A New Beginning for Empirical Dream Research

Erin J. Wamsley & John Antrobus

City College Department of Psychology
New York, NY

NOTE: If you use this paper in research, please use the following citation, as this on-line version is simply a reprint of the original article:
Wamsley, E. J., & Antrobus, J. (2005). A new beginning for empirical dream research [Review of the book The Scientific Study of Dreams]. American Journal of Psychology, 119(1), 129-135.

Review of:
The Scientific Study of Dreams: Neural Networks, Cognitive Development, and Content Analysis
By G. William Domhoff. Washington, DC: American Psychological Association, 2003. 209 pp. Cloth, $49.95.

William Domhoff's The Scientific Study of Dreams breaks new ground in the field, not by proposing grandiose, premature answers to questions about the nature of dreams but rather by showing us with unprecedented clarity and scope where we have erred in the past and allowing us to start over again. In a careful, reasoned critique, Domhoff demonstrates the shortcomings of the dominant dream theories that haunted the last century. Based on a thorough examination of converging evidence from neurophysiological, cognitive, and content analysis approaches to the study of dreams, he then lays the groundwork for the construction of future theories of the dreaming process from a neurocognitive perspective. Although a new, complete theory is not presented here, its insightful and unbiased assessment of where dream research stands at this time should make The Scientific Study of Dreams required reading for everyone interested in an empirical approach to understanding dreaming.

With the publication of Freud's Interpretation of Dreams in 1900, popular thought about the nature of dreams became centered around the idea that our dreams are disguised wishes originating from unconscious influences in early childhood. Dreams allowed us to hallucinatorily gratify socially unacceptable wishes repressed in our unconscious, and the transformative process of dreamwork ensured that these themes were well enough disguised in our dream content to allow us to sleep through the night, without being startled awake. Until the early 1950s, this psychoanalytic conception dominated dream psychology and was explored largely using the same techniques of clinical observation that led Freud to the formulation of his original theory.

The modern era of empirical research on dreaming was kicked off in the 1950s with Aserinsky and Kleitman's (1953) groundbreaking discovery of rapid eye movement (REM) sleep As a graduate student working on a summer project, Eugene Aserinsky serendipitously discovered that about every 90 minutes throughout the night, humans go through short periods during which their eyes dart back and forth underneath their eyelids, accompanied by dramatic increases in brain activity as measured by electroencephalography. Soon it was reported that subjects were much more likely to report a dream when awakened from REM sleep than when awakened from non-rapid eye movement (NREM) sleep (Dement & Kleitman, 1957). This discovery kindled hope in many researchers that the physiological cause of dreaming had been found. It seemed that all we needed to do was to study the physiology of REM sleep and its connection to dream reports, and we could unravel the mystery of what dreams were and how they came to be. Unfortunately, the large numbers of psychophysiological studies that ensued, aiming at understanding the relationship between dreaming and features of REM sleep, failed to lead to a successful explanation of dreaming. This line of research probably was encumbered by the fact that the widespread idea that dreaming is confined exclusively to REM sleep turned out to be false (Foulkes, 1967).

In the 1970s, the primary exponent of the idea that REM sleep was the key to understanding dreaming was Allan Hobson and his activation synthesis theory. One of Hobson's motivations in the creation of this theory was to debunk the still-prominent Freudian idea that dreams contained a hidden meaning masked by manifest content. Beginning from a neurophysiological model of REM sleep, activation synthesis theory holds that essentially random stimulation of the forebrain originating in pontine brain stem areas involved in the control of REM sleep is the impetus for dream creation. The higher brain then takes this random activation, seeking to make sense of it, and synthesizes it into the semicoherent structure that we experience as the dream (Hobson & McCarley, 1977). Thus, there is no latent content being masked; the dream content is simply a loose synthesis of a number of randomly activated discrete elements, the meaning of which are apparent on a surface level.

Freud's and Hobson's theories remain the most prominent frameworks for understanding dreaming to this day. One of the major strengths of Domhoff's book is his unbiased assessment of psychoanalytic, activation synthesis, and other traditional theoretical approaches to understanding dreams. Domhoff exposes the weak points of these theories, showing how each is unable to encompass the full scope of available data on the nature of dreams.

In a thoughtful analysis of the main tenets of Freud's theory of dreams, Domhoff shows how each of Freud's major claims about the nature of dreams has failed to stand up to empirical investigation. For example, in regard to the idea that dreams are disguised wish fulfillments, Domhoff points out that although Freud's ideas on this point were based largely on the dreams of children, David Foulkes's numerous empirical studies on children's dreams have revealed that, contrary to Freud's claims, the hallucinatory gratification of simple desires is very uncommon in the generally impoverished dreams of children. In addition, the well-documented repetitive nightmares of particular life incidents common in victims of posttraumatic stress disorder pose a serious objection to the idea that we dream in order to hallucinatorily gratify wishes. Freud's concepts of repression, dreamwork, and the influence of very early childhood experience are similarly challenged on the basis of empirical investigations. Domhoff nonetheless credits Freud for popularizing the idea that dreams reflect important psychological qualities of individuals, an idea that has found consistent support from content analysis studies of dreaming by Domhoff and others.

