The lack of thematic continuity in dreams with scene and plot discontinuities

Abstract

A central tenet of Freudian dream theory holds that there is thematic coherence within all dreams, even those containing scene and plot discontinuities. While other models support varying degrees of dream coherence, none address the question of how, or even whether, coherence can be identified in dreams with such discontinuities. Here, we objectively test the ability of judges to evaluate the coherence of individual dream narratives. Twenty reports with complete scene and plot discontinuities were collected, and half were cut apart at their discontinuities and their two halves spliced together with segments from dreams of other subjects. The remaining 10 reports were left intact. Judges correctly identified reports as intact or spliced only 57% of the time, a rate only slightly better than chance. Only 3 of the 20 reports, one intact and two spliced, were reliably scored correctly, while one intact report was mis-scored by 80% of the judges. Judges had no greater confidence in correct decisions than incorrect ones. Dream report features identified by the judges that were effectively used in scoring dreams included characters, locations and objects, while psychoanalytic content and writing style were least effectively used. In sum, we find no evidence that dream construction consistently results in identifiable thematic coherence. Rather, scene and plot discontinuities in many cases represent such complete breaks as to be unrecognizable. We conclude that the finding of continuity by those reading these reports reflects ineluctable synthetic activity in the mind of the dream researcher or analyst rather than in the mind of the dreamer.

This paper is part of the Festschrift in honor of Dr. Robert Stickgold.

Statement of Significance.

The function and mechanism of construction of dreams remain hotly contested. Understanding the structure of dream reports would provide important insights into both of these outstanding issues. Some dream reports contain apparently complete scene and plot discontinuities. But the classic Freudian theory of dream construction argues that such complete discontinuities cannot exist. Our finding here that neither psychoanalysts nor cognitive neuroscientist could reliably find such continuity in 80% of such dream reports (while also erroneously identifying such continuity reliably in 10% of “spliced” dreams reconstructed from dreams of two different dreamers) indicates that such continuity cannot, in fact, be reliably recognized in dream reports with such discontinuities and brings a central tenet of Freudian dream theory into question.

Introduction

Since Freud first published The Interpretation of Dreams in 1900, most descriptions and theories of dream content have emphasized their thematic coherence [1, 2]. Globus [3] has even argued for the existence of “unifying concepts that bridge across radical thematic discontinuities” (p. 27). In contrast, Seligman and Yellen have suggested that dream construction operates on a “principle of adjacency” [4] (p. 17), whereby dream segments are logically connected to those immediately preceding and following them, but not to more distant segments. Unfortunately, all of these claims were based on anecdotal reports and subjective evaluations.

Claims of coherence in dreams or any other reports of mental content must be greeted with particularly keen skepticism because the judgments are so highly dependent on the subjective evaluations of the clinician or researcher, a problem that extends across the fields of psychology and psychiatry. It is therefore still unclear when thematic coherence is actually present in a dream and when it is merely read into the dream report post hoc by the researcher or analyst or dreamer, themself. Indeed, each of these three groups could reasonably be expected to be biased toward concluding that such coherence is present. To tackle this problem, we have developed a technique of dream analysis, called “dream splicing,” which permits the objective evaluation of thematic coherence [5]. “Spliced dreams” are created by combining segments from different dream reports. A set of spliced dreams is then mixed with a set of intact dreams, and judges are asked to determine which of the reports are spliced and which are intact.

In our initial study, we reported that in most cases judges could accurately distinguish intact from spliced dreams, with correct scores outnumbering incorrect ones by almost 5 to 1 [5]. That study, however, only looked at reports without major scene and plot shifts. In the current study, we hypothesized that judges would be unable to distinguish intact from spliced dreams constructed from dream narratives containing complete scene and plot shifts. Indeed, we demonstrate that very little thematic coherence can be found between adjacent segments that span major scene and plot discontinuities.

