attention, however, until Freud’s publication of three papers on the topic early in the 20th century.[1] Freud was skeptical and even hostile to such accounts, which had become associated with telepathy, and declared the phenomenon of shared dreaming irrelevant to psychoanalytic theory (Freud, 1925, in Devereux, 1953, p. 85). Indeed, Freud asserted that a shared dream is not really a dream: “Let us call it instead a telepathic experience in a state of sleep. A dream without condensation, distortion, dramatization, above all, without wish fulfillment, surely hardly deserves the name” (Freud, 1925, in Devereux, 1953, p. 77).[2]

Although Freud insisted that shared dreams could not inform psychoanalysis, he argued conversely (Freud, 1925, in Devereux, 1953, p. 85), which was interpreted by some of his contemporaries as an invitation for psychoanalytic investigation of extrasensory phenomena (Fodor, 195 1 ). Reports soon followed, detailing instances of presumed telepathy between analyst and client, including specific cases of dream sharing between analyst and client or between clients (Eisenbud, 1947; Fodor, 1947; Stekel, 1943). Deutsch (1926) suggested that dream sharing might represent a special type of “counter-transference” resulting from the unusually close intuitive bond between analyst and client. Similarly, Fodor (1951) proposed that strong emotional ties between subjects may be required for dream sharing, whereas Hitschmann (1933) emphasized the role of the intensity of the transmitted thoughts or images.

Reflecting the general lack of scientific rigor in most of these reports, however, Ellis (1947) suggested that they proved little more than the power of biased imagination when applied to coincidence. Proponents responded in kind (see Devereux, 1953), but the debate did not stimulate empirical research capable of clarifying or resolving the underlying issues.

The emphasis on telepathy as a mechanism of shared dreams waned, and more recent accounts of dream sharing stress interplay between individuals (Hunt, 1989) or counter-transference (in the psychoanalytic tradition) (Bernstein, 1987). Several other investigators have reported that people occasionally, and perhaps regularly, have similar dreams (Calogeras, 1977; Graves, 1924; O’Flarety, 1984; Shor, 1988; Ullman, 1981; Wolff, 1952). Jung asserted that “the authenticity of this phenomenon can no longer be disputed” (Jung, 1974, p. 47). Dream sharing has never been demonstrated conclusively, however, using controlled, quantitative methodology, blind measurement, and appropriate statistical analysis. Indeed, the largest source of current activity regarding shared dreams is the community of “dream workers,” whose interest is generally nonempirical and even purposefully nonscientific (Reed, 1985, p. 19).

Dream sharing, if properly documented, could have implications not only for the origin and meaning of dreams, but also for related topics such as interpersonal communication and linguistic symbolism. Therefore, when we learned of a couple who alleged that they frequently experienced similar dreams, we solicited their cooperation in evaluating their claim rigorously. We considered a priori that dream sharing would be demonstrated conclusively for this couple only if two general criteria were realized: Several blinded analysts must independently confirm significant overlap in the content of the putatively shared, but not unshared, dreams, and this overlap could not be explained by chance. Both criteria were met in this case study, which therefore provides the first rigorous corroboration of dream sharing.

Method

Subjects

The subjects were a heterosexual couple who agreed to participate in the research on the condition that their anonymity was preserved at all stages of the project. Both were healthy, successful, middle-aged, monogamous Caucasians with no medical history of alcohol or drug abuse. Both were also professionals, although active in different fields and at different developmental stages in their careers. The subjects were of similar socioeconomic backgrounds (upper middle class), similar educational levels (college graduates), and held similar political views (self-reported as liberal/progressive). At the time of the study they had known each other for less than 1 year and were unmarried cohabitants of 2 months in the early stages of an intimate personal relationship.

Procedure

Every remembered dream of both subjects was tape recorded independently by each subject within 1 hour of awakening, frequently under the supervision of the senior author, for a period of 10 weeks. The subjects were instructed in advance to report as much detail as they could remember about each discrete dream recalled from the previous night. The subjects did not discuss their dreams with each other before the dreams had been tape recorded independently, but after making the recording they discussed the previous night’s dream(s) with each other and recorded a commentary in which they identified putatively shared dreams.

A total of 153 dreams were recorded by the couple, of which 72 were recorded by the woman. The recorded dreams were subsequently transcribed by computer and printed as dream reports of approximately 100 to 2,000 words each that represented the raw data of this study. The dream reports were edited minimally by the male subject to preserve anonymity. Editing consisted mainly of substituting fictitious names of persons and places and/or deleting personal details that could have compromised the couple’s identity, without, however, altering the substantive content of the dream report. Editing never entailed adding new material to the report.

The written and edited dream reports were grouped into 93 pairs, including 13 dreams identified by the couple as shared (17% of the total sample) and 80 randomly paired nonshared dreams from the same (78%) or different (22%) nights. Statistical analyses were performed on all the dream pairs and same-night dream pairs with identical conclusions, but only the analysis on all dream pairs is reported here.

The 93 pairs of dream reports were distributed to 12 undergraduate student analysts without revealing to them the identity of the putatively shared dreams. The analysts had been trained and tested previously in the dream content analysis methodology of Hall and Van de Castle (1965), and received course credit for their work. In advance of scoring the dream reports of the couple, interanalyst reliability (which usually exceeded 90%) was tested repeatedly using the Hall and Van de Castle content analysis methodology applied to scored reference dreams.

Each of the 93 pairs of dream reports (both putatively shared and unshared) was analyzed blind by each of the 12 analysts by scoring overlap in content between them in seven specific categories, modified from Hall and Van de Castle (1965) in consultation with the latter author (personal communication). The seven categories were setting (location), specific objects (e.g., cars and animals), characters (people appearing in the dream), themes (central topics), plots and/or subplots, affect (emotional content), and effect (specific actions and events). On the basis of content-overlap scores in the seven individual categories, an overall overlap score was assigned to each dream pair by each analyst.

