Chapter 1:
Toward a Neurocognitive Model of Dreams


The neurocognitive model of dreaming and dreams proposed in this chapter has three basic components. First, there is the neurophysiological substrate that underlies and activates the process of dreaming. Second, there is the conceptual system of schemas and scripts that constitutes the process of dreaming. Third, there is the dream content that results from this cognitive process. This chapter discusses each of these components and suggests some of the specific ways in which they may relate to each other.

The Neural Network for Dreaming

Research in neuropsychology and neuroimaging converged in the late 1990s to suggest the broad outlines of a neural network for dreaming that is very different from what had been imagined in the past. There are some disparities between the lesion and imaging results, and a few disagreements among the imaging studies, all of which utilize PET-scan technology, but the overall findings are strikingly consistent and provide the starting point for more detailed studies. However, it would be extremely premature to overspecify the network at a time when new discoveries are being made each year and so much remains to be learned (Antrobus, 2000a; Morrison & Sanford, 2000).

The neural substrate for dreaming provides the necessary level of activation needed for dreaming, and perhaps constrains the types of thinking that are possible. It seems to be responsible for determining the degree of vividness and intensity experienced in dreaming, and may account for other formal features of dreaming, such as its realistic and self-participatory nature, its general lack of self-reflectiveness, and its occasional incongruities of form. However, the neural substrate cannot account for the narrative nature of dreaming or the substance of dream content, which are products of the conceptual systems discussed later in this chapter.

There are four major areas of agreement abut the contours of the neural network among researchers with varying theoretical perspectives. First, the mechanisms that generate REM "support our most vivid and elaborate dreaming" (Foulkes, 1999, p. 6). Second, there are forebrain controls of the REM generator located in the tegmental region in the middle of the pons. Third, a complex forebrain network is necessary for dreaming. Fourth, this forebrain network plays the major role in terms of shaping dream content (Hobson et al., 2000b; Solms, 2000). Based on these agreements, this book concentrates on the development of a neurocognitive model of dreams that can encompass the whole range of dream content and relate that content to waking conceptions and concerns.

There are also some disagreements among theorists. For example, there are varying opinions concerning the mix of neurochemicals that modulates the brain during REM sleep (Gottesmann, 2000; Hobson et al., 2000b; Perry & Piggott, 2000). There are also differences as to whether the neural network for dreaming always includes the area in the pons that is necessary for REM (Foulkes, 1999; Hobson et al., 2000b; Solms, 2000). These and various other unresolved issues related to the neural substrate for dreaming are discussed in chapter 6 as part of the critique of activation-synthesis theory.

The neuropsychological and neuroimaging results are in some ways interchangeable, but the neuropsychological studies provide the best starting point because they always include a crucial psychological component, the presence or absence of the subjective sense of dreaming. The neuroimaging studies, on the other hand, are a mapping of sleep stages, although one research group did collect dream reports from several REM awakenings as well as one non-REM (NREM) awakening (Maquet, 2000, p. 224). Dreaming is highly correlated with REM, but sleep stages are an imperfect indicator of dreaming because there is at least some degree of dreaming in NREM (Foulkes, 1966; Foulkes, 1985; Hobson, Pace-Schott, & Stickgold, 2000a). To connect neuroimaging work more closely to the process of dreaming, there is a need for studies of NREM periods from which dreams are reported, and of REM periods where no dreams are recalled upon awakening (Maquet, 2000, p. 224). It would be especially useful to have studies of Stage II NREM after the first four REM periods because many dreams seem to occur at this time (Antrobus, Kondo, & Reinsel, 1995; Cicogna, Natale, Occhionero, & Bosinelli, 1998).

The primary source of neuropsychological information on dreaming is a study by Solms (1997) in which 361 consecutive neurological patients were asked in great detail between 1985 and 1989 about any changes they had noticed in the frequency and nature of their dreaming since their injury or illness. Solms then integrated the results with the findings from 73 published studies in the neurological literature that mention deficits and excesses in dreaming. Twenty-nine of the 361 patients turned out to be free of any brain lesions. They were used as a control group because they had been faced with the possibility of brain injuries, admitted to the hospital, and subjected to the same routines and tests as the patients who did suffer lesions.

The responses from the remaining patients concerning changes in their dreaming were correlated with the findings from their neurological tests and brain scans. Solms then focused on those patients with focal brain lesions so that causal inferences about specific regions of the brain could be made. These analyses led to the conclusion that there are two different types of dreaming "deficits"-- loss of visual dreaming and complete loss ("cessation") of dreaming. There are also two types of dreaming "excesses"-the intrusion of dreaming into waking thought and increased nightmare frequency. It is noteworthy that all four types of changes in dreaming correlate with waking cognitive defects that are mentioned at appropriate points in the later discussion. In addition, they relate to relatively specific brain sites. The result is the general outlines of a neural network for dreaming that can be linked at many points to waking cognition on the one side and to the results of neuroimaging studies on the other (Solms, 1997; 2000). Figure 1.1 presents an overview of this network.

The Solms study provides seven specific findings relating dreaming and neurological structures. First, 200 of the 332 patients with brain lesions reported no changes in dreaming. This is highly useful information because it reveals those parts of the brain that are not necessary for dreaming. Instead of the diffuse cortical activity suggested by EEG recordings using scalp electrodes, the neural network for dreaming is surprisingly localized and does not include vast areas of the brain that are essential to waking cognition-- the dorsolateral prefrontal cortex, the sensorimotor cortex, and the primary visual cortex. For example, lesions in the dorsolateral prefrontal cortex that cause waking deficits of self-monitoring and decision-making have no effect on dreaming. These findings are supported and supplemented by a neuroimaging study revealing that all of these areas are as inactive during REM as they are during NREM, along with the opercular cortex and posterior cingulate cortex (Braun et al., 1997).

Second, Solms found changes in dreaming due to injuries in the medial occipito-temporal region of the visual association cortex in two patients. One lost all visual imagery in dreams for a short time, and the other was able to see static dream images from time to time. Both had highly specific lesion sites and deficits in waking mental imagery. These findings correspond with 13 cases that go back to the 1880s in the neurological literature, including cases with losses of facial imagery or color vision in dreaming. They also parallel findings in sleep laboratory studies by Foulkes and his colleagues. The case with no visual imagery in dreams is similar to Kerr, Foulkes, and Jurkovic's (1978) laboratory study of a patient with damage in her visual association cortex. This patient had neither waking mental imagery nor any visual imagery in her dreams. As to Solms' case with static dream imagery, it is described by him (1997, p. 105) as "strikingly reminiscent" of a second patient studied in the laboratory by Kerr and Foulkes (1981).

These findings are of theoretical interest because they show that the neural network for dreaming has considerable specificity. They are also of interest because the visual impairments in dreaming have parallel waking deficits, which suggests that the network has close relationships with at least some aspects of waking mental imagery. Once again, the neuroimaging findings are very consistent with the neuropsychological findings, because the visual association cortex-and the auditory association cortex as well-are reactivated during REM (Braun et al., 1998).

Third, Solms found 47 patients with either unilateral or bilateral injuries in or near the region of the PTO junction who reported complete loss of dreaming. They also showed a decline in waking visuospatial abilities. This discovery led to the hypothesis that the cortical network for spatial representation, centered in the inferior parietal lobes and important in the creation of waking mental imagery, is essential for dreaming. Solms also reports there is even some evidence that the left parietal region "contributes symbolic (quasispatial) mechanisms to the dream process whereas the right parietal region contributes concrete spatial mechanisms," but he also stresses that this claim needs further investigation (1997, p. 271). These results, which are supported by many individual cases in the literature, are also consistent with the neuroimaging findings, which show reactivation of the right inferior parietal lobe, an area thought to be important in spatial cognition (Maquet, 2000; Maquet et al., 1996). As shown later in the chapter, these findings also may provide a crucial link to developmental studies of dreaming.

However, any claim by a patient that dreaming has been lost raises the possibility that it may be memory for the dream that has failed. Evidence that these patients actually have ceased to dream comes first from the fact that those who reported loss of dreaming were no more likely to have memory disorders than those who reported that they continued to dream (Solms, 1997, pp. 160-161). Second, there are two laboratory studies of neurological patients in which awakenings from REM did not produce any dream recall in most participants. In the first study, 9 leucotomized schizophrenics who claimed they no longer dreamed were awakened from all REM periods during two nights in the laboratory and compared with a control sample of hospitalized schizophrenics who had not been leucotomized (Jus et al., 1973). Out of 66 awakenings, only two produced dream reports, and they were both from the same patient, whereas half or more of the awakenings with the control group led to dream reports. In the second study, only 3 of 12 patients who reported they had not dreamed over the course of a 10-day observation period could recall a dream from REM awakenings, as compared to 75 percent dream recall after REM awakenings by 7 neurological patients who said they continued to dream (Murri, Massetani, Siciliano, & Arena, 1985). It therefore seems safe to conclude that claims concerning the loss of dreaming are credible.