Methods

Selection of reports

Dream journals containing 10 or more dream reports at least 10 lines in length were collected from 36 psychology students who were required to keep dream journals as part of a class and scored for scene/plot discontinuities, characterized by abrupt changes in location, thoughts, and actions. A subset of these were “complete discontinuities,” wherein no character (except the dreamer) or object is reported at both locations and no reference to one location, its characters or objects is made at the other location. The identical physical object or character must be referred to at both locations to violate this constraint. Complete discontinuities were always noted in the report by a sentence starting with, “Then I was in [a new location],” or some similar wording. Dream journals were read by RS until 20 reports containing complete scene/plot discontinuities (one from each of 20 subjects) were identified. For the purpose of this study, a single dream was operationalized as a written report numbered and dated by the participant. Multiple dreams from a single night were numbered sequentially by the participant.

Preparation of spliced-dream reports

Spliced-dream reports were prepared as previously described [5]. In brief, reports were first edited for spelling and gross grammatical errors, and references to the start or middle of the dream (e.g. the phrase “At the start of the dream...”) were then removed. Ten of the 20 reports, selected at random, were each divided into two segments at the point of discontinuity, and segments from different reports then randomly recombined to include the first segment from one report followed by the second segment of another. The 20 final reports—10 intact and 10 spliced—were put in a random order and printed out, one per page. For each report, the two segments were separated by a row of asterisk. Two examples, one spliced and one intact, are given here. The reader may wish to try to determine which is which:

Report #1

I am in a big city. The city is really the City in the Clouds. I know we are going to be invaded by the Imperial Fleet. My son G. and I are going to see the ruler of the City to warn her. We walk to the palace. On a narrow covered bridge a guard stops us. She says: “The queen cannot see you. She is not seeing anyone.” We know suddenly that the queen is our friend Z. We say to the guard: “The queen is an old friend of ours.” The guard lets us in. We go to see Z and she is pregnant.

* * * * * *

I am in a sunny sea resort, but I stay inside and never go out—I have a conversation with somebody on that theme. I am lying in bed with my ex-husband. He says: “I have to leave, will you still be here when I come back?” “Probably not,” I say. “Maybe I will be here, maybe somewhere else.” He leaves and I take off my shoes. They are made out of rubber. I have trouble doing that and my skin starts peeling off. All of a sudden I see I have a big raw wound on my foot, bleeding. My head starts to itch. I comb my hair and lice fall out, three different kinds, big white ones and small brown and red ones. I am furious. I somehow know I got those lice from the seat I was sitting on in a boat which brought me to that place.

Report #14

I believe I am in my hometown. M (from work) comes by to visit. He is worried about the weather. My brother is home. We start to look for the forecast on the radio. The receiver is very complicated, but I recognize it and am familiar with how it works. We need to find a station near Baltimore and S asks M for a Baltimore zip code. M answers and S asks if it has changed often. M says no; S tells us his has changed five times in as many years. So we add the zip code to some other number that seems to be a standard number or at least one you can look up if you need to and we get a station near M’s home. I start to tune in the station, push an incorrect button (a & b are for fast searching?). S is irritated and tells me I’m doing it wrong. I tell him I simply made a mistake; I’m mad at him. M’s station is playing music, so we try others. Almost all the stations are playing classical music. So we go back to M’s station and wait. We are sitting in a room similar to my own living room, but it seems to be a mirror image. M and I talk, finally get to the weather, and then M decides he will try to get to Baltimore.

* * * * * *

There is a party. All the men are wearing suits, very odd. I am following someone home. (To make sure he is safe?) I think he is gay. I come back to the party, driven by a policeman who is black and a friend of mine. I can’t pay for the ride? There is some kind of parade or celebration going on.

Sixteen of the 20 judges scored Report #1 as spliced. Despite their general agreement, the report in fact was not spliced; it contains the two halves of a single dream report. The second report (Report #14) was correctly scored as spliced by 17 of the 20 judges.

Judges

One group of 10 judges (referred to below as “analysts”) was solicited from the membership of two psychoanalytic societies and received the dream reports by email. A second group of 10 (referred to below as “cognitivists”) were members of a seminar group involved in the formal study of dreams and cognition and received the reports by hand. Judges were given the packet of 20 reports along with a score sheet and were informed of the nature of the reports, including that half were spliced. They were then instructed to score each report as either intact or spliced and to indicate how confident they were of each score, on a scale of 1–5. After the scoring was completed, judges were asked to immediately review their scoring and to indicate “any special tricks,” such as a gender change in the dreamer from one segment to the next or similar locales in the two, that they used for scoring individual reports.