Overlap between the dreams in all eight categories (seven specific and one overall) was scored by marking on a linear scale of constant length whose two extremes were labeled 0% and 100%. The percentage of overlap was subsequently measured, calculated, and normalized across the cumulative scores of each analyst as the percentage of the maximum score assigned.[3] Each of the 12 analysts also recorded on each score sheet the specific item(s) of dream content (if any) on which overlap occurred for subsequent analysis of the categorical basis of the overlap and the calculation of the joint probability of overlap.

Each of the 12 analysts scored all 93 pairs of dream reports, generating 1, 1 16 score sheets. In addition, the male subject was trained in the same methodology and independently scored all 93 dream pairs. The resulting 1,209 evaluations constituted the reduced data of this study on which subsequent quantitative analysis was undertaken. Each analyst also recorded weekly the degree of her or his belief in the phenomenon of dream sharing, also on a linear scale labeled 0 to 100%. These data permitted subsequent evaluation of the possibility that the belief of the analysts was related to the scoring of dream content overlap.

Subsequent analysis was designed to test three hypotheses. The first was the overlap in content between putatively shared dreams (identified by the dreamers) would be significantly greater than overlap between randomly matched unshared dreams. The mean content overlap score of the 12 analysts was computed and compared for both putatively shared and unshared (randomly matched) dreams. The comparison was undertaken both for individual analysts and for the mean scores of all analysts. Assessment of the specific category in which the overlap of dream content took place was also determined within this first type of analysis.

The second hypothesis was that quantitative overlap scores assigned by the blinded analysts would correlate positively and significantly with overlap scores assigned by the male subject. This hypothesis, which was intended to assess whether unbiased observers also perceived the content of putatively shared dreams as similar, was tested by calculating Pearson correlation coefficients between the scores of blinded analysts and the male subject, both for individual analysts and for the grouped means of all 12 analysts.

The third hypothesis was that overlap in dream content revealed by the previous two criteria, if any, could not be ascribed to chance, as evidenced by a sufficiently low frequency of occurrence of the specific items for which content overlap was scored. The frequency of occurrence of specific overlap items (rather than categories) was measured using word- and phrase-searching software applied to the data files of the transcribed dreams. Overlap items were those that were identified by at least 2 (but usually more and often all) of the 12 analysts.

On this basis, the probability of chance overlap of the identified items was calculated as the joint probability (i.e., the product of the frequencies) of all the specific items for which overlap was scored. The items on which overlap was scored were typically more specific than the seven categories enumerated earlier. In the pair of shared dream reports reproduced in the Appendix, the category of objects was represented in both dreams by painted pictures, whereas the category of themes was represented by the visions of boiling, burning fire.

Two additional issues were evaluated retroactively. First, the temporal pattern of shared dreams was examined by plotting all overlap scores as a function of cumulative time for overall overlap and overlap in the seven specific categories. The purpose of this analysis was to reveal any nonrandom temporal structure that might characterize dream sharing. Because we had made no specific hypothesis, the evaluation was retroactive. Second, we evaluated the effect of the beliefs held by the analysts regarding the validity of dream sharing on the overlap scores they assigned, also by hypothesis-free, retroactive analysis.

Results

The mean overall content-overlap score of 13 putatively shared dreams (as identified by the dreamers) assigned by the 12 blinded analysts was 39.15% (SD = 19.56). The mean overall overlap score of the 80 randomly matched nonshared dreams was 5.23% (SD = 5.96). These two means were significantly different, one-tailed unpaired t(91) = 12.5 8, p = .00005.[4]

The overall overlap scores across the 93 sets of both shared and nonshared dreams were not correlated with the length of the dream report, r = .03, twotailed t(91) = 0.29, p = .5. Additionally, the mean length of the 13 shared dream pairs (65.1 lines, SD = 30.5) was not significantly different from the mean length of the 80 unshared dream pairs (56.7 lines, SD = 29.8), two-tailed unpaired t(91) = 0.935, p = .35. Therefore, differences in dream report length did not account for the difference in overlap scores.

Analysis of overlap scores of the 13 shared dream pairs, by specific categories, rather than by overall overlap scores as reported earlier, showed that overlap was strongest for the categories of objects and weaker for the other six categories (Figure 1). A univariate between-subject analysis of variance (ANOVA) between the overlap scores of the 12 different observers revealed a significant simple main effect of category on mean overlap score, F(6,72) = 2.63 3, p = .0052. Each mean was then compared with the smallest mean (that associated with characters) using the Dunnett procedure, which showed that the mean overlap score for objects (25.3%, SD19.3) was significantly greater than the mean for characters (12.5 %, SD9.0), one-tailed Dunnett t(12) = 2.69, p = .01. Similarly,the mean overlap score for themes (22.6%, SD = 13.6) was significantly greater than the mean for characters, one-tailed Dunnett t(12) = 2.11, p = .03; and the mean overlap score for effect (20.8%, SD = 15.3) was significantly greater than the mean for characters, one-tailed Dunnett t(12) = 1.857, p = .05. Remaining mean overlap scores (Figure 1) were not significantly different from the mean of characters, one-tailed Dunnett t(12) = 0.675-1.73, p > .05.

The correlation coefficient between the overall overlap scores assigned by the male subject and the corresponding mean scores of the 12 blinded analysts was r = .72, significantly different from zero, two-tailed t(91) = 9.84, p = .0001. For comparison, the mean correlation coefficient between scores of the 12 analysts for the same dreams (autocorrelation) was r = .65. The correlation coefficient between scores was therefore greater between the blinded analysts and the male subject than between blinded analysts evaluating identical dreams, although the difference between the correlation coefficients was not significant, Z(90) = 0.54, p = .59. Comparable correlations were significantly different from zero for each analyst individually (t tests in all cases, p < .05, Bonferroni adjusted), presumably reflecting both high interanalyst correspondence and the reliability of independent judgments.