Although praising Allan Hobson's activation synthesis hypothesis for working toward the integration of detailed neurological data with dream features, Domhoff's assessment of activation synthesis is that it, like Freudian dream theory, is unable to explain all of the available data and to account for some of the most salient characteristics of dreaming. Ultimately, the problems with activation synthesis rest on the theory's leap from a well-supported model of REM sleep generation to specific dream features, with little attention to what neurocognitive processes might justify such a leap. For example, Hobson and his colleagues attribute bizarreness in dreams such as sudden changes in scene and improbable combinations of elements to the forebrain's inability to make complete sense of random stimulation originating in the brain stem (Mamelak & Hobson, 1989). Domhoff criticizes such explanations for placing too much emphasis on the activity of cells in a small area of the brain stem as an explanation for features of dreams while failing to explore what cognitive mechanisms might underlie such a link between low-level neuronal activity and dream content.

Domhoff argues additionally that activation synthesis theory has not acknowledged some research that is incompatible with its claims. For example, reports of vivid and complex dreaming from NREM sleep (Foulkes, 1967) and David Foulkes's finding that complex dreaming depends on the development of general cognitive skills seem to contradict activation synthesis's emphasis on REM-related activity in the pontine brain stem as the major determinant of dream features. Recent work by Solms (2000) also raised questions for activation synthesis by demonstrating that patients with damage to the pontine brain stem that eliminates REM sleep may still report vivid dreaming. Domhoff (p. 157) concludes that the connections between REM sleep brain stem activity and dream features postulated by activation synthesis are "premature at best," failing to consider the cognitive mechanisms that might mediate such connections and being incompatible with evidence suggesting that dreaming occurs in the absence of REM sleep.

Domhoff argues that we need to start over from the beginning. We need to throw out the old theories and set out again from the few solid empirical findings on the nature of dreams accumulated thus far. It is in its introductory chapter that The Scientific Study of Dreams does its best work. Here, based on the data from Foulkes's developmental studies, content analysis studies, and recent brain imaging and lesion studies of dreaming, Domhoff lays out for the reader an extremely useful, detailed discussion of the facts from which we should proceed.

In the last 10 years, the study of dreaming has been revitalized by new data from brain imaging studies elucidating how patterns of neural activation differ across sleep stages and by clinical observations from Solms (2000) describing types of brain lesions that lead to the cessation of dreaming or allow dreaming to be preserved. These studies have been widely cited in the dream research literature in the past decade, but very few writers have reviewed this work with the depth and unbiased clarity that Domhoff does here. Beginning in the late 1990s, positron emission tomography studies by Braun, Balkin, and Wesensten (1997), Maquet, Peters, and Aerts (1996), and others began to give us insight not previously available into the precise patterns of activation occurring in the human brain during REM and NREM sleep. Here, Domhoff not only points out the widely cited agreements between these neuroimaging studies (e.g., the deactivation of executive systems in all of sleep and the high levels of brain stem and limbic activity found in REM sleep) but also stresses disagreements in the interpretation of these data. Domhoff also rightly cautions us about making too many inferences about dreaming based on these imaging studies. Most of these studies did not collect dream reports from participants at the time of the brain imaging and therefore provide us with important information about regional brain activation in REM sleep but do not necessarily inform us about the aspects of this activity that might be important for dreaming.

Domhoff therefore places more emphasis on Solms's work with interviewing brain-damaged patients about their dreaming habits, rather than on the imaging studies, as the key neurophysiological evidence addressing the brain basis for dream production. Domhoff presents Solms's work in detail, describing what he considers to be the seven key findings related to the neural substrate for dreaming, including the finding that the pontine brain stem, prefrontal cortex, and early visual areas do not appear to be necessary for dreaming, whereas some higher-level visual areas, areas involved in spatial representation, and a motivation-related area targeted by lobotomies in the mid-1900s seem to be crucial to the ability to experience dreaming. Domhoff also makes some fair critiques of this work, however, pointing out that although Solms argues that his data demonstrate that the brain stem area crucial to activation synthesis is not a part of the neural network for dreaming, this finding may not be entirely convincing because it is unclear whether the damage that most patients sustained to this area was sufficient to noticeably interfere with REM sleep. It should also be noted that the dream data collected by Solms were not obtained from laboratory awakenings but relied instead on self-reported summaries of subjects' dream characteristics, leaving open the possibility that these data may have been influenced by processes occurring during the time between the actual dream experiences and the data collection. Nonetheless, Solms (2000) has provided us with some valuable clues as to the possible structure of a neural network for dreaming.