While all judges scored the reports, only 15 of the judges (7 cognitivists and 8 analysts) completed the second portion of the protocol, giving their reasons for judging reports as intact or spliced.

Statistical tests

Statistical analyses were performed using the Student t-test, binomial distribution analyses, Pearson regression analysis, and chi-square test as indicated. Binomial distribution analyses were performed using Microsoft Excel; all other analyses were performed using SPSS v23 for the Macintosh.

Informed consent

At the time of data collection (1996), the Massachusetts Mental Health Center, where the study was based, did not require informed consent in research studies (cf. Stickgold R, Pace-Schott E, and Hobson JA. A new paradigm for dream research: mentation reports following spontaneous arousal from REM and NREM sleep recorded in a home setting. Consciousness Cogn. 1994;3:16–29.)

Results

Accuracy in scoring reports

Overall scoring

Much to our surprise, judges correctly identified intact and spliced dreams at rates only 7% better than the 50% expected by chance (t = 2.4, df = 19, p = .03, dʹ = 0.52, 95% CI = 51 to 53), identifying intact and spliced reports with equal accuracy (intact = 56%, spliced = 58%; t = 0.51, df = 19, p = .62, dʹ = 0.09, 95% CI = 41% to 70%).

Accuracy of individual judges

While the accuracy of individual judges varied from 35% to 75%, the distribution of accuracies did not differ from that expected by chance (χ² = 27.1, df = 1, p = .10, w = 0.82). Thus there were no significant differences among the 20 judges.

Accuracy scoring individual reports

Individual reports were scored with accuracies ranging from 20% to 95%. In contrast to the scoring by individual judges, the distribution of scores for individual dreams was highly significant (χ² = 56.1, df = 19, p < .001, w = 1.67), with the actual scores skewed towards better than chance accuracy (Figure 1). Four of the 20 dreams, 2 intact and 2 spliced, were scored at significantly better (n = 3) or worse (n = 1) than chance (Figure 1). With these removed, the distribution of scores on the remaining 16 reports did not differ significantly from chance (χ² = 20.7, df = 19, p = .15, w = 1.02).

Figure 1.

Distribution of dream scores. Bar graph: dream report scores; red line: random binomial distribution; black lines: 95% confidence interval. Green bars: intact dreams scored significantly better or worse than chance (Rpt-I1: 4 correct, Rpt-I4: 19 correct); blue bar: spliced reports scored significantly better than chance.

While most reports, both intact and spliced, were scored at near-chance levels (one-half scored with accuracies of 40%–60%, three-quarters scored with 30%–70% accuracy), some were scored with accuracies unexpected for a random binomial distribution, with the content of four reports leading 80%–95% of the judges to agree in their scores. But these near-unanimous agreements included both correct and incorrect judgments, demonstrating that reports can consistently lead judges to an incorrect evaluation of the report’s intactness as well as a correct evaluation. An example of such a report, incorrectly scored as spliced by 16 of the 20 judges, was given above in Methods (Report #1). By way of contrast, an intact report correctly scored by 19 of the 20 judges is given below:

Report #4

I am an observer, as I watch two children, one 4 years-old and the other an adolescent, walk out onto the roof of their house to look at something on the ground. We are all on the second floor of the house, including their parents who do not appear concerned. I then become a participant when I move toward them to guide them back into the house. I am feeling anxious that they will fall off the roof.

* * * * * *

I am in a gathering room, with the family dog present. The dog is black and of medium build. I approach the dog who is sitting on the floor by giving it a big hug, saying, “what a good dog.” Suddenly, it attempts to bite my face, though it doesn’t feel like the dog did so. I spend some time trying to let the dog know I am not fearful, while inside I am surprised that a dog would bite me.