Overlap of content between different dreams is meaningful only if the criteria for assigning such overlap are sufficiently infrequent in the entire dream set that chance overlap is unlikely. This issue has already been addressed indirectly by the greater overlap between shared dreams (39.15%) than randomly matched unshared dreams from the same couple (5.23%) as detailed earlier. Coincidental overlap in dream content is excluded explicitly here.

Toward this end, the frequency of those criteria used to establish content overlap was determined in the entire set of 153 dream reports and in the smaller set of 13 pairs of shared dreams (or 26 dream reports). The frequency of occurrence in the entire dream set of those criteria on which overlap was scored was low, often 1/153. The estimated population sample probability of chance overlap in putatively shared dreams, that is, the product of individually calculated frequencies, ranged from 4.3 X 10(sup -5) to 1.3 X 10(sup -20) (mean, 2.0 x 10(sup -10), i.e., from about 1 in 2,000 to about 1 in 1 billion billion.

For the subset of shared dreams (identified by the dreamers and confirmed by independent blind analysis), 46 specific overlap items were identified, of which four (approximately 9%) occurred in other shared dreams. Therefore, the content of each shared dream was virtually distinct from every other, reducing the possibility of random or chance overlap due to high frequency of overlap criteria.

These conclusions are illustrated qualitatively with reference to a specific dream pair, in which both the man and the woman dreamed of pictures and visions of burning or boiling fire (see Appendix). These elements occurred in no other dreams in the set. That is, these particular two dreams were novel. This pair of shared dreams received the highest overall overlap score of all shared dreams from the blinded analysts (M = 93%). Only two elements overlapped, objects (paintings) and themes (tire), however, and hence this dream pair also exhibited the highest probability of chance overlap (4.3 X 10 (sup -5)), owing to the small number of overlap items. This probability of chance overlap for this particular pair of shared dreams was nonetheless approximately three orders of magnitude smaller than the accepted statistical standard for rejection of the null hypothesis (5 x 10(sup -2).

The joint probability of occurrence of specific items of shared dream overlap (i.e., the product of frequencies of overlap items) decreased significantly as the length of the paired shared dream reports increased, r = -.62, two-tailed t(11) = 3.34, p = .008. This significant negative correlation is consistent with the increasing number of overlap items as dream report length increased, which yielded progressively smaller joint probabilities (the product of increasing numbers of fractions).

Cumulative plots of overall overlap of dream content as a function of time revealed an apparently nonrandom temporal clustering of shared dreams (Figure 2). Similar plots for each of the seven categories did not show clear clustering, although the relatively small sample size available within each category was probably incapable of revealing clustering even if it occurred.

The temporal cycle of dream sharing was approximately 30 days in length and correlated, perhaps coincidentally, with the menstrual cycle of the female subject, as deduced from the dates and duration of menstruation she provided retroactively. Maximal clustering of shared dreams corresponded approximately to the onset of menses in the woman.

The mean overall overlap score of the 12 analysts calculated weekly over the 8-week scoring period was not correlated significantly with the corresponding analysts’ mean level of belief in the phenomenon of dream sharing as averaged over the corresponding weeks, r = -.34, two-tailed t(10) = 1.143, p 10. Therefore, the personal beliefs of the analysts did not influence their scoring of dream pairs. Indeed, the weak (but insignificant) negative correlation suggests the trend that increasing belief in dream sharing induced cautionary weaker scoring of overlap.

Discussion

The data and analyses reported here demonstrate frequent dream sharing, entailing approximately one fifth of the total sample of 153 dreams in a single case study of one couple’s dreams recorded over a 10-week period. The overlap in content between randomly selected unpaired dreams (5.23%) was significantly less than content overlap in shared dreams (39.15%), suggesting indirectly that the overlap cannot be explained by chance. The possibility of chance overlap was evaluated directly by measuring the frequency of occurrence of specific overlap items and determining their products. These ranged from 1 in 2,000 to less than 1 in 1 billion billion (M = 1 in 5 billion). Additionally, shared dreams occurred at apparently regular intervals in perhaps coincidental synchrony with the female subject’s menstrual cycle. These findings suggest that mechanisms other than chance are responsible for the observed overlap in dream content.

This demonstration of dream sharing raises questions about the underlying mechanisms. The findings also have implications for contemporary dream theory, including the origin and interpretation of dreams. The purpose of the following discussion is to illustrate that (a) dream sharing can be explained, in principle, by relatively straightforward mechanisms (i.e., psychic explanations are unnecessary although not excluded); (b) dream sharing is inconsistent with existing theories of the role of dreams, and therefore, (c) dream sharing requires a new theoretical framework, one of which is briefly outlined here.

Possible Mechanisms of Dream Sharing

As evidenced in the historical account of dream sharing presented in the introduction to this article, past reports exhibited confusion between the phenomenon itself and the underlying mechanisms. In particular, dream sharing (the phenomenon) was widely accepted as evidence of telepathy (one possible mechanism among many)-an inaccurate identity that, in hindsight, impeded empirical research on the phenomenon itself.

Although the present data do not bear directly on the mechanism of dream sharing, several straightforward possibilities are consistent with the data, and the data enable preliminary discussion of each. Perhaps the most straightforward possible mechanism of dream sharing is similar or common daytime experiences by the two dreamers, day residue (Freud, 1925; Hall & Van de Castle, 1965). The specific content of shared dreams in the present study seldom bore an obvious relation to plausible real events, however (e.g., the “fire” dream pair reproduced in the Appendix). Thus, if day residue does contribute common elements to nighttime dreams, an additional symbolic conversion from waking to dream events (”distortion,” in Freudian terminology) must take place in one or both dreamers; and this symbolic conversion must be similar in both dreamers. Arguing against a prominent role for day residue is the fact that the two subjects of this study typically had different, if overlapping, daily experiences.