Fourth, Solms found that patients with bifrontal lesions in the white matter inferior to the frontal horns of the lateral ventricles, in the ventromesial region, also reported loss of dreaming. This area provides a crucial link between the basal forebrain and limbic structures on one side and many parts of the frontal cortex. The evidence for the importance of this area is based on only nine cases, but the finding is strengthened by the results from studies of leucotomized schizophrenics that were previously overlooked in the dream research literature. These studies report the loss of dreaming in 70-90% of the hundreds of schizophrenic patients who were leucotomized between 1940 and 1975 as a way to control their symptoms (Frank, 1946; 1950; Solms, 1997; 2000). Moreover, as just noted in the discussion of whether dreaming is actually lost, the absence of dreaming in this type of patient has been confirmed with awakenings in the laboratory during REM (Jus et al., 1973). Once again highlighting the parallels between waking cognition and dreaming, most of these patients were lacking in initiative, curiosity, and fantasy in waking life. These findings also fit with the neuroimaging studies, which show that the basal forebrain and limbic region are highly active during REM (Braun et al., 1997; Maquet, 2000; Nofzinger et al., 1997).

Fifth, 10 of Solms's patients reported an increased frequency and vividness of dreaming, often accompanied by the intrusion of dreaming into waking life. They also said that their dreams now seemed more realistic. The most frequently involved areas were the medial prefrontal cortex, the anterior cingulate cortex, and the basal forebrain. Some of these patients said that they felt like they were always dreaming, or that their thoughts quickly turned into pictures or realistic events, which suggests to Solms (1997, pp. 198-199), following an idea presented by Luria (1973), that they had lost the "selectivity of mental processes." Observations by members of the hospital staff support the idea that these patients were suffering from a confusion between dreaming and waking thought. Whitty and Lewin (1957) report several similar cases. Damasio et. al. (1985, p. 269) describe their patients with similar lesions as suffering from "waking dreams." It is also noteworthy that the medial prefrontal cortex is involved in processes of arousal and attention in waking life, and that injuries to this area can lead to confabulation and compulsive fabrication (Braun et al., 1997; Hobson et al., 2000b, p. 808).

Neuroimaging studies reveal that the medial prefrontal cortex, anterior cingulate cortex, and basal forebrain are reactivated during REM (Braun et al., 1997; Maquet et al., 1996; Nofzinger et al., 1997). Adding a new dimension to the picture, Nofzinger et al. (2001; 1999) discovered that depressed patients have much lower reactivation of the medial prefrontal cortex, anterior cingulate cortex, and right anterior insula in REM than normal participants. It is as if their neural network for dreaming has shrunk to a small core area. Since there is reason to believe that depressives may dream less (Armitage, Rochlen, Fitch, Trivedi, & Rush, 1995), and that the few dreams they do have are very bland (Barrett & Loeffler, 1992; Kramer & Roth, 1973), this finding demonstrates the potential of using atypical cases to learn more about the relationship between dream content and the functioning of the neural substrate for dreaming.

Sixth, Solms found that injuries to the temporal lobe caused increased nightmares of a repetitive nature for 9 patients, 5 of whom had symptoms of epilepsy. In keeping with this discovery, there are many instances in the literature of epileptics reporting nightmares, which are often caused by temporal-lobe seizures during NREM. These patients sometimes suffer from daytime hallucinations as well (Solms, 2000, p. 847, for a summary and references). Then, too, studies using stereotaxic electrodes to locate the sites causing seizures show that the "dreamy state" sometimes experienced as part of the diagnostic process is related to the temporal-limbic region. In one such study, the amygdala, anterior hippocampus, and temporal cortex were involved in every spontaneous occurrence of this state during the procedure (Bancaud, Brunet-Bourgin, Chauvel, & Halgren, 1994). Thus, the possibility arises that the seizures may be activating the neural substrate for dreaming (Hobson et al., 2000a, p. 1031; Solms, 1997, p. 243). In addition, the neuroimaging results are consistent with the inclusion of the temporal lobe in the neural network for dreaming. They show that the occipital-temporal region is reactivated during REM, along with nearby limbic areas (Braun et al., 1997; Maquet, 2000).

Seventh, and finally, Solms had 53 patients with brainstem lesions who were able to state whether or not they continued to dream. Forty-three "reported a preservation of the subjective experience of dreaming" (Solms, 1997, p. 154 ). Solms believes these results show that REM is not necessary for dreaming. However, his findings are not fully convincing on this issue because it is very difficult to eliminate REM even with experimental lesions in animals (Hobson, Stickgold, & Pace-Schott, 1998). Since these patients were not tested in the sleep laboratory for the absence of REM, it is therefore not certain that the relevant areas of the brainstem were affected. Further, it may be that "any lesion capable of destroying the pontine REM sleep generator mechanism would have to be so extensive as to eliminate consciousness altogether" (Hobson et al., 1998, p. R10). Still, there are reports of patients who lost REM and remained sentient (Gironell, Calzada, Sagales, & Barraquer-Bordas, 1995; Lavie, 1984; Lavie, 1990). Unfortunately, these patients were not awakened to see if they could report dreams, but their positive case histories show that future studies of the presence or absence of dreaming in people without REM may be possible.

As expected on the basis of earlier lesion studies with cats (Hobson, McCarley, & Wyzinski, 1975; McCarley & Hobson, 1975), the neuroimaging studies found that the pontine tegmentum is far more active in REM than NREM, with one study suggesting it is even more active in REM than in waking (Braun et al., 1997). This reactivation seems to extend to the thalamus through cholinergic pathways, and then to the basal ganglia, basal forebrain, and limbic/paralimbic regions.

These overall findings, while far more specific than anything that could have been imagined in the early 1990s, nonetheless leave open many questions about the exact contours of the dream network due to methodological differences in the various studies and possible individual differences in participants (Hobson et al., 2000b; Maquet, 2000). They also lead to inevitable speculation about the functioning of the network. Braun et. al. (1997, p. 1190) stress that the activation levels during REM are comparable to those in waking, but without the involvement of the frontal areas so important to waking cognition: "REM sleep may constitute a state of generalized brain activity with the specific exclusion of executive systems that normally participate in the highest order analysis and integration of neural information." They also emphasize the functional connections among the pons, basal forebrain, limbic structures, and medial prefrontal cortex during REM.

According to Maquet (2000, p. 222), the amygdala may be the central structure in the modulation of cortical activity in REM, as evidenced by the fact that it is tightly connected to the anterior cingulate cortex and the inferior parietal lobule, which are reactivated in REM, but has few connections to the dorsolateral prefrontal cortex and parietal lobes, which are relatively inactive throughout sleep. Meanwhile, Nofzinger et. al. (1997; 2001) highlight the importance of the anterior cingulate cortex, which plays a role in attentional states, performance monitoring, and error detection in waking thought.

Solms (2000) believes that the neuroimaging findings are generally very consistent with his neuropsychological findings, but doubts that the REM generator is a necessary part of the neural substrate for dreaming. Instead, he argues that dreaming is generated by the dopaminergic system that has its origins in dopaminergic cells in the ventral tegmentum, just above the pons, and then fans out to the amygdala, anterior cingulate gyrus, and frontal cortex. He agrees that the cholinergic pathways originating in the pons are the most frequent instigators of the necessary level of forebrain activation, but asserts that dreaming occurs "only if and when the initial activation stage engages the dopaminergic circuits of the ventromesial forebrain" (Solms, 2000, p. 849). Activation-synthesis theorists, on the other hand, see the neuroimaging results as strong support for their emphasis on the brainstem generator, while at the same time welcoming the insights into the forebrain network provided by both the neuroimaging and neuropsychological findings (Hobson et al., 2000a; Hobson et al., 2000b).