Interestingly, while agreeing that this report was intact, analysts and cognitivists gave different reasons for their judgments. Four of six cognitivists cited a theme of family or the location in the home as their reason for judging the report intact, while none of the eight analysts cited either of these reasons. Instead, they relied on a sense of anxiety or defensiveness (five judges), a reason cited by two of the eight cognitivists as well. This issue is discussed in greater detail below.

Confidence in scoring reports

Overall confidence ratings

Judges used the full 5-point scale, with one-third of the scores being ratings of 1 (minimal confidence) or 5 (maximal confidence). Overall, judges showed moderate confidence in their confidence ratings (2.98 ± 1.12 [SD]). Surprisingly, judges showed no greater confidence in their correct judgments than in incorrect ones (average confidence: correct = 3.01, incorrect = 2.98; t = 0.40, df = 18, p = .69, dʹ = 0.03, 95% CI = 2.59 to 3.40).

Confidence ratings of individual judges

Judges showed considerable variation in their overall confidence levels, with mean values for individual judges ranging from 1.4 to 4.0 and averaging 2.99 ± 0.80 (SD) Ratings were significantly different across judges (repeated measures ANOVA, main effect of judge: F = 264, df = 1, p < .0001.

Confidence ratings of individual reports

Averaged across all judges, certain reports were scored with greater confidence than others (repeated measures ANOVA, df = 17.9, F = 3.14, p < .0001). However, there was no correlation between the confidence ratings and scoring accuracies for individual reports (Figure 2; Pearson’s correlation test, df = 18, r = .15, p = .52), and those reports with highest and lowest confidence ratings were not scored correctly at rates significantly different from chance.

Figure 2.

Correlation of dream report confidence levels with scoring accuracy. Green circles = intact reports; red circles = spliced reports.

Justifications for report scores

For each report scored, judges were asked to indicate the basis for their judgment. Free-form answers were solicited and subsequently categorized. No categories were suggested to the judges. More than one category could be scored for a given report score. Pre hoc categories included (1) the gender of the dreamer and similarities or differences between the two report segments in (2) characters, (3) objects, (4) locale, (5) affect, (6) theme, (7) psychoanalytic (Freudian) content, or (8) writing style. Psychoanalytic content was identified post hoc as use of psychoanalytic terminology (e.g. “homosexual ideation,” “defense against anxiety,” “narcissism,” “Oedipal,” “denial”) in the judges’ answers.

Fifty-seven percent of the justifications given were in conjunction with correct judgments, with no significant differences between categories in their correlation with correct judgments (χ² = 10.0, df = 7, p = .19, w = 0.19). Justification based on characters and locale were most likely to be associated with correct scores (binomial distributions—character: 15 of 20, 15 of 19, χ² = 4.0, df = 1, p = .78, w = 0.12). By contrast, reports based on writing style or psychoanalytic content were least likely to be associated with correct scores (style: 40 of 78, p = .99, w = 0.07; psychoanalytic: 15 of 32, p = .98, w = 0.07).

Comparison of analysts and cognitivists

Although there were subtle clues to the presence of dream coherence in these reports, the ability of judges to identify the coherence might depend on their training and perspective. For this reason, we used 10 judges expected to score reports from a cognitive neuroscience perspective and 10 judges expected to use a more psychoanalytic approach.

If the two groups used different strategies in scoring the reports, the outcomes were indistinguishable. Each group scored reports correctly 57% of the time, and their accuracies were highly correlated on a dream-by-dream basis (df = 18, r = .66, p = .001). Each group showed the lowest accuracy for report I-1 and the highest for report I-4 (see Figure 1).

In contrast to these similarities, the two groups reported using different strategies in scoring the reports (Figure 3, χ² = 35.7, df = 7, p < 0.0001, w = 0.36). Cognitivists based their judgments on characters, objects, and locales 2.2 times more frequently than did analysts (27 vs. 12; χ² = 24.3, p < .0001, w = 0.29), but never used psychoanalytic content (0 vs. 32; p < .0001, w = 0.28).

Figure 3.

Justifications for scores. Number of times each justification was used, by group. G: gender; C: characters; O: objects; L: location; A: affect; T: theme; S: style; P: psychoanalytic.