A second possible mechanism for dream overlap is common sensory inputs during sleep. Berger (1963) has shown that meaningful verbal stimuli can influence dream content, and common sensory experiences of other modalities may likewise induce common dream experiences. If this is the case, however, the apparent disparity between shared dream content and plausible sensory input during sleep again suggests the intervention of a symbolic conversion, or distortion, common to both dreamers.

A third and fourth possible and related mechanisms for dream sharing are common psychologies, that is, the realignment of unconscious symbolism so that they match in the two dreamers (Perlmutter & Babineau, 1983), and/or enhanced “intuitive understanding” that accompanies pair bonding (Hunt, 1989; Jung, 1974). These possible mechanisms are not mutually exclusive from each other or from the first and second possible mechanisms. Indeed, the third and fourth mechanisms could underlie the “symbolic conversion” implicit in both the first and second. In this case, the study of shared dreams could lend insight into the related issue of linguistic symbolism.

A fifth possible mechanism of shared dreams is telepathy, the extrasensory transmission of mental activity between individuals, as was widely assumed in the earlier history of shared dreams (Ehrenwald, 1942; Eisenbud, 1947; Fodor, 1947, 195 1; Mason, 1899; Stekel, 1943). This mechanism would appear the least likely, because the modem neurosciences provide no known physical basis. The present data provide no direct evidence for telepathy or, for that matter, for any other mechanism of dream sharing, although neither do they rigorously exclude any specific mechanism as a contributor to shared dreams.

Implications of Dream Sharing for the Origin and Interpretation of Dreams

Prior to Freud’s introduction of the unconscious, dreams were conceived as nonsensical collections of random mental impressions arising from somatic stimulation. The dream was considered a spontaneous, undirected, and apparently purposeless mental event, constrained only by the intrinsic organization of the mnemonic systems from which it emerges (Hobson, 1989). In contrast, Freud conceived of dreams as nonrandom structures rich in meaning that originate in the neuroses, fears, and/or latent wishes of the dreamer (Freud, 1952). Contemporary models have revisited pre-Freudian concepts, by considering dreaming as random mental noise that is interpreted artfully (Hobson, 1989; Foulkes, 1985) or “mental detritus,” discarded by the brain as irrelevant information (Crick & Mitchison, 1986).

There are, however, dreams that fit none of the above categories. Creative dreams, for example, such as Lowei’s famous dream about the experiment that led to the discovery of chemical synaptic transmission in the frog heart (Loewi, 1960), suggest a possible problem-solving, goal-directed function for dreams. Similarly, intuitive dreams have been interpreted as enhancing empathy between two or more individuals (Hunt, 1989). Neither these nor other dream categories have been subjected to systematic analysis that could disconfirm their function, however, and therefore they do not furnish a rigorous challenge to contemporary models of dreaming.

In contrast, dream sharing that cannot be explained by chance overlap of content, as documented here, implies a common origin of dreams in separate individuals. Common origin of dreams in different individuals in turn cannot be reconciled easily with existing theories on the origin and meaning of dreams. For example, Freud’s interpretation of dreams is difficult to reconcile with shared dreaming. If two individuals have the same dream, the dream could not have originated in the latent wishes, fears, or neuroses of both, except under the implausible hypothesis that latent wishes, fears, or neuroses are identical in both individuals. A shared dream must therefore originate independent of the latent wishes, fears, or neuroses of at least one of the dreamers. As noted by others, “. . . telepathic [shared] dreams posed a special threat to Freud … their existence would oppose the claim that dreams are a determinate result of psychological processes, like symptoms” (Friedan, 1990, p. 98).

Neither can shared dreams be explained as pseudo-random mental noise (Foulkes, 1985; Hobson, 1989). Shared dreams exhibit common and highly improbable content overlap, and the content cannot therefore be random or pseudorandom in both dreamers. Likewise, shared dreams appear inconsistent with the “mental detritus” theory (Crick and Mitchison, 1986), for similar dream content in the two dreamers would then imply correspondingly similar mental detritus. Moreover, if any dream (shared or unshared) were irrelevant “neural noise,” as required by the mental detritus theory, the subsequent recall and discussion of the dream implies the storage of irrelevant information. In this case dreams would have the counterproductive role of generating useless new information that occupies presumably finite memory space at the presumed expense of more adaptive information-an unlikely and unsatisfying explanation for a human behavior as universal as dreaming.

A New Theoretical Framework for the Interpretation of Dreams

The incapacity of contemporary dream theories to explain dream sharing suggests the need for a new theoretical framework. In formulating such a framework, we assume that shared dreaming, in common with other evolved behavior, conferred selective reproductive advantage during evolution and evolved in early humanoid, prehistorical tribal societies. In this context, we speculate that shared dreams and their subsequent discussion and evaluation by the dreamers contributes to interpersonal bonding-pair (in the case of the present study), familial, social, tribal, cultural-which could in turn contribute to group cohesion, enhanced individual survival, and ultimately to reproductive success. Under this hypothesis, shared dreaming would have reinforced the social bonding that presumably assisted the survival of early humanoid societies. Seen in this perspective, shared dreaming would meet the necessary and sufficient criteria for evolution through natural selection.

This evolutionary social bonding theory of dreams demands that a significant percentage of dreams are shared, as occurred in the present case study. This theory requires further that dreams are discussed among members of the corresponding social unit, as was the case for the couple studied here and is common among numerous contemporary tribal cultures (Domhoff, 1985; Lincoln, 1935; Margolin, 1978). Consistent with this theory, several authors have noted a strong relationship between dream elements and culturally shared mythology (e.g., Campbell, 1973; Jung, 1974; Roheim, 1952). The shared dream could contribute reciprocally to the emergence of mythology, which likewise might have served the interests of interpersonal cohesion within the corresponding social unit. Finally, the evolutionary theory of dreaming requires that shared dreaming contributes to bonding and enhanced reproductive success, which is plausible but untested.