These slightly different perspectives share in common the idea that the association cortices, paralimbic structures, and limbic structures may operate as a closed loop to generate the process of dreaming. This is the starting point for the neurocognitive model proposed in this book. On the one hand, this subsystem is cut off from the primary sensory cortices that provide information about the external world, and on the other from the prefrontal cortices that integrate incoming sensory information with memory and emotion in the process of decision-making (cf. Braun et al., 1998, p. 94). This model implies that an unconstrained and freewheeling conceptual system can operate when there is sufficient activation. Its relative isolation may account for the "single-mindedness" of dreams, that is, the lack of parallel thoughts and reflective awareness (Rechtschaffen, 1978; Rechtschaffen, 1997). At the same time, as evidence presented throughout this book shows, the neural network for dreaming contains enough cognitive processing areas, such as the medial frontal cortex and anterior cingulate cortex, and perhaps the orbital-frontal cortex, to produce coherent dramatizations that often portray the dreamer's conceptions and concerns in waking life (Foulkes, 1985, pp. 209-213; Hall, 1953b). This emphasis on conceptions and concerns, based on inferences from detailed studies of dream content, provides the cognitive dimension that is lacking in activation-synthesis theory.

While the basic outlines of the neural network for dreaming seem clear, it is equally certain that much remains to be learned about its functioning. In addition to further neuroimaging studies using fMRI and transcranial magnetic stimulation as well as PET scans, this process could be greatly aided by clinical neuropsychologists who familiarize themselves with the new model and then look for patients with either pure lesions in relevant areas, or complaints about changes in their dreaming. Such studies could be especially helpful with patients who previously kept a dream journal and are willing and able to report dreams as their lesions heal.

For now, what seems certain is that progress toward an increasingly detailed mapping of the neural network for dreaming is inevitable in an era in which neuropsychology is making rapid strides and neuroimaging studies are becoming increasingly sophisticated and commonplace. Due to the potential of transcranial magnetic procedures, it might even be possible to shape dream content by stimulating different regions within the neural network for dreaming (Mazziotta, Toga, & Frackowiak, 2000; Stewart, Ellison, Walsh, & Cowey, 2001). The stage is therefore set for a consideration of how this network might be integrated with other areas of dream research to build a neurocognitive model. For example, the forebrain portion of the neural substrate for dreaming seems to be a good starting point for understanding the occasional occurrence of an awareness of dreaming during a dream.

The Question of "Lucid Dreaming"

The phenomenon of becoming aware of a dream while it is ongoing enjoyed a flurry of attention and speculation in the 1980s under the morally toned label of "lucid dreaming," implying a superior or elite status for "lucid dreamers," and efforts were made to link it to meditation and other altered states of consciousness (Gackenbach & Bosveld, 1989; LaBerge, 1985). While often remarked upon in books on dreams in the pre-laboratory era, lucid dreaming could not be studied systematically until it was shown in the laboratory that it occurs during REM (LaBerge, Nagel, Dement, & Zarcone, 1981). Although there have been too few people studied in the laboratory to establish the frequency of lucid dreaming, there are laboratory and non-laboratory studies suggesting that the degree of self-awareness and sense of conscious control can vary greatly from person to person and even within any given lucid dream (Barrett, 1992). However, as Foulkes (1990b, p. 121) notes, much more laboratory work needs to be done concerning "the conditions under which certain kinds of generic and autobiographical knowledge prove to be accessible during dreaming in the service of an ongoing comprehension and evaluation of dream events."

If dreaming is the form that consciousness takes during sleep (Foulkes, 1999), and if changes in the neural network for dreaming underlie different dreaming states, then lucid dreaming may be a product of a dream state in which the higher-order neural patterns that give human beings "core consciousness" and an "autobiographical self" are more active than usual (Damasio, 1999). This speculation is consistent with Rechtschaffen's (1997) use of the confabulations caused by frontal lobe injuries to argue that the loss of reflective awareness in dreams is due to the lack of frontal lobe activity. It also fits with the finding that higher levels of alpha activity during REM are related to lucid dream reports (Ogilvie, 1982; Tyson, Ogilvie, & Hunt, 1984), and with the fact that self-awareness during REM is associated with "phasic" (intermittent) activation within the REM period (Bradley, Hollifield, & Foulkes, 1992). The content of lucid dreams also has a more "realistic" nature, which would be expected from this line of reasoning (Gackenbach, 1988).

Then, too, it is noteworthy that dream reports in an exploratory PET scan study of l2 male participants showed a greater sense of control when the medial frontal cortex is more active, and a greater sense of things being out of control when the amygdala is most active (Shapiro et al., 1995). There is also non-laboratory evidence suggesting that the neural network for dreaming includes more frontal cortex during lucid dreaming: they seem to occur most frequently in the home setting after an early morning awakening-between 5 and 6:30 a.m.-- that is followed by imagery rehearsal and a conscious attempt to be aware of dreaming upon falling back to sleep (LaBerge, 1985). Thus, the new question of interest is the overall state of the neural network for dreaming during the experience of lucid dreaming, and how that overall state relates to indicators of REM and Stage II NREM (Hobson et al., 2000a, p. 1020; Hobson et al., 2000b. p. 837).

The Development of Dreaming Cognition

The serendipitous discovery of REM sleep in 1953, especially the finding that the four or five REM periods of the night occupy 20-25% of adult sleep time and lead to dream reports from 80-90% of awakenings in normal adults, triggered an enormous advance in the understanding of both sleep and dreaming (Dement & Kleitman, 1957a; 1957b; Foulkes, 1966; Kamiya, 1961). These studies demonstrate that dreaming is far more ubiquitous in both REM and NREM than any previous dream theorist ever imagined, which has major implications, discussed in Chapter 6, for traditional clinical and functional theories of dreaming. They also reveal that dreaming has many important parallels with waking cognition.

In addition, laboratory dream studies show that dreaming cannot be triggered by external stimuli, and that it is very difficult to influence dream content with either presleep stimuli, such as fear-arousing or exciting movies, or with concurrent stimuli administered during REM, such as a spray of water, sounds, or the names of significant people in the dreamers' lives (Foulkes, 1985; 1996a; Rechtschaffen, 1978). For example, in a large-scale study comparing the influence of neutral and affect-arousing presleep films on the REM dreams of 24 adult participants, only 5% of 179 awakenings showed any sign of incorporation (Foulkes & Rechtschaffen, 1964); similar results were obtained in a study of boys between ages 7 and 11 (Foulkes, Pivik, Steadman, Spear, & Symonds, 1967). In the most frequently cited study of the influence of external stimuli applied during REM, the sound of a bell was only incorporated on 20 of the 204 instigations with 12 participants (Dement & Wolpert, 1958a). Somatosensory stimuli from sprays of water, electrical pulses to the hand, and pressure from a blood pressure cuff seem to have the highest rate of incorporation at around 40 percent (Sauvageau, Nielsen, & Montplaisir, 1998, p. 132). However, the criteria for incorporation are very loose in most of these studies, sometimes including what are assumed to be metaphoric expressions of the stimulus (Arkin, 1991).

Overall, Foulkes's (1996a, p. 614) judgment seems to be the best starting point in terms of developing a neurocognitive model of dreaming: "Probably the most general conclusion to be reached from a wide variety of disparate stimuli employed and analyses undertaken is that dreams are relatively autonomous, or 'isolated,' mental phenomena, in that they are not readily susceptible to either induction or modification by immediate presleep manipulation, at least those within the realm of possibility in ethical human experimentation." However, on the rare occasions when stimuli are incorporated, "the speed and ingenious fit of the incorporation into the meaning and imagery context of the ongoing dream" are often "remarkable," suggesting the high cognitive level of the dreaming brain. (Antrobus, 2000b, p. 474).

Laboratory studies further show that the content of dream reports, whether from REM or NREM awakenings, is in large measure a coherent and reasonable simulation of the real world. This conclusion joins with the neuropsychological findings on brain-lesioned patients in suggesting there is a greater parallel between waking thought and dreaming than is assumed by either clinical or activation-synthesis theorists (Cavallero & Foulkes, 1993; Foulkes, 1985; Meier, 1993; Snyder, 1970; Strauch & Meier, 1996). Some of these laboratory findings on dream content are discussed in more detail in Chapter 2.