Discussion

The central issue addressed in this study is the mechanism of dream construction and specifically the extent to which it is necessarily produces coherent dream narratives with identifiable continuity that extends through the entire dream. While psychoanalysts have emphasized dream continuity, cognitive neuroscientists have emphasized their discontinuity. Of course both could be partially true. For example, we have reported that the bizarre transformations of objects and characters seen in dreams and which we had described as the products of chaotic brainstem activity [6], in fact follow strict transformation rules, and that judges can readily match initial objects with their transformed products [7]. Thus, an objectively identifiable coherence can be found linking these apparently chaotically introduced dream elements. The current study was prompted by the question of whether another potentially chaotic form of dream transformation, namely the complete plot/scene discontinuities found in about 10% of the dream reports in our sample, also follows transformation rules that permit the objective identification of coherence between the original and final scenes and plots. Our dream splicing technique allows such coherence to be objectively measured [5]. It should be acknowledged that it is unclear what causes a dream to suddenly undergo such a discontinuity and whether such discontinuities serve a function.

Definition of a single dream

There is much discussion in the dream research field over what constitutes a single dream. But just as there is no agreement on what constitutes a dream—whether daydreams, hypnagogic imagery, non-REM dreams more generally, or reports of thoughts during sleep without sensory imagery should be considered “true” dreams—there may be no consensus on what constitute a “single” dream. Indeed, for dreams more generally, a consensus agreement was reported in 2001, stating that, “a single definition for dreaming is most likely impossible given the wide spectrum of fields engaged in the study of dreaming, and the diversity in currently applied definitions” [8]. As a result, many researchers simply refer to “sleep mentation,” avoiding the issue of defining what is or isn’t a dream.

While some researchers might argue that a dream report with a complete discontinuity by definition represents two separate dreams (thus making this study meaningless), there seem to be no pre hoc reason to come to this conclusion. Rather the argument seems to be based on a particular definition of what a dream is and a particular theory about what they can or cannot include. It is for this reason that we have operationalized our definition of a dream simply as what the dreamer reports as a single dream.

Frequency of occurrence of dreams with compete plot/scene discontinuities

While reporting, 10% of the dream reports in our sample were scored as having complete plot and scene discontinuities. We are not suggesting that this represents their frequency in the general population of dream reports. In the only known study looking specifically at the frequency of various forms of dream bizarreness, 1.9% of individual sentences in dream reports were scored as complete discontinuities [9]. Reports were at least five lines long and contained 30–600 words. The average number of sentences per report was not given. But if they averaged five sentences, then the findings would translate into a discontinuity rate of 9.5%, essentially the same as ours.

Identification of dream coherence

Contrary to our own expectations, we found surprisingly little support for the idea that dreams with scene and plot discontinuities contain identifiable coherence; judges correctly scored reports as spliced or intact only 57% of the time, a rate just 7% above chance. In addition, there was no evidence that individual judges scored better or worse than others or were more confident of reports they scored correctly. In short, there was no evidence that any of the judges could reliably distinguish intact reports from spliced reports. These findings are in agreement with a pilot study in which 14 spliced and intact dreams were scored correctly in only 37 of 70 attempts, or 53% of the time (unpublished results).

Although judges correctly scored reports at rates only slightly better than chance, this difference was significant (p < .03, dʹ = 0.54, 95% CI = 51 to 63). (To achieve 80% power in this analysis (α = .05) would require an n = 23 judges.) Further analysis indicated that judges scored 3 of the 20 reports at significantly better than chance rates (80%–95% correct). Still, the remaining 85% of the reports were not scored at rates significantly better than chance, and one was scored significantly worse (20% correct). The rate at which reports were consistently scored correctly was not significantly higher than the rate at which they were consistently scored incorrectly, although the numbers are too small to allow significance to be reached. Thus, it is unclear whether or not the presence of report content permitting consistent scoring across judges is a product of a coherent dream construction process extending across scene and plot discontinuities. But if present, such coherence is detectable in no more than 10%–15% of such reports.