In conclusion, this study demonstrates the regular occurrence of shared dreaming in a single couple over a limited period of time. The generality of the phenomenon is unknown, as are the underlying mechanisms. Further research on dream sharing could be directed toward replicating the phenomenon in a different and larger cohort, defining the circumstances under which it occurs, evaluating its frequency of occurrence in different cultures, exploring the underlying mechanisms, and testing parameters of the evolutionary model and/of alternative theoretical frameworks.

APPENDIX

Verbatim Transcription of Shared Dream Pair 160/172

Dream 160 (Male Subject’s Dream Report)

I was looking through a bunch of pictures and paintings that S. [young daughter] had made. They were big pictures on big papers, and I was turning them. There was one that was done in red lines, but very complex, and the lines all together made a pattern, and the pattern was something like a-I’m not sure what it was, it looked like red and orange boiling fire, and in the comer was a streak of orange/yellow. I had the feeling it was a man, like a god-like man in the sky blowing. I was very impressed with it, because I thought S. made it, and then I realized she couldn’t possibly have made it; it was done by a professional artist. I was looking through this pile of pictures and wondering.

Dream 172 (Female Subject’s Dream Report)

I had a dream that involved fire, and that was a bright red and orange and yellow painting that I kept coming back to, and I would kind of glance out of the comer of my eye and finally I got the word “fire” and it was kind of out of my peripheral vision, and I kept seeing this painting. It became fire and the whole room became fire I think. It was kind of like something you glance at, and then you glance at it again, and then you glance at it again and again and again and it becomes just everywhere. So that’s what it is that painting of red and orange and yellow, seeing that it was fire.

1. The papers, “Dreams and telepathy” (first published in 1922), “The occult significance of dreams” (first published in 1925), and “Dreams and the occult” (first published in 1933), are reproduced fully in Devereaux (1953).
2. Freud’s adamant and circular rejection of dream sharing suggests that he realized the phenomenon presented significant difficulties to his own theory of the origin and meaning of dreams, proposed also by Friedan (1990).
3. The purpose of normalizing each analyst’s score as a percentage of maximum score assigned was to avoid giving undue weight in the mean to the scores of any single analyst. In practice, the maximum scores of all analysts ranged from about 85% to nearly 100%, and hence the normalized scores differ by only a few percentage points from absolute scores.
4. The test was one-tailed because the hypothesis to be tested was directional (i.e., that shared dreams display greater overlap than nonshared dreams).

GRAPH: FIGURE 1. Histogram showing the distribution of content overlap in shared dreams across the seven specific categories for which overlap was scored, namely, objects (0), themes (T), effect (E), settings (S), plots/subplots (P), affect (A), and characters (C). Shown are the mean overlap scores across all analysts for each category for the 13 shared dreams in this study. Bars show standard deviations.

GRAPH: FIGURE 2. Cumulative time histogram showing overall mean content overlap score of matched dream pairs as a function of time. The mean content overlap score assigned by the 12 analysts is shown on the ordinate. Shared dreams are those whose overlap score exceeded 20%. Note discrete temporal clumps of activity recurring on an approximately 30-day cycle. The non-shared dream pairs shown here are necessarily limited to those that took place on the same night (78% of the total sample of data).

REFERENCES

Berger, R. J. (1963). Experimental modification of dream content by meaningful verbal stimuli. British Journal of Psychiatry, 109, 722-740.

Bernstein, K. (1987). When supervisor and therapist dream: The use of an unusual countertransference phenomenon. Journal of the American Academy of Psychoanalysis, 15, 261-271.

Calogeras, R. (1977). Husband and wife exchange of dreams. International Review of Psychoanalysis, 4, 71-82.

Campbell, J. (1973). The hero with a thousand faces. Princeton, NJ: Princeton University Press.

Crick, F., & Mitchison, G. (1986). REM sleep and neural nets. Journal of Mind and Behavior, 7, 229-249.

Deutsch, H. (1926). Occult processes occurring during psychoanalysis. In G. Devereux (Ed.), Psychoanalysis and the occult (pp. 133-146). New York: International Universities Press.

Devereux, G. (Ed.) (1953). Psychoanalysis and the occult. New York: International Universities Press.

Domhoff, W. G. (1985). The mystique of dreams: A search for utopia through Senoi dream theory. Berkeley: University of California Press.

Ehrenwald, H. (1942). Telepathy in dreams. British Journal of Medical Psychiatry, 19, 313-323.

Eisenbud, J. (1947). The dreams of two patients in analysis interpreted as a telepathic reve’a deux. The Psychoanalytic Quarterly, 16, 39-69.

Ellis, A. (1947). Telepathy and psychoanalysis: A critique of recent “findings.” In G. Devereux (Ed.), Psychoanalysis and the occult (pp. 297-338). New York: International Universities Press.

Fodor, N. (1947). Telepathy in analysis. In G. Devereux (Ed.), Psychoanalysis and the occult (pp. 283-296). New York: International Universities Press.

Fodor, N. (1951). New approaches to dream interpretation. New York: The Citadel Press.

Foulkes, D. (1985). Dreaming: A cognitive-psychological analysis. Hillsdale, NJ: Erlbaum.

Freud, S. (1925). Dreams and telepathy. In G. Devereux (Ed.), Psychoanalysis and the occult (pp. 69-86). New York: International Universities Press.

Freud, S. (1952). On dreams. New York: W. W. Norton.

Friedan, K. (1990). Freud’s dream of interpretation. New York: State University of New York Press.

Graves, R. (1924). The meaning of dreams. London: Cecil Palner.

Hall. C., & Van de Castle, R. (1965). The content analysis of dreams. New York: AppletonCentury-Crofts.