In addition, there are three laboratory studies suggesting that waking thought can have dreamlike qualities when participants are relaxing in a darkened room. In the first two of these studies, awake participants monitored by the EEG gave dreamlike responses to 15-20% of the requests for reports of what was going through their minds (Foulkes & Fleisher, 1975; Foulkes & Scott, 1973). In another laboratory study, judges who compared REM reports with thought reports from awake participants reclining in a darkened room rated the waking reports as more dreamlike (Reinsel, Antrobus, & Wollman, 1992; Reinsel, Wollman, & Antrobus, 1986). Furthermore, there is a field study of waking consciousness--using pagers to contact participants-- that discovered that 9% of the 1,425 thought samples had "more than a trace" of dreamlike thought and another 16% had a "trace" of such thought (Klinger & Cox, 1987/1988). Taken together, these studies lead to the idea that dreaming may not always be a function of sleep, thereby providing another possible linkage between waking cognition and dreaming. Instead, at least some forms of dreaming simply may require a high level of brain activation, combined with a reduction in external stimulation and a decrease in self-control (Antrobus, 1991; Foulkes, 1999; Llinas & Pare, 1991).

It is within the context of this general evidence for the overlap of waking cognition and dreaming that two-large scale studies of dreaming in children, one longitudinal, one cross-sectional, provide systematic evidence that can provide a developmental dimension to a neurocognitive model of dreams (Foulkes, 1982; Foulkes, 1999; Foulkes et al., 1990). The longitudinal study began with 7 boys and 7 girls ages 3-4 to cover the ages 3-9 over the five-year span of the study. It also included 8 girls and 8 boys ages 9-10 to account for the years between 9 and 15. Remarkably, all of the 14 children in the younger group participated in all five years of the study. Twelve of the 16 in the older group completed the study; the other 4 moved out of town.

To check on the possibility that participation in the study improved dream recall and accounted for any increases in the frequency and narrative complexity of dream reports, 6 boys ages 11-13 were added to the older group in the third year and 7 girls ages 7-9 were added to the younger group in the fifth year. The new participants generally did not differ on any dream measures from the original participants. In total, 26 children between the ages of 3 and 15 participated for five full years, 34 for at least three years, and 43 for at least one complete year. Normative dream data for each group were collected during the first, third, and fifth years of the study, when children slept in the laboratory for 9 nights each. They responded to 3 awakenings a night from either REM or NREM for a total of 2,711 awakenings. All the awakenings were carried out by Foulkes to insure experimenter consistency. During the second and fourth years the children participated in a variety of methodological studies, the most important of which compared dreams collected after a night of uninterrupted sleep in the laboratory with dreams collected in the morning at home by parents.

In addition to information on the frequency of dream recall and the content of the dream reports, a wide range of personality and cognitive tests was administered by other members of the project team. Information about school performance was obtained. Observations of the youngest group were made at a two-week nursery school during the first three summers of the study. A total of 657 non-dream variables were correlated with the dream data because "it would have constituted criminal neglect to have collected so many dream data and not to have searched far and wide for waking variables related to them" (Foulkes, 1999, p. 49).

The cross-sectional study focused on children ages 5-8 to see if the most interesting results of the longitudinal study could be replicated. It included 20 children at each age who were within one month of their birthdays, so a total of 80 children spent three nights in the sleep laboratory. They were each awakened 10 times, with all of the 800 awakenings once again carried out by Foulkes. The children also took several cognitive tests measuring visuospatial, verbal, descriptive, and memory abilities that had correlated with dream recall or length of dream reports in the first study. They also took three interview-based tests that claim to measure aspects of the development of self-awareness. In neither study did Foulkes know the results of the daytime tests until he had collected all of the dream data.

There are several replicated results from these two studies that are important for a neurocognitive model of dreams. None of the findings on rate of recall, report length, or narrative complexity showed any gender differences. First, and most unexpected, the median rate of dream recall was only 20-30% from REM awakenings until ages 9-11, when the median recall rate of 79% from REM awakenings approached adult levels. Recall from NREM awakenings went from 6% at ages 5-7 to 39% at ages 11-13. For both REM and NREM awakenings, recall came first from awakenings late in the night, then from awakenings in the middle of the night, and finally from awakenings early in the sleep period.

Second, the children's dream reports had very different content until ages 13-15 than what is reported by adults. For children under age 5, the REM reports consisted primarily of static and bland images in which they saw an animal, or were thinking about eating or sleeping. The dreams of children ages 5-8 showed a sequence of events in which characters moved about and interacted, but the dream narratives were not very well developed. The dreamer did not appear regularly as an active participant in her or his dreams until around age 8. Compared to the dream reports of adults, those of the young children were notable for their low levels of aggressions, misfortunes, and negative emotions (Domhoff, 1996; Foulkes, 1982; 1999). Gender differences in dream content did begin to appear in late childhood (Domhoff, 1996; Foulkes, 1982), but were more prevalent by adolescence (Trupin, 1976). The findings on children ages 9-15 have been replicated and extended in a major longitudinal project by Strauch (1996; 1999) based on 12 boys and 12 girls studied at two-year intervals in the sleep laboratory and at home.

The results on both recall and content are of great theoretical importance because they suggest that young children do not dream in the fashion assumed by all previous theorists on the basis of anecdotal and clinical accounts. Instead, they reveal dreaming to be a cognitive achievement that develops gradually in the same way most other cognitive abilities develop in children. The frequency and cognitive structure of children's dreams is not adult-like until ages 9-11, and the dream reports are not adult-like in length or content until ages 11-13. And the content is in general very different from what was expected on the basis of anecdotal accounts and non-laboratory studies.

Foulkes' findings on the waking correlates of dreaming and dream content in children provide further surprises because verbal and linguistic skills do not play a role until dreaming is fully developed, and none of the personality measures correlated with dream content until preadolescence. The one good and consistent predictor of the frequency of dream reporting in children ages 5-9 in both studies is visuospatial skills, as best measured by the Block Design test of the WISC and the Embedded Figures Test. This leads to the hypothesis that visual imagination may develop gradually and be a necessary cognitive prerequisite for dreaming.

There has been no rush to draw out the implications of these findings, perhaps because they do not agree with common sense: everyone has anecdotal examples of dreams from young children, and children seem to understand the concept of dreaming. In addition, as many as half of college students in two different samples claimed to remember a dream from childhood, although it may be significant that the average estimated age for such dreams is 6.5 years, as compared to the usual 3.5 years for their earliest memory (Domhoff, 1993a). Skeptics therefore argue that the low rates of recall in young children may be due to waking cognitive factors rather than a lack of dreaming.

For example, Hunt (1989) thinks the problem may be an inability to distinguish the "embedded" experience of a dream from similar subjective states; others say that children simply may lack the linguistic skills to translate the non-verbal experience of dreaming into the narrative report necessary to show evidence of dreaming (Hobson et al., 2000b; Weinstein, Schwartz, & Arkin, 1991). Foulkes finds these alternative explanations unlikely because none of the several linguistic, descriptive, memory, or story telling tests administered to the children correlated with rates of recall. Such explanations are also contradicted by the fact that both REM and NREM reports are first given late in the sleep period; it does not seem likely that either discriminatory or narrative skills would be unavailable earlier in the night once they had developed.

The idea that young children do not dream very well until their visuospatial skills are developed is supported by Foulkes' unanticipated findings with two of the boys ages 11-13 ages who were added to the study during its third year. Both of them had average memory and verbal skills, and both were adequate students in school, but both turned out be very low on visuospatial skills. Neither reported very many dreams during REM awakenings, far below the average for all other children in their age group. Since neither of these boys lacked the linguistic skills claimed by critics to be the reason why younger children do not report dreams when awakened in the laboratory, it seems more likely that they were not dreaming (Foulkes, 1982, pp. 180-181, 225-226).

Findings on the presence or absence of visual imagery in people who lose their sight through disease or accident before or after ages 5-7 also support Foulkes's argument. As is well known, those who become blind after this critical period "continue to be able while awake to conjure up mental images of persons, objects, and events, and they continue to dream in imagery" (Foulkes, 1999, p. 15). This point includes visual dream images of people they met after they became blind, which supports the generally accepted idea that they have a system of imagery independent of perceptual capabilities. On the other hand, people who become blind before ages 5-7 do not have waking visual imagery or visual dreams.

The likelihood that preschool children do not dream often or well may have implications for an unexpected finding in studies of how children come to understand imagination, pretense, and dreams. Several studies suggest that by age 3 children understand mental states and readily distinguish between the real and the imaginary. However, preschool children do less well on questions inquiring about dreams: "Whereas 3- and 4-year-olds are reported to have a sensitive understanding of the origins of imagination, early work on dreams suggests that children of this same age are quite confused about their origins" (Woolley, 1995, p. 195). Some 3-year-olds also "appeared to conceive of dreams as shared fantasies, claiming that dream content is shared between sleeping individuals" (Woolley, 1995, p. 189).