The difficulty of judges in accurately scoring reports is not due to the dream splicing technique. When the same procedure was carried out with reports that lacked scene and plot shifts [5], judges scored reports correctly 82% of the time, and 13 of 22 reports were correctly scored by all five judges. Thus, the inability to accurately score the current dream reports results directly from the presence of scene and plot discontinuities in the reports.

The representative nature of dream reports with discontinuities selected for study

To be clear, we are not suggesting that we are analyzing a representative collection of dream reports more generally, and we are specifically saying that the reports analyzed here are not even representative of our collection of reports. Indeed, they are not representative, as they contain complete scene and plot discontinuities.

Similarly, we are not trying to determine the frequency or characteristics of these discontinuous dream reports, beyond that they are discontinuous. If one is testing the hypothesis that all dreams have identifiable coherence across their length, then in theory only one dream report without identifiable coherence would be needed to disprove the hypothesis. Since 16 of our 20 reports were correctly scored by less than 15 of the 20 judges, and 8 by less than half, it is clear that the hypothesis that all discontinuous dream reports have enough objective coherence to be correctly identified as intact by most, in not all, judges is false.

Strategies for identifying dream coherence

The nature of the continuity identified by judges is difficult to summarize. Overall, cognitivists were most likely to be correct when they based their judgments on characters and locations. This is in agreement with the results of an earlier study with abridged dream reports, in which both intact and spliced reports with the highest accuracies of scoring were found to have distinctive similarities (for intact reports) or differences (for spliced reports) in characters, objects, or locations [5]. But in the current study, judges with psychoanalytic training gave consistently different reasons for their judgments and yet scored as well as cognitivists. In one case, 19 of the 20 judges correctly identified a report as spliced, but cognitivists and analysts gave very different justifications for their scores; of six categories of justifications indicated by one or more judge, only two were used by both groups.

One possible explanation of this finding is that the two groups indeed based their decisions on two different types of information, both of which happened to be present in this report. But if this were true, one would expect a general lack of correlation in the accuracies between the two groups across the 20 reports, some of which would provide clear evidence of a type preferred by one group, while others would contain evidence of a type preferred by the other. Yet the correlation between scores was quite high (r = .66, p < .001), accounting for 44% of the variance in scores between groups.

Thus, it seems more likely that judges actually used the same information as the bases for their judgments, but perceived it differently, presumably based on their differing views of where coherence would be expected to be found. This suggests that most descriptions of coherence within these reports are post hoc constructs of the judges, rather than reflecting inherent coherences in the reports themselves. This is true even in the reports for which judges were reliably accurate in their scoring. For these reports, judges’ strategies still differed, and confidence ratings were not unusually high. Thus, judges’ confidence levels reflected something other than their objective ability to correctly identify spliced and intact reports.

Confidence in dream scores

More generally, there was no significant correlation between the accuracy of individual judges and their overall confidence in their judgments. Nor was there any correlation between the accuracy with which individual reports were scored and the confidence of judges in their scorings of those reports. In short, there is no evidence to support the idea that judges had any means of assessing the likelihood that their judgments were correct.

Consideration of covariates

We have shown that the theoretical orientation and the confidence of judges do not significantly affect their accuracies. But we do not seek to probe further into such questions as whether males scored better or worse than females or whether the length of a report predicted its correct scoring, as these are not germane to the central question being investigated. Indeed, we believe that adding such details would distract from the main point of the paper.

Constraints on dream construction

The question of dream coherence is a complex one, conflating dream construction, dream meaning, and dream interpretation, and a detailed discussion of this question is deferred to a future paper. The definition of dream construction, however, is relatively straightforward. It consists of those mechanisms and processes which lead to the construction of dreams including their sensorimotor hallucinations, emotions, thoughts, and narrative development. Studying dream reports can provide clues to these underlying mechanisms. Thus, the previously noted constraints on object transformations [7] suggest that these bizarre transformations are constrained by preexisting associative networks. We therefore postulate, as did Hartley [10], that associative processes form a critical element in the dream construction process. Our findings in this and our earlier study [5] indicate that specific rules also constrain narrative development during dream construction:

• The development of dream narratives is normally constrained by the principle of adjacency [4], whereby each segment is objectively related to the preceding one [5].