Hitschmann, E. (1933). Telepathy during psychoanalysis? In G. Devereux (Ed.), Psychoanalysis and the occult (pp. 128-132). New York: International Universities Press.

Hobson, A. J. (1989). The dreaming brain. New York: Basic.

Hunt, H. (1989). The multiplicity of dreams. New Haven, CT: Yale University Press.

Hutchinson, H. (1901). Dreams and their meanings. New York: Longmans Green.

Jung, C. G. (1974). Dreams. Princeton, NJ: Princeton University Press.

Lincoln, J. S. (1935). The dream in primitive cultures. London: Cresset.

Loewi, O. (1960). An autobiographical sketch. Perspectives in Biology, 4, 3-25.

Margolin, M. (1978). The Ohlone way: Indian life in the San Francisco-Monterey Bay area. Berkeley, CA: Heyday.

Mason, R. O. (1899). Telepathy and the subliminal self New York: Henry Holt.

O’Flarety, W. D. (1984). Dreams, illusions and other realities. Chicago: University of Chicago Press.

Perlmutter, R., & Babineau, R. (1983). The use of dreams in couples therapy. Psychiatry, 46, 66-72.

Reed, H. (1985,September-October). Dreaming for Mary. Venture Inward, pp. 14-19.

Roheim, G. (1952). The gates of the dream. New York: International Universities Press.

Shor, B. (1988,January-February). Shared dreaming. Dream Network Bulletin.

Stekel, W. (1943). The interpretation of dreams. Vol. 2. New York: Liveright.

Ullman, M. (1981). Psi communication through dream sharing. Parapsychology, Review,, 12, 1-9.