These differences are often explained by noting that pretense and imagination are deliberate mental activities that are facilitated by toys and interactions with adults. Woolley (1995, p. 195) speculates that "dream origins are simply more difficult for children," perhaps because dreams are not willful mental states. If Foulkes's findings are used as a starting point, however, this failure of understanding may be due to a lack of personal experience with dreams. In fact, many of the explanations for dreams offered by preschool children--that they are shared fantasies, that they come from God, that they are produced by the people who appear in them--seem to reflect what they are told by their parents, along with what they deduce from story books. This alternative hypothesis suggests there is a need for new research on the way in which children's "theory of mind" interacts with what they learn about dreams from their culture to produce possibly fabricated reports when they sense an expectation or pressure to tell a dream (Ceci, Bruck, & Battin, 2000).

Once children have the ability to dream, their linguistic and descriptive skills begin to correlate with the length and narrative complexity of their dream reports. Still, it is not until ages 11-13 that dream content shows any relationship to personality dimensions. For example, individualistic and assertive children portray themselves as more active in their dreams. Children with more violence in their waking fantasies have more aggressive interactions in their dreams, and those who display the most hostility before going to bed in the laboratory more often dream of themselves as angry. These findings on the continuity of dream content with waking thought support findings in earlier studies of children in the laboratory (Foulkes, 1967; Foulkes, Larson, Swanson, & Rardin, 1969; Foulkes et al., 1967). They suggest that dreams can reflect personal concerns and emotional preoccupations once there is an adequate level of cognitive development. As shown in Domhoff (1996) and by evidence presented throughout this book, this finding is all that remains of the large claims by Freud and Jung. This point is explored in more detail in Chapter 6.

Foulkes' overall findings raise the possibility that the development of dreaming may be based on the maturation of the neural network for dreaming discussed in the previous section. This hypothesis is the first and most crucial one in an effort to create a neurocognitive model of dreams. It is suggested most strongly by the parallel between the dependence of dreaming in children on visuospatial skills, which are based primarily in the parietal lobes (Robertson, 1998), and the loss of dreaming in adults with injuries to either parietal lobe. It is also suggested by the static nature of preschool children's dreams, which may relate to the absence of movement imagery in the dreams of adults with lesions in specific areas of the visual association cortex.

More generally, if the low levels of dreaming in children and the differences in their dream reports from normative adult findings are treated as if they are "deficits," then the search could be made for possible causal "defects" in the neural network necessary for dreaming. This strategy has been followed by Welsh, Pennington, and Groisser (1993) in studying the development of frontal lobe executive functions in children, employing neuropsychological tests in conjunction with standard developmental tests. It could be widened to include neuroimaging studies of the developing brain and myelination studies as well (Chugani, 1999; Paus, Zijdenbos, Worsely, Collins, & others, 1999; Rivkin, 2000; Thatcher, 1996). Indeed, the fact that myelination of the inferior parietal lobules is not functionally complete until ages 5-7 may be part of the reason why dreaming is not fully developed until after that age period (Janowsky & Carper, 1996; Solms, 1999). The fruitfulness of this approach is also seen in studies showing that the presence or absence of visual imagery in blind adults depends upon whether they lost their sight before or after age 5-7 (Hurovitz, Dunn, Domhoff, & Fiss, 1999).

The integration of the neural network for dreaming with Foulkes' developmental findings would provide a solid basis for a model that is genuinely "neurocognitive" instead of simply "neuropsychological," in the sense that it would have the potential to relate a general neural system to the general development of dreaming. However, it is necessary to include what is known about the nature of dream content before the cognitive dimension of dreams can be folded into modern-day cognitive theory and then incorporated into a neurocognitive model of dreams.

The Nature of Dream Content

Although there are several systems of content analysis that have made one or more contributions to the overall understanding of dream content (Foulkes & Shepherd, 1971; Gottschalk & Gleser, 1969; Winget & Kramer, 1979), the largest and most systematic body of findings on what people dream about comes from a comprehensive set of descriptive empirical categories developed by Hall (1951), and then finalized with the help of Van de Castle (Hall & Van de Castle, 1966). There are four general findings with this Hall/Van de Castle system that must be encompassed by a neurocognitive model.

First, several different studies reveal that the dream lives of college men and women in the United States remained the same throughout the second half of the 20th century despite major cultural changes (Domhoff, 1996; Dudley & Swank, 1990; Hall, Domhoff, Blick, & Weesner, 1982; Hall & Van de Castle, 1966; Tonay, 1990/1991). Second, there is little or no change in dream content with age once adulthood is reached. That is, older dreamers do not differ from college students, except perhaps for a decline in physical aggressions and negative emotions (Cote, Lortie-Lussier, Roy, & DeKoninck, 1996; Hall & Domhoff, 1963; Hall & Domhoff, 1964; Howe & Blick, 1983; Strauch, 2000; Zepelin, 1980). Nor does dream content change much in longitudinal studies of dream journals provided by adults, a claim that holds true for periods as long as four or five decades and for people still keeping journals in their seventies (Domhoff, 1996; Hall & Nordby, 1972; Lortie-Lussier, Cote, & Vachon, 2000; Smith & Hall, 1964).

The third relevant result with the Hall/Van de Castle system is that there is a stable pattern of cross-cultural similarities and differences in dream content. The vast body of research on which this conclusion is based, much of it unpublished work by Hall, is summarized in Domhoff (1996, Chapter 6). There have been no significant additions to this literature since that time. Everywhere in the world, for example, women and men have the same differences in the percentage of gendered characters who are males or females. Women dream equally of men and women, but 67 percent of the gendered characters in men's dreams are other men (Hall, 1984). The same gender differences are also found in short stories by male and female authors (Hall, 1963), and in stories told by pre-school children (Domhoff, 1996, p. 89). The low percentage of males in the dreams of Japanese women seemed to be the major exception to this generalization (Yamanaka, Morita, & Matsumoto, 1982), but a later study of several different samples reveal the same percentages as elsewhere (Nishigawa, Brubaker, & Domhoff, 2001).

For both men and women cross-culturally, there is usually more aggression than friendliness, more misfortune than good fortune, and more negative emotions than positive emotions. In addition to these similarities, there are also a few differences that make sense in terms of large-scale cultural differences. For instance, small traditional societies have a higher percentage of animal characters. Further, there are large variations from society to society in the percentage of all aggressive interactions that are physical in nature, although it is also the case that men in most societies have a higher physical aggression percent than women (Domhoff, 1996; Gregor, 1981; O'Nell & O'Nell, 1977).

Finally, studies of dream journals have demonstrated wide individual differences on a variety of Hall/Van de Castle content indicators that are explained in Chapter 3. These differences generally relate to the waking concerns or past emotional preoccupations of the dreamers. Thus there is a continuity between most aspects of dream content and waking thought (Bell & Hall, 1971; Domhoff, 1996; Hall & Lind, 1970; Hall & Nordby, 1972). This finding leads to the hypothesis of a "continuity principle" operating in dreams that is compatible with Foulkes's (1967; 1982; 1999) findings in laboratory studies with both children and adults.

The continuity principle is best demonstrated by blind analyses of dream journals, where nothing is known about the dreamer until she or he later answers questions developed on the basis of the content analysis. In particular, blind analyses lead to accurate portrayals of the dreamers' conceptions and concerns in regard to the important people in their lives. This emphasis on questions developed from the results of content analyses follows from three conclusions based on earlier attempts to find correlations between dream content and standard personality measures. First, the findings with projective techniques are meager and inconsistent (Domhoff, 1996; Hall, 1956), which may be a function of the inadequacies of these instruments (Lilienfeld, Wood, & Garb, 2000). Second, the results with structured personality tests, while usually consistent with the continuity principle, did not lead to new insights, so such tests were seldom used after the early 1970s (Domhoff, 1996). Third, past research shows that dreams most directly reveal concerns, interests, and worries rather than personality traits, which suggests that an open-ended neurocognitive approach may be more useful at this juncture (Domhoff, 1996, Chapter 8; Hall, 1953c; Hall & Nordby, 1972). Chapter 5 explains how studies using questions based on a dream series are conducted.