As a part of his overview of what we know about dreams, Domhoff also relies on developmentally oriented studies of dreaming conducted by David Foulkes. Foulkes's cross-sectional and longitudinal studies on the dreaming patterns of children have demonstrated that young children report dreaming much less frequently than adults and that the dreams children do report are shorter and less complex than adult dreams, often consisting of mere static images, rarely including the dreamer as a participant in ongoing action, and not exhibiting the kind of narrative qualities common in adult dream reports. Children's dreams gradually become more complex as they get older but do not resemble adult dreams until age 13-15. These studies suggest an important continuity between dreaming and waking cognition: Dreaming can occur in an adult-like form only when waking-life cognitive systems are fully matured. According to Domhoff, full-fledged dreaming may depend on the activation of a neural network that gradually matures during childhood.

Domhoff also brings our attention to some evidence from cognitive studies by Foulkes and others suggesting that the activation of this neural network for dreaming can occur in all stages of sleep and perhaps outside the sleep state. Several studies have shown that dreamlike mentation can occur during periods of quiet wakefulness, at sleep onset, and during NREM sleep (Foulkes, 1967). Work in our own laboratory has recently demonstrated that dreams become longer and more dreamlike in the late morning in NREM as well as REM sleep. In fact, late in the sleep phase, NREM dreams -- which have traditionally been described as less dreamlike than REM dreams -- are comparable on these dimensions to REM dreams early in the night (Wamsley et al., 2004; Antrobus, Kondo, & Reinsel, 1995). Such findings demonstrate that, rather than being the result of physiology unique to REM sleep, dreamlike mentation occurs whenever sensory input from the outside world is attenuated and sufficient levels of activation are present in a neural system underlying dreaming.

Contributions to our understanding of dreaming made by studies using a content analysis approach are also included as a key portion of Domhoff's review. In addition to providing support for Foulkes's developmental theory of dreaming, content analysis studies have provided the most powerful evidence for what has sometimes been called the continuity hypothesis of dreaming. The continuity hypothesis holds that dreams are neither hidden messages from the unconscious nor mere random brain activation but are instead transparent reflections of experiences, thoughts, and concerns from waking life. Support for the continuity hypothesis from content analysis includes Calvin Hall's (1947) blind analyses of dream journals, in which he used the content analysis method to construct accurate profiles of individuals' waking lives, with dream journals as his only source material. Studies of gender and cultural differences in dream content also seem to support the continuity hypothesis, in that people with different waking experiences exhibit different patterns of dream content. In general, such studies have consistently demonstrated that people dream of the things that they experience and think about during their waking lives. These data support a conception of dreams as being similar to our waking thought processes, focusing on the ideas and experiences that are important to our waking lives, as opposed to being either symbolic or random in nature.

Domhoff ends this chapter with an attempt to synthesize these neurological, developmental, and content analysis findings into a set of general principles from which a neurocognitive theory of dreaming can be built. Domhoff proposes that dreaming can occur in any stage of sleep, when in the absence of incoming sensory information a neural network supporting dream creation reaches some critical threshold of activation. This neural network is likely to include the motivation-related areas identified as crucial to dreaming by Solms (2000) and is not likely to include the pontine brain stem or the prefrontal cortices. Adult-like dreaming depends on the maturation of this network, which, following Foulkes's work, does not appear to be achieved until at least early adolescence. According to Domhoff, the dream content itself mirrors waking thought and experience through its reliance on the dreamer's conceptual system, creating "reasonable simulations of the real world" (p. 32) from our memories, knowledge, and schemata. Of course, some portions of even this general framework that Domhoff describes are still speculative and untested. In particular, little empirical work has been done to address the cognitive and neural mechanisms by which dreams come to reflect waking thoughts and experiences.

In between Domhoff's synthesis of what we know about dreaming and his critique of traditional theories, an insightful discussion of methodological issues in the study of dreaming leads up to three chapters devoted to the Hall-Van de Castle method of dream content analysis.

Domhoff begins this section by discussing the merits and drawbacks of various methods of dream collection including collecting dreams in a laboratory setting, from home journals, from classroom settings, and from therapeutic sessions. As a former student of early cognitive dream theorist Calvin Hall, Domhoff has long advocated the use of the Hall-Van de Castle method of content analysis as a reliable way to understand dreaming, and in his discussion of methods of dream analysis, he makes a strong case for the superiority of the content analysis approach over many other less reliable dream analysis techniques. In his final assessment of the experimental methods most commonly found in dream research literature, Domhoff ends up concluding, "Generally speaking, the literature is anecdotal, contradictory, and dispute ridden, and is based on clinical case studies, poorly collected and overly small samples, unreliable and unvalidated rating scales, and inappropriate statistical methods" (p. 65). Domhoff presents content analysis as a methodologically sound alternative to other, less reliable content rating scales.