Wolff, W. (1952). The dream-mirror of conscience. New York: Grune and Statten.

~~~~~~~~

By W. JACKSON DAVIS and MARCOS FRANK, Department of Biology, University of California, Santa Cruz

An expanded version of this study was submitted by Marcos Frank in partial fulfillment of the requirements for the Bachelor of Arts with Honors in Psychobiology at the University of California, Santa Cruz

We thank W. G. Domhoff and R. Berger for useful discussions and constructive criticism, Dotty Hollinger for transcribing the dream reports, Peter Coopersmith for assistance with data reduction and analysis, and the following individuals for patient hours of skilled dream content analysis: Natalie Barnett, Karen Chandeysson, Ryder Gwinn, Chris Henderson, Nadine Kadekian, Valery Kaplan, Lisa Leo, Tom McCabe, Jay McLaughlin, Laura Medina, Ralph Porras, Veronica Tonay, and Parker Waechte

Marcos Frank is currently at the Department of Biological Sciences, Stanford University, Stanford, CA 94305

Address correspondence to W. Jackson Davis, Biology Department, University of California, Santa Cruz, CA 95064.

A HUNDRED YEARS AFTER FREUD, ONE MAN MAY HAVE FIGURED OUT WHY WE DREAM. YOU’LL NEVER THINK THE SAME WAY ABOUT NIGHTMARES AGAIN.

Jay Dixit (2007, November). NIGHT SCHOOL. Psychology Today, 40(6), 88-94.

I THE DREAM ROBBERS

WHAT HAPPENS WHEN A RAT STOPS DREAMING? IN 2004, researchers at the University of Wisconsin at Madison decided to find out. Their method was simple, if a bit devilish. Step 1: Strand a rat in a tub of water. In the center of this tiny sea, allot the creature its own little desert island in the form of an inverted flowerpot. The rat can swim around as much as it pleases, but come nightfall, if it wants any sleep, it has to clamber up and stretch itself across the flowerpot, its belly sagging over the drainage hole.

In this uncomfortable position, the rat is able to rest and eventually fall asleep. But as soon as the animal hits REM sleep, the muscular paralysis that accompanies this stage of vivid dreaming causes its body to slacken. The rat slips through the hole and gets dunked in the water. The surprised rat is then free to crawl back onto the pot, lick the drops off its paws, and go back to sleep-but it won’t get any REM sleep.

Step 2 : After several mostly dreamless nights, the creature is subjected to a virtual decathlon of physical ordeals designed to test its survival behaviors. Every rat is born with a set of instinctive reactions to threatening situations. These behaviors don’t have to be learned; they’re natural defenses-useful responses accrued over millennia of rat society.

The dream-deprived rats flubbed each of the tasks. When plopped down in a wide-open field, they did not scurry to the safety of a more sheltered area; instead, they recklessly wandered around exposed areas. When shocked, they paused briefly and then went about their business, rather than freezing in their tracks the way normal rats do. When confronted with a foreign object in their burrow, they did not bury it; instead, they groomed themselves. Had the animals been out in the wild, they would have made easy prey.

The surprise came during Step 3. Each rat was given amphetamines and tested again; nothing changed. If failure to be an effective rat were due to mere sleep deprivation, amphetamines would have reversed the effect. But that didn’t happen. These rats weren’t floundering because they were sleepy. Something else was going on-but what?

II WHAT DREAMS ARE MADE OF

DREAMING IS SO BASIC TO HUMAN EXISTENCE, it’s astonishing we don’t understand it better. It consumes years of our lives, and no other single activity exerts such a powerful pull on our imaginations. Yet central as dreaming is, we still have no idea why we dream. Freud saw dreams as convoluted pathways toward fulfilling forbidden aggressive and sexual wishes; frightening dreams were wishes in disguise-wishes so scary, he believed, they had to transmute themselves into fear and masquerade as nightmares.

Later came the idea that dreams are the cognitive echoes of our efforts to work out conflicting emotions. More recently, dreams have been viewed as mere “epiphenomena”-excrescences of the brain with no function at all, the mind’s attempt to make sense of random neural firing while the body restores itself during sleep. As Harvard sleep researcher Allan Hobson puts it, dreams are “the noise the brain makes while it’s doing its homework.”

“There’s nothing closer to a consensus on the purpose and function of dreaming than there’s ever been,” says Deirdre Barrett, a Harvard psychologist and editor of the forthcoming The New Science of Dreaming. Indeed, no theory has been able to reconcile the findings of various subdisciplines of dream science. Until now.

Finnish psychologist Antti Revonsuo believes the marooned rats lost their ability to defend themselves not because they were exhausted but because they were robbed of their dreams. Dreams, he contends, are a training ground in which animals and people alike go over the behaviors that are key to their survival. Prevented from dreaming, the rats were unable to rehearse their survival behaviors. In other words, they were defenseless because they were out of practice.

III A THERTER OF THRERTS

SAY YOU’RE IN A FIGHT AND SOMEBODY wraps his arms around you from the front, pinning your arms to your sides-a bear hug. Most people reflexively stiffen their body. But this is actually the worst thing to do; making your body rigid makes you easier to lift-and lets your assailant pick you up and drop you on your head, or worse, haul you off somewhere. Better to bend your knees and lower your center of gravity so you’re harder to lift. You’re then free to punch your aggressor’s testicles, claw the skin on his back, kick out his knee, stomp his foot, even bite his neck-unappetizing options, but effective against even the biggest thug.

The difference between the typical and optimal response could save your life. But making such a reaction swift and automatic takes practice. It’s the reason martial arts students drill their movements over and over. Frequent rehearsal prepares them for that one decisive moment, ensuring that their response in an actual life-or-death situation is the one they practiced.

Dreams may do the same thing. A dream researcher at the University of Turku, in Finland, Revonsuo believes that dreams are a sort of nighttime theater in which our brains screen realistic scenarios. This virtual reality simulates emergency situations and provides an arena for safe training. As Revonsuo puts it, “The primary function of negative dreams is rehearsal for similar real events, so that threat recognition and avoidance happens faster and more automatically in comparable real situations.”

Faced with actual life-or-death situations-traffic accidents, terrorist attacks, street assaults-some people report entering a mode of calm, rapid response, reacting automatically, almost without thinking. Afterward, they often say the episode felt unreal, as if it were all a dream. Threat simulation, Revonsuo believes, is why.

IV A SEASON IN HELL

AS A GRAD STUDENT IN PSYCHOLOGY IN the early 1990s, Revonsuo often had bad dreams. What struck him the most was how lifelike they were. “I would say to myself, in my dream, ‘Oh shit! I’ve dreamt of this before, but now this is really happening!’” he recalls.

“Credible world analogs” are what cognitive psychologist David Foulkes calls dreams. Although we tend to dwell on the bizarreness of dreams, most dreams are quite mundane, Foulkes notes. You move around, talk, run, interact with others, experience emotions, and feel the passage of time, just as in everyday life.

When Revonsuo began studying dreams, he asked his students to start keeping logs of their own nocturnal escapades. He noticed something striking. The dreams were filled with dangerous events, negative emotions, monsters, chases, escapes, fights, and near-death experiences. The dream world was a hellscape of danger, teeming with threatening events far more sinister than in waking life.

These weren’t the misfirings of diseased brains. Threat dreams were the norm, accountingfor astaggering two-thirds of all dreams. Revonsuo discovered that we grossly underestimate the number of nightmares we have. As it turns out, we have 300 to 1,000 threat dreams per year-one to four per night. Just under half are aggressive encounters: physical aggression such as fist-fights, and nonphysical aggression such as verbal arguments. The rest are about car crashes, falling and drowning, missing a meeting or a test, being lost or trapped, and being naked in public. The whole dream world seemed to have a negative bias: more negative emotions than positive ones, more misfortune than good fortune, more nightmares than fantasy.

V A THEORY IS BORN

IN THE ANCESTRAL ENVIRONMENT, Revonsuo reasoned, our dreams served to protect us, teachingus how to respond when a wild animal was chasing us or when we got lost in the forest. That was why the dream world was so filled with peril: to simulate the potential threats and prepare us to react quickly. But how could dreams help us select the optimal response, given that dream recall is so fragile? After all, we remember only a few of our dreams, and even those fade fast in the tumult of the day.