Several of the discoveries with the Hall/Van de Castle system, and especially the consistency in what adults dream about throughout their lives, lead to the idea that there is a "repetition principle" that is operating in the dream process at least some of the time (Domhoff, 1993b; Domhoff, 1996). This tendency to repeat has gone unnoticed by those who study one dream at a time with clinical patients, use samples of individual dream reports from groups of people, or hold to Jung's (1974) theory that a dream series shows a pattern of symbolic change toward greater personal integration. However, the relative absence of the repetition of themes in dreams collected over several weeks from participants in laboratory studies suggests that the pervasiveness of repetitive dreaming may be overestimated by selective dream recall in everyday dream journals (Foulkes, 2001). Thus, the fact of repetition is solidly established, but its relative frequency remains to be determined.

The idea of a repetition principle in dreams not only describes the consistency over years and decades in characters, social interactions, activities, and settings in the longitudinal studies using the Hall/Van de Castle system. It also encompasses three other repetitive aspects of dream life that must be comprehended within a neurocognitive model of dreaming. First, there is a large clinical literature on the repetitive nightmares of people suffering from post-traumatic stress disorder that fits well with the idea of a repetition principle (Hartmann, 1984; Hartmann, 1998; Kramer, 2000b; Kramer, Schoen, & Kinney, 1987). This literature shows that such dreams are more frequent and persistent than was realized until systematic studies began in the aftermath of the Vietnam War (Barrett, 1996).

Second, the repetition principle can encompass the recurrent dreams that 50-80% of people claim to have had at one time or another in their lives, often starting in late childhood or early adolescence, sometimes lasting for the rest of their lives, and usually highly negative in content and emotionally upsetting (Cartwright & Romanek, 1978; Domhoff, 1996; Zadra, 1996). Third, the idea of a repetition principle can incorporate the repeated themes found in most series of 20 or more dreams (Hall, 1947; 1953c). In other words, it is not just Hall/Van de Castle indicators that are consistent over many years, but also more general themes, such as being lost, preparing meals, or being late for an examination. In a study based on 649 dreams over a 50 year period, for example, just six themes accounted for at least part of the content in 71% of the dream reports (Domhoff, 1993b).

The concept of a repetition principle suggests several potential linkages between dream content and the neural substrate for dreaming, particularly in terms of its possible relationship with the vigilance/fear system that seems to be centered in the amygdala (Le Doux, 1996; Whalen, 1998). The best examples of this point, of course, are the repetitive nightmares of post-traumatic stress disorder. These nightmares sometimes happen in Stage II of NREM (Van der Kolk, Blitz, Burr, Sherry, & Hartmann, 1984) and seem to have parallels with the nightmares suffered by epileptics due to seizures in NREM (Solms, 1997; 2000). In addition, as noted in the discussion of the neural network for dreaming, the dreamy states sometimes experienced by epileptics are usually related to the temporal-limbic region (Bancaud et al., 1994). Thus, future neuroimaging work on both post-traumatic stress disorder and epilepsy may hold promise for linkages between the repetition principle and the neural network for dreaming.

However, there need not be an exclusive focus on patients. The consistency of emotionally painful themes and of heightened scores on Hall/Van de Castle indicators in the dreams of many normal participants suggests that their dream life is often "stuck" in the past in a way that fits with the persistence of negative memories stored in the vigilance/fear system (Domhoff, 1996, and Chapter 5). Both dreams and the vigilance/fear system seem to provide a neurocognitive record of traumas, upsets, and tensions over a lifetime. Moreover, both may persist even when the person is emotionally recovered and unhampered by the past during waking life. This possibility suggests that dreams may not always be symptomatic of present-day problems, contrary to what all clinical theories assume.

Systematic studies showing the effects of different drugs on dream content, when done in conjunction with neuroimaging studies, might help to pinpoint relationships between repetitive dream content and specific components of the dream-generation network. The promise for such studies is seen in the fact that both the anticholinergic beladonna alkaloids (Ketchum, Sidell, Crowell, Aghajanian, & Hayes, 1973; Wichlinski, 2000) and dopamine (Hartmann, Russ, Oldfield, Falke, & Skoff, 1980; Solms, 2000) intensify the dream experience. Patients suffering from epilepsy or Parkinson's Disease might be potential candidates for such content studies because it already is known that the medications that eliminate epileptic seizures also reduce or eliminate the patients' nightmares, and that L-dopa potentiates the dream experience for Parkinson's patients (Hartmann, 1984; Perry, Walker, Grace, & Perry, 1999; Solms, 1997).

Although earlier studies concerning the effect of drugs on dream content led to few clear results for a variety of reasons (Roth, Kramer, & Salis, 1979), the potential for pre/post studies of individual cases is shown in the large positive changes in the dream content of a 21-year-old woman after she began taking sertraline (Zoloft), an antidepressant selective serotonin reuptake inhibitor, to cope with anxiety attacks (Kirschner, 1999). These positive changes include more friendly interactions, fewer aggressive interactions, and fewer negative emotions. It is also of interest that she showed a decline in "elements from the past," which might be an indication that the repetition principle is having less influence on her dreams.

Dream content and the neural network for dreaming also might be linked by investigations that correlate specific neurological defects with atypical scores on Hall/Van de Castle indicators. Patients who have suffered damage to the amygdala might be ideal candidates for future defect studies because they have lost their capacity for fear in waking life and express predominantly positive emotions (Adolphs & Damasio, 1998; Damasio, 1999; Pace-Schott, 2000). It therefore could be hypothesized on the basis of the continuity principle that their negative emotions percent would be far lower than the 80% figure that has been found in several different studies (Hall et al., 1982; Hall & Van de Castle, 1966; Roussy, Raymond, & De Koninck, 2000b; Tonay, 1990/1991).

The potential for such studies is demonstrated in older reports cited by Solms (1997) showing a decline in "narrative complexity" in the dream reports of patients with specific defects through injuries or operations. It is also seen in a study showing that 17 male chronic brain syndrome patients had more family members, less aggression, and less emotional content in the 31 dreams they reported than does the Hall/Van de Castle normative sample (Kramer, Roth, & Trinder, 1975). This pattern of findings suggests that their dreams were very bland, a characterization that fits with the waking personalities of such patients (Torda, 1969). It might even be that there is a different profile on Hall/Van de Castle indicators for each type of defect, a possibility that is demonstrated in a sample of 104 dream reports from 20 male schizophrenics (Domhoff, 1999b, p. 127)

It also could be useful to look for changes in dream content as the process of dreaming returns in patients with injuries to one or the other parietal lobe. It might be that content is more simple and banal at first, reflecting only a partial recovery. Then, too, this approach could be used to test the idea that the left parietal lobe is more involved in symbolic (quasispatial) constructions and the right parietal lobe in concrete spatial constructions (Solms, 1997, p. 271).

Dream Content and Waking Cognition

Findings from the study of dream content not only suggest links with the neural network for dreaming, but also with waking cognition. In particular, the continuity principle provides the same kind of strong connection between dreaming and waking cognition that has been demonstrated by the neuropsychological and developmental evidence presented earlier in this chapter. This continuity leads to the hypothesis that both dreaming and waking cognition are dealing with the same psychological issues to a large extent. This hypothesis provides the basis for linking a neurocognitive model of dreams with what is known about waking cognition.

However, as shown by the evidence concerning the repetition principle in the previous section, the continuity principle does not operate entirely in terms of current personal interests and concerns. Dream content is also continuous in varying degrees for different individuals with past waking concerns. Discrepancies between current waking concerns and current dream content, such as dreaming about painful events that are no longer thought about in waking life, could be used to see how the continuity and repetition principles interact with each other to shape dream content.

The starting point for adding a cognitive dimension to the model is the concept of a "conceptual system," or system of schemas and scripts, which is the organizational basis for all human knowledge and beliefs. Most of this system is thought to be unconscious, in the sense of being outside of conscious awareness, but some of it can become conscious as well. It consists of both experientially based and figurative concepts, both of which are processed and understood equally fast and well according to experimental studies (Gibbs, 1994; Gibbs, 1999). The conceptual system builds on three types of experiential categories that are based upon bodily sensations and interactions with the world: basic level, spatial relations, and sensorimotor (Lakoff & Johnson, 1999).