The Hall-Van de Castle system of content analysis, designed to be used on large samples of dream reports, essentially counts the instances of dream elements (such as friendly social interactions, aggressions, and emotions) and then classifies these elements into well-defined categories. Using content analysis, a researcher can determine what percentage of emotions in a sample of dreams are positive, what percentage of social interactions in the sample are aggressive, or how many dreams in a sample contain at least one sexual interaction, for example. Domhoff argues for this system by pointing out that, compared with other content-rating systems, Hall-Van de Castle content analysis enjoys the advantages of high interrater reliability, widespread usage, the existence of established norms, and procedures to avoid the pervasive confound of dream report length that has plagued much of the dream content literature. A large section of the middle of The Scientific Study of Dreams is essentially a researcher's guide to the use of content analysis, tackling practical questions related to the scoring and analysis of dream reports using this system.

Although Domhoff makes a strong case that content analysis is superior to other types of content rating scales used in dream research, we must realize that any content rating system will only take us so far. As mentioned earlier, studies using content analysis have contributed importantly to the field by showing us that dream content demonstrates continuity with waking thought and experience. However, it should be noted that many dream researchers argue that questions about the overall structure of dreaming are more important than questions about the specific content elements in dreams. For example, if we want to examine bizarreness, self-reflection, qualitative aspects of visual imagery, or qualities of thought in dreaming, content analysis may not be the most appropriate tool to use. Content analysis has been useful in showing us that we dream about things that are prominent in our waking lives, but a theory of dreaming ultimately must seek to explain the qualitative, formal differences between our dream experiences and our waking thought processes.

Although The Scientific Study of Dreams does not culminate in a complete theory of dreaming, Domhoff accomplishes several very important goals. He challenges much of the past literature on dreaming, effectively demonstrating the shortcomings of both psychoanalytic and activation synthesis dream theory and rightly concluding that much of the other literature on dreaming is riddled with methodological flaws. Domhoff then gives us an insightful analysis of what we should consider known about dreaming at this time, calling on us to take the next steps by encouraging academics to be open to new ideas about dreaming and to help loosen the hold of flawed traditional theories on young researchers. The Scientific Study of Dreams is a much-needed wake-up call, telling us that we need to rethink our strategies and start anew, salvaging the bits of sound knowledge that we have accumulated so far and adopting more effective research and theoretical strategies as we move into the future of dream research. The exact direction in which we should proceed is not entirely clear, however. It may be up to the next generation of researchers to find the path that will eventually lead us to a successful theory of dreaming.


Antrobus, J., Kondo, T., & Reinsel, R. (1995). Dreaming in the late morning: Summation of REM and diurnal cortical activation. Consciousness and Cognition, 4, 275-299.

Aserinsky, E., & Kleitman, N. (1953). Regularly occurring periods of eye motility and concomitant phenomenon during sleep. Science, 188, 273-274.

Braun, A. R., Balkin, T. J., & Wesensten, N. J. (1997). Regional cerebral blood flow throughout the sleep-wake cycle: An H250 PET study. Brain, 120, 1173-1197.

Dement, W., & Kleitman, N. (1957). The relation of eye movements during sleep to dream activity: An objective method for the study of dreaming. Journal of Experimental Psychology, 53, 339-346.

Foulkes, D. (1967). Nonrapid eye movement mentation. Experimental Neurology, 4, 28-38.

Hall, C. S. (1947). Diagnosing personality by the analysis of dreams. Journal of Abnormal and Social Psychology, 42, 68-79.

Hobson, J. A., & McCarley, R. W. (1977). The brain as a dream state generator: An activation synthesis hypothesis of the dream process. American Journal of Psychiatry, 132, 1335-1348.

Mamelak, A., & Hobson, J. (1989). Dream bizarreness as the cognitive correlate of altered neuronal behavior in REM sleep. Journal of Cognitive Neuroscience, 1, 210-222.

Maquet, P., Peters, J., & Aerts, J. (1996). Functional neuroanatomy of human rapid eye movement sleep and dreaming. Nature, 383(6596), 163-166.

Solms, M. (2000). Dreaming and REM sleep are controlled by different brain mechanisms. Behavioral & Brain Sciences, 23, 1083-1121.

Wamsley, E. J., Hirota, Y., Tucker, M. A., Smith, M. R., Doan, T., Chaklader, A., Emery, R. N., & Antrobus, J. (2004). Circadian influences on sleep mentation. Sleep, 27(supplement), A59.

Go back to the Dream Library index.

dreamresearch.net home page dreamresearch.net contact info