Revonsuo believes that by providing rehearsal, dreaming helps us recognize dangers more quickly and respond more efficiently. We don’t need to be aware of this rehearsal, just as you don’t have to recall exactly where you practiced your tennis serve in order to reap the rewards.

The idea that dreams are a dojo for perfecting waking activities fits well with what is already known about practice. Mental rehearsal through visualization improves skills, enhances learning, and changes the brain, polishing performance in almost any domain, from sports to piano playing.

The single most pervasive theme in dreaming is that of being chased or attacked. Just as athletes in training repeat parts of their performance, we may, in our nightmares, be attacked and chased over and over again, not to solve a particular problem but to actually practice efficient escape behavior.

Saber-toothed tigers no longer stalk our villages, but Stone Age themes still rule our dreams. “Nowadays, the evolutionary footprint is clearest in the dreams of children, who often dream about being chased by monsters, much the same way we were once chased by predators,” says Revonsuo. As life has evolved, so have the threats we rehearse. “You insert a modern danger into that ancestral key and get a bizarre combination,” says Revonsuo. “We dream of being chased, shot, or robbed, getting into traffic accidents, a burglar in our house, or perhaps smaller mishaps such as losing our wallets-and that prepares us for our waking life.”

The dreamingbrain, explains Revonsuo, scans emotional memories. When it detects a memory trace with a strong negative emotion, it constructs a nightmare around that theme. The more traumatic the event, the more intense the nightmare. The brain’s system for detecting threats is sensitive and flexible: Anything the brain tags with a strong negative charge gets thrown into the threat bin and dredged up at night.

Sometimes this system works well: Dreaming about a boy running in front of our car better prepares us should that danger crop up in real life. But sometimes the modern world throws the threat-detection mechanism out of whack: Watching horror movies can trigger nightmares about vampires, ghosts, aliens, or zombies. Such “nonsense nightmares” don’t rehearse any useful threats; they’re like an allergic reaction, says Revonsuo. Just as our immune system can mistake pollen for a pathogen and mount a defensive campaign, the threat-detection system misperceives horror movies and deploys its defenses by generating a nightmare.

VI HEROES OF OUR OWN DREAMS

IN THE JUNGLES OF THE AMAZON LIVES a tribe called the Mehinaku. The Mehinaku lead the traditional life of hunter-gatherers. They spend their days fishing and gathering roots. Since they believe that dreams predict the future, they are scrupulous about remembering them and sharing them with others. That makes them perfect for an ethnographic study of dreams. In 1981, anthropologist Thomas Gregor surveyed their dreams and analyzed the content.

As it turns out, the Mehinaku dream profusely about the dangers in their everyday lives: being attacked by wild pigs; chased by jaguars; bitten by snakes; stung by wasps, ants, or bees-all potentially lethal. “Their dreams simulate over and over again what to do and how to do it quickly when they spot these animals in the wild,” reports Revonsuo. Across a tribesman’s lifespan, a single failure to react efficiently could be fatal. If threat simulation even marginally increases the likelihood that such fatal failures won’t occur, it would prove adaptive.

If the threat-simulation theory is correct, dreams should focus on the self, and when confronted with a threat, the dream self should react realistically to ensure its own survival and that of its loved ones. And so it is. We are the heroes of our own dreams. We don’t dream about other people’s adventures or about fictional superheroes battling monsters. We dream about ourselves.

If dreams evolved to simulate the threats in our environment, then being exposed to more dangers in real life should activate the nightmare function, overstuffing our dreams with threats. This is precisely what happens. Even a single exposure to a life-threatening situation can plunge a person into an inferno of post-traumatic nightmares, dreams in which the threatening event-the attack, the rape, the war-is repeated over and over in every possible variation.

Studies of traumatized Iraqi and Palestinian children who grew up in extremely violent environments, some of whom witnessed their parents’ deaths, show that their dreams are phantasmagoric carnivals of threatening events. People who watched more television on September 11,2001, and saw threatening images were more likely to dream about the events of that day; people who merely talked about it with others were less likely to dream about it.

Traumatic dreams do seem to rehearse relevant threats. Just four weeks into the first Gulf War, as Scud missiles were raining down on Tel Aviv and Haifa, the war was already encroaching on the dreams of Israeli college students, according to a study. The most prominent topic: gas masks.

But not all our dreams contain threats. That’s not surprising, says Revonuso. There’s no reason a biological system has to express its function at all times. Many bodily systems spring into action only in critical situations. Take sperm cells. The average man ejaculates over 100 million sperm at a time, yet over the course of his life, only a few will ever accomplish their biological mission of fertilizing an egg. Every day, millions of sperm are wasted-and while this may, as Monty Python sings, make God quite irate, it doesn’t mean that sperm cells have some function other than fertilizing eggs and competing with other sperm.

VII THE NIGHTTIME EDGE

INTRIGUING AS REVONSUO’S THEORY IS, not everyone is sold on the idea that dreams are primarily a theater of threat rehearsal. Dream researchers have known for centuries that dreaming helps problem solving, for example-but they still do not know why.

Some researchers argue that dreams are designed specifically to help us come up with creative solutions. But if that’s the case, it’s infuriatingly inconsistent-and complicated by the fact that we rarely remember our dreams.

Those who awake with brilliant solutions to scientific or artistic problems are the exception. German chemist Friedrich August Kekulé struggled to find the molecular structure of benzene until he dreamed about a snake devouring its own tail and realized benzene was a closed circle-a ring. The selftaught Indian mathematician Srinivasa Ramanujan came up with every one of his proofs in dreams. Paul McCartney dreamed “Yesterday,” woke up, and wrote it down.

Problem solving may be a side effect of the simulation system. The mere fact of running scenarios over and over may inevitably generate new solutions. That’s why when we have an important decision to make, we like to “sleep on it” first, why, according to a study by University of Maryland psychologist Clara Hill, couples who dream about their relationship are more likely to resolve their conflicts than couples who don’t.

It’s also known that we get better at tasks just by dreaming about them. Robert Stickgold, a sleep researcher at Harvard Medical School, has found that if you time people as they tap out the sequence 4-1-3-2-4 with their fingers, then ask them to do it again later that day, they are no better.

But let them sleep in between and their performance improves-literally overnight. The implication seems obvious: Sleep provides practice. People given brainteasers before bed dream about the answers. Math students are all too familiar with dreams about algebra problems. Anyone who’s ever played too much Tetris knows you can start having Tetris dreams.

Stickgold holds that dreaming is much more complex than rehearsal. He points, for example, to the ability of sleep to allow us to integrate and consolidate knowledge. Duringsleep, our brains are making sense of the world, discovering new associations among existingmemories, looking for patterns, formulating rules. “That’s how we create meaning,” says Stickgold. “Our brain puts things together.”

Dreams do have a certain edge over conscious thought Neuroimaging work has shown a distinct pattern of activation and inhibition in the dreaming brain. Visual and emotional centers are abnormally activated, while censoring mechanisms are deactivated. When we try to visualize during the day, imagery is thin and insubstantial, less real than the real world. But studies suggest that vivid hallucinations during dreaming rival the clarity and detail of vision itself.

“Dreaming is a sensitive system that tries to pay much attention to the threatening cues in our environment,” Revonsuo concludes. “Their function is to protect and prepare us.”

‘Yes,” says Harvard’s Barrett, “dreams are worrying about disasters. But they’re also planning for nice things and they’re fantasizing and they’re problem solving.”

She contends that the purpose of dreaming is “as broad as all waking thought. That’s why I say dreams are reallyjust thinking in adifferent biochemical state.”