Basic-level categories arise through the interaction of inherited neural structures with patterns of stimuli from the environment. They reflect distinctions among types of animals, such as cows, horses, and goats, or types of social interactions, such as friendly and aggressive interactions, or types of actions, such as walking and running. Basic-level categories are most directly distinguished from other categories by the fact that a single mental image can represent an entire category, such as a "dog" or a "cat," a "boat" or a "car" (Murphy & Lassaline, 1997). In addition to the large number of basic-level categories, there are also spatial relations categories that are experiential in nature, such as "up," "down," "in front of," and "in back of." Comparative linguistic studies show that "there is a relatively small collection of primitive image schemas that structure systems of spatial relations in the world's languages" (Lakoff & Johnson, 1999, p. 35). Finally, there are sensorimotor categories that are based on direct experience of such varied qualities as temperature, motion, and touch.

Dreams are thought of as highly "symbolic" in many different cultures, including Western civilization, but the findings from content analysis suggest that dreams may consist primarily of constructions arising from experiential categories. Based on his reading of thousands of dreams collected from children, teenagers, and adults in the sleep laboratory, Foulkes (1985) concludes that most dreams are "simulations" of real-world experiences. Young adult dreamers are often shopping, playing sports, visiting with their friends, arguing with their parents, worrying about the faithfulness of their lovers, or feeling tempted to be unfaithful themselves. The content of young children's dreams is usually even more realistic.

Although the Hall/Van de Castle coding system is accurately described as "empirical" and "descriptive," it is noteworthy that most of its coding categories are basic-level categories. This point holds true for all of the social interaction, activity, and emotions categories, and for most of the character categories. This coding system therefore makes good theoretical sense to the degree that dreams are constructed from experiential categories. Perhaps this focus on basic-level categories also explains why the system can be learned and used with high intercoder reliability by new researchers in many different countries.

The theory of cognitive functioning sketched out in the above paragraphs provides a basis for adding a cognitive dimension to the neurocognitive model because it fits well with earlier work on dream content by Hall (1953b) Foulkes (1985), Antrobus (1978; 1991), Fiss (1986), and other cognitive dream researchers. The model begins with the proposition that dreaming is what the mature brain does when (1) the neural network for dreaming outlined earlier in the chapter is at an adequate level of activation; (2) external stimuli are occluded, and (3) the self has been relinquished (Foulkes, 1999). This view accounts for dreaming at sleep onset, in REM sleep, and at times of sufficient activation during NREM (Antrobus, 2000b; Antrobus et al., 1995; Vogel, 1991). It also accounts for the fact that there is sometimes dreaming in awake participants who are resting quietly in a darkened sleep laboratory, where EEG recordings verify that the participants are in fact awake (Foulkes & Fleisher, 1975; Foulkes & Scott, 1973).

Once instigated, dreaming draws on memory schemas, general knowledge, and episodic memories to produce reasonable simulations of the real world (Antrobus, 1991; Foulkes, 1985; Foulkes, 1999), with due allowance for an occasional highly unusual or extremely memorable dream (Bulkeley, 1999; Hunt, 1989; Knudson & Minier, 1999; Kuiken & Sikora, 1993). Generally speaking, these simulations express the dreamer's "conceptions," which are also the basis for action in the waking world from the standpoint of cognitive theory. In particular, dreams express several key aspects of people's conceptual systems, especially self conceptions and conceptions of family and friends (Hall, 1953b).

The emphasis is on conceptions of "self" and "others" because studies of adult dream content show that dreams reflect relatively little about a person's attitudes toward current events and politics (Hall, 1951). Similarly, (Foulkes, 1982; 1999) found that children between ages 5 and 15 dreamed very little of their two most time-consuming daytime activities, going to school and watching television; instead, they dreamed about recreational activities. An emphasis on the highly personal nature of dreams may explain why the dreams of college students in the United States have not changed over the past 50 years; the culture has changed, but personal concerns probably remain very stable. This emphasis also may explain why dreams are more similar than they are different around the world. As anthropologist Thomas Gregor (1981, p. 389) suggests at the conclusion of his detailed study of 385 dream reports from men and women in a very small native group deep in the Amazon jungle, "it may be possible to show that the dream experience is less variant than other aspects of culture."

Starting with the idea that dreams usually express highly personal conceptions, it is possible to build a complex picture of a dreamer's overall conceptual system because people usually have more than one conception of themselves and the important people in their lives. Moreover, these conceptions of self and others can be contradictory as well as numerous. Some of the apparent contradictions may disappear, however, if closer analysis shows that a parent is seen as "supportive" in some contexts, such as when facing exams or problems at work, but "restrictive" in others, such as when the dreamer wants to engage in sexual activities (Hall, 1947). It is possible that conceptual maps of the dreaming mind based on content analysis findings could be expressed in network terms (e.g., Markman, 1999; Osgood, 1959).

In addition, it is possible that the use of conceptions is more diffuse during dreaming because the cognitive system is unconstrained by the requirements of the waking world (Foulkes, 1985). This hypothesis might help account for the repetitive nature of dream content related to significant people and interests. It is as if the "updated" versions of key concepts are no more likely to be used than the older ones.

This neurocognitive model also contains a way to assess the weight to be given to the conceptions expressed in dreams: by determining the relative frequency of their occurrence. Since findings with the Hall/Van de Castle system show that frequency reveals the "intensity" of a "concern" or "interest," it can be said that dreams reveal both "conceptions" and "concerns," and therefore have at least some degree of psychological "meaning." This point further integrates the Hall/Van de Castle coding system with a neurocognitive model because its categories not only relate to basic-level concepts, but its frequencies relate to conscious concerns.

Even though dreams seem to be based to a large extent on experiential-level categories, the emphasis in a neurocognitive model on the close parallels between waking thought and dreaming raises the possibility that some of the unusual and not immediately understandable features of dreams may be the product of figurative thinking: conceptual metaphors, metonymies, ironies, and conceptual blends (Fauconnier, 1997; Gibbs, 1994; Lakoff & Johnson, 1999). Figurative concepts are sometimes thought of as mere embellishments of speech that are not necessary for thinking, but many cognitive scientists now see them as an important part of people's conceptual system due to a wide range of experimental studies summarized by Gibbs (1994), many of them carried out by him and his students. This system of conceptual metaphors is learned anew by each individual due to repeated experiences within the course of childhood development.

Lakoff and Johnson(1999) estimate that there may be "hundreds" of "primary" conceptual metaphors that "map" well-understood experiential categories--the "source" domain--to more complex or abstract matters of human concern--the "target" domain. For example, basic experiences like warmth and motion are used to understand more difficult concepts like "friendship" ("they have a warm relationship") and "time" (time often "goes by slowly," but sometimes "time flies by"). Just as in waking thought, figurative thinking may be used in dreams when it expresses a conception better and more succinctly than an experiential concept does (Hall, 1953a; Lakoff, 1997). This idea also provides a plausible explanation for why there seem to be many different metaphoric expressions in dreams for one "referent:" each metaphor provides a slightly different conception of the referent object.

One avenue into the possibility of a linkage between waking figurative thought and dream content might be found in "typical" dreams, such as flying under one's own power or finding oneself inappropriately dressed in public. A content analysis of 983 dream reports in two-week journals kept by 126 students in a college course demonstrates that flying dreams accounted for only 0.5 percent of the total, and the figures for other typical dreams-such as teeth falling out, falling in space, or finding money-are even lower (Domhoff, 1996, p. 198). However, several survey studies suggest that at least a significant minority of respondents claim to have had one or more of such dreams (Griffith, Miyago, & Tago, 1958; Nielsen, Zadra, Germain, & Montplaisir, 1999; Ward, Beck, & Rascoe, 1961). These rare dreams may be examples of "primary" metaphors, which are based on repeated correlations between two dimensions of experience that are common in childhood development. For example, tasting something sweet (a physiological process), and then experiencing pleasure (an emotion), leads to the metaphor that "Pleasing is Tasty" (Grady, 1999).

Consider dreams of flying under one's own power, which are reportedly experienced by a little over half of college students in two surveys, and said by them to be generally positive in feeling tone (Domhoff, 1996). Searching for a metaphor related to flying, the possibility arises that these dreams may be instances of the primary metaphor "Happiness is Up," as found in such expressions as "high as a kite," "walking on air," and "floating on cloud nine." This speculation also fits with the fact that people sometimes become apprehensive about falling during their positive flying dreams, just as people worry that they may "crash" or "have the air let out of their balloon" when they are too elated in waking life.

Similarly, it may be that dreams of appearing inappropriately dressed in public, which are reportedly experienced by 40 to 50 percent of college students, usually in their mid-teens, and sometimes more than once, are instances of the conceptual metaphor "Embarrassment is Exposure" (Domhoff, 1996, p. 203). This metaphor is expressed through such well-known phrases as "caught red-handed," "caught with egg on your face," and "caught with your pants down" (Holland & Kipnis, 1994). It might be evidence for this conjecture that when college students are asked to write down the dream in which they experienced the greatest feeling of embarrassment, they most often spontaneously report one in which they are inadequately attired in a public place (Domhoff, 1996).