• This constraint does not extend to distantly connected segments; between the initial and final segments of a dream the content shifts so much that it is generally impossible to identify any objective continuity between these two segments [5].

• The principle of adjacency does not normally hold for adjacent segments which span a discontinuity in scene and plot; in these cases, there again is no identifiable relationship between the neighboring segments that can be consistently identified by judges.

• This lack of coherence across complete scene and plot discontinuity does not seem to have any significant impact on other models of dream construction, such as the continuity hypothesis of Hall and Nordby [11] or the NEXTUP model of Zadra and Stickgold [12].

The objection can be raised that thematic coherence in fact does exist across discontinuities and even across the length of a dream, but that our methodology is simply inadequate for the task of identifying it. Such an objection is, of course, legitimate, but must be seen in proper perspective. Whenever scientific studies fail to find an effect, it is possible that the effect was simply too small to be measured by the methodology employed. But our hypothesis—that there is no coherence across scene and plot discontinuities—was falsifiable and could have been disproven by this experiment. Indeed, similar hypotheses, that adjacent segments of dreams without discontinuities are not coherently connected and that object transformations are not constrained by coherence have both been disproven by this and similar analyses [5, 7].

Indeed, the current study found strong evidence for about 10% of the dreams within our sample that had scene and plot discontinuities maintaining enough coherence across the discontinuity to be reliably identified as intact reports. What this percentage would look like in a larger sample of dreams is, of course, unknown. This coherence most likely lies in similarities in persons, places, and things found in the two dream segments. But for the remaining 90% of such dream reports, no objective evidence of coherence can be found. It remains for those who object to our conclusions to carry out experiments which objectively show that such coherence does exist. In this regard, we anticipate the claim that it is only by knowing the dreamer that an independent judge could be expected to recognize thematic continuity in apparently discontinuous dream reports. Such a theory could be tested by collecting and splicing dream reports from patients well known to their analysts and then having the analysts attempt to distinguish the intact from spliced reports in a blind study. We encourage our analytic colleagues to conduct such studies and would be happy to collaborate with any who wished to do so.

We wish to emphasize that objective evidence cannot be obtained by merely having judges read a complete dream report and stating whether or not discontinuous segments are coherently connected. Whether considering our specific finding that judges “identified” such coherence in 42% of spliced reports (which clearly did not contain such continuity) or, more generally, the suppositions behind such instruments as the Rorschach and thematic apperception tests, it is clear that the human mind can and regularly does attribute meaning and coherence to events independent of their actual existence. In this sense, even the judgment of the dreamer, perhaps based on free association to the dream content, is not likely to be useful. While of potential clinical interest, using such additional information simply substitutes the subjects’ post hoc creative processes for that of the researcher, and objectivity is again lost. But this more complex issue is at the core of the questions of dream interpretation and meaning, topics which are addressed more directly in the accompanying paper [9].

We conclude that there is no coherence, thematic, or otherwise, across the vast majority of scene and plot discontinuities found in dream reports. While there is a pervasive sense within the dream research community that such continuity does exist, we ascribe this misperception to the ineluctable synthetic activity of mind the researchers rather than that of the dreamers. Testing this hypothesis further by asking dreamers to identify links between apparently discontinuous segments in their own dreams would be of value.

Funding

Funded by National Institute of Mental Health grants MH048832 and MH65292.

Conflict of interest statement

The authors have no financial arrangements or connections to disclose. The authors have no potential conflicts of interest to disclose.

Data Availability

The data for this article will be available to share on reasonable request to the corresponding author.

Author contributions

Robert Stickgold (Conceptualization [lead], Data curation [lead], Formal analysis [lead], Funding acquisition [lead], Investigation [supporting], Methodology [lead], Project administration [equal], Resources [lead], Supervision [lead], Validation [equal], Visualization [lead], Writing—original draft [lead], Writing—review & editing [equal]) and Funke Sangodeyi (Conceptualization [supporting], Formal analysis [supporting], Investigation [lead], Visualization [supporting], Writing—original draft [supporting], Writing—review & editing [equal])