These two hypothetical examples aside, the few attempts to undertake systematic studies of metaphor in dreams suggest that most dreams do not seem to relate very obviously to primary metaphors (Hall, 1953a). Most dreams seem more like dramas or plays in which the dreamer acts out various scenarios that revolve around a few basic personal themes (Greenberg & Pearlman, 1993; Hall, 1947). They seem to be instances of the "thematic" point on the repetition dimension, that is, specific "episodes" or "examples" relating to more general emotional preoccupations, usually negative in nature. They appear to take the form of proverbs or parables, which can be understood only by extracting "generic" information from specific stories (Lakoff, 1993a; Lakoff & Turner, 1989).

These more complex dreams may rely on "resemblance" metaphors, which depend upon the perception of the common aspects in two representational schemas (Grady, 1999), or on conceptual blends that often start with basic conceptual metaphors and then are elaborated into highly novel thoughts (Grady, Oakley, & Coulson, 1999). Hall (1953a) has shown that blind analyses of a series of dreams can lead to very plausible and potentially verifiable inferences when figurative forms of thought relating to a major concern are utilized several times in the dream series. To take his best example, a young woman who provided a series of dreams had an especially striking one in which she was searching for her wedding gown because she and her husband were to be married again on their first wedding anniversary. However, she was very disappointed when she found the gown: it was dirty and torn. With tears in her eyes, she put the gown under her arm and went to the church, only to have her husband ask why she had brought the gown. She reports she was "confused and bewildered and felt strange and alone" (Hall, 1953a, p. 179).

Looking at the dream from a figurative point of view, Hall hypothesized that the state of the dress might express her conception of her marriage. In today's terms, the dream may be a conceptual blend based upon a metonymy. To test this hypothesis, Hall looked to see if there were other dreams in the series that might suggest the marriage is in difficulty, and there were several: (1.) the stone from her engagement ring is missing; (2.) her husband has tuberculosis; (3.) one of her women friends is going through a divorce; and (4) a friend who is about to be married receives a lot of useless bric-a-brac for wedding presents. If the Hall/Van de Castle system had been available when this analysis was made, the case could have been improved by comparing the dreamer's aggressions-per-character ratio with her husband to the same ratio with other adult males. If it were higher with her husband than with other adult males, and if there were a lower rate of friendly interactions as well, then the metaphoric hypothesis would have been supported by means of a non-metaphoric content analysis.

Two later chapters provide methods that might aid in the search for figurative meaning in dream content. Chapter 4 suggests new ways to do empirical studies on metaphors in dreams, using sophisticated software to search for phrases and strings of words in large numbers of on-line dream reports. Chapter 5 presents findings with this method as one part of a larger study of 3,116 dreams from one person over a 20-year period. Since some of the findings presented there contradict the continuity principle, it may be that the contradictory findings involve dream elements that are figurative in nature. For example, the series contains instances of the dreamer riding horses or shooting guns well, but contrary to expectations, she does not ride or shoot in waking life, and is fearful of both horses and guns.

The possibility that some dreams may be based on figurative thinking provides a way for a neurocognitive model to incorporate the interesting idea that past experiences are sometimes used as personal metaphors to express current conflicts that have similar emotions and feelings at their core (Kramer et al., 1987). This idea comes from a study of Vietnam veterans who had recovered from their post-traumatic stress disorder. However, they later came back to the Veterans Administration for help when war-related themes began to appear in their dreams in the face of new life stressors, such as marital conflict, conflicts with children, or work-related tensions. In effect, these new war-related dreams may be conceptual blends that combine past experiences with aspects of the stressful situations the veterans are now enduring. The resemblance is in the similarity of the feelings in both the war and the new situation. "It's a war zone out there," they might be thinking in relation to their current problems.

If dreaming is in part figurative, especially in terms of primary metaphors, resemblance metaphors, metonymies, and conceptual blends, then a neurocognitive model could advance in parallel with new understandings in cognitive linguistics. However, it still would be necessary to do the same kinds of thematic and Hall/Van de Castle content analyses of a dream series that have been carried out in the past in order to understand any given series of dreams. This is in part because many resemblance metaphors and most conceptual blends are likely to be unique to the dreamer. In addition, to the degree that dreams are like proverbs and parables, then it remains necessary to study many dreams in a search for the "generic" or underlying pattern.

For now, it needs to be stressed that there is little or no systematic evidence that dreams make use of the vast system of figurative thought available to most individuals in waking life through a combination of developmental experiences and cultural heritage. Of all the possible linkages among the three areas of dream research suggested in this chapter, the idea that dream content studies may provide bridges to waking figurative thought is by far the most speculative. It is also an issue that divides dream theorists. For example, both Foulkes (1999, p. 110) and Hobson, Pace-Schott, and Stickgold (2000b), who disagree on many issues, are together in doubting that unusual constructions in dreams are meaningful. For Foulkes, they reveal the limited nature of the cognitive abilities possessed by the sleeping brain. For Hobson et. al, they are often reflections of the unique neuromodulation of the neural network for dreaming, a form of delirium during sleep.

Even if it turns out that dreams make little or no use of figurative thought, a cognitive theory is useful in explaining why dreams hold great fascination for many people in many different cultures: dreams seem to have parallels with waking figurative thought. The parallels with the metaphoric dimensions of waking thought may be why some societies have made use of dreams in their cultural practices and rituals. In that sense, dreams have "emergent" uses that have been developed in the course of history and passed on through culture. This view also explains the use of dreams in psychotherapy: dream interpreters make use of metaphoric interpretations that are plausible to the client. It may be that dreams simply provide a platform from which the client and therapist can develop a new narrative about the client's life through a process of negotiation about metaphoric meanings. For the foreseeable future, then, metaphoric interpretations are the fool's gold of dream theories. With their glitter of seeming insight and the accompanying feelings of enrichment and closure, metaphoric interpretations deceive interpreters and dreamers alike.

A neurocognitive model also can incorporate the unexpected finding that nightmares often can be eliminated by having people write out and visually rehearse a new ending of their own choosing for the dream (Krakow, Kellner, Pathak, & Lambert, 1995). This process may be an instance of the cognitive distancing that many people achieve by writing about personal feelings and events (Pennebaker & Graybeal, 2001; Pennebaker & Keough, 1999; Pennebaker & Seagal, 1999).

In closing this discussion of dreaming and cognition, it is worth mentioning that a new neurocognitive model might turn out to be useful in understanding the development of consciousness. Foulkes (1990a; 1999) offers this fresh idea based on his cross-sectional study of children ages 5-8. If it is assumed that dreaming is the form that consciousness takes during sleep, then the origins of consciousness can be explored by doing detailed studies of the development of the ability to dream. As one part of this general idea, Foulkes further suggests that the ability to include oneself in a dream, which is not fully developed until around age 8, may be an index of when a child has a full sense of self. He reached this conclusion after finding that three waking tests designed to assess the development of the "self" concept did not correlate with each other and did not predict the inclusion of the dreamer as a character in his or her dream reports (Foulkes, 1999, p. 95).


There are other possible linkages among the three areas of dream research discussed in this chapter. However, enough has been said to demonstrate that there is a large body of established empirical findings upon which to base a new model. Moreover, the research tools are now available to do the many studies that would be necessary to test and develop the model. The rapid advances in neuroimaging and neurochemistry are the most obvious examples of this point. As noted earlier, the growing number of neuropsychologists in clinical settings are also a potentially important resource for developing this model, because they could easily screen for changes in dreaming as they examine patients with pure lesions in relevant areas of the brain.

The advent of personal computers and the constant improvements in software are also important because they have made content analysis somewhat less labor-intensive and far more accurate than it was in the past. These advances include a spreadsheet that calculates all the Hall/Van de Castle content indicators (Schneider & Domhoff, 1995). This spreadsheet is discussed as part of Chapter 3. It is also now possible to use a new search program to find single words, strings of words, or phrases in over 11,000 dream reports available on (Schneider & Domhoff, 1999). This search program may prove especially useful for making the metaphoric studies that are necessary to determine the degree to which there are limits to the meaning in dreams. The use of for new approaches to content analysis is demonstrated in Chapter 4.

Return to