The Neurocognitive Theory of Dreams at Age 20 (PDF)

The Neurocognitive Theory of Dreams at Age 20: An Assessment and a Comparison With Four Other Theories of Dreaming G. William Domhoff University of California, Santa Cruz This article assesses the neurocognitive theory of dreams on the occasion of its 20th anniversary. The theory synthesizes findings from 3 strands of dream research, which focus respectively on the neural substrate that subserves dreaming, the development of dreaming in children, and quantitative findings on adult dream content, all 3 of which are necessary ingredients in any theory in cognitive neuroscience (Ochsner & Kosslyn, 2014). The article compares the current standing of the theory with that of 4 other theories with a neural dimension: the Freudian, activation-synthesis, memory-consolidation, and threat-simulation theories of dream- ing. It concludes that the neurocognitive theory differs from the other 4 in that many of its key building blocks were created and have since been replicated by independent investigators in 3 different research areas. The other theories lack a developmental dimension, and their claims sometimes do not accord with established findings on dream content. On the other hand, the neurocognitive theory has been strengthened by neuroimaging findings revealing that the neural substrate that enables dreaming is a subsystem of the default network, which supports imagination in waking; it also includes key hubs in the waking self-system, which may help explain the focus on personal concerns in dreams. This subsystem of the default network, when uncon- strained and activated, leads dreamers to experience themselves as being in hypo- thetical scenarios that include vivid sensory environments, which also usually portray interpersonal interactions. Dreaming is an intensified and enhanced form of sponta- neous thought that can be characterized as an “embodied simulation.” Keywords: dreaming, imagination, embodied simulation, neurocognitive, default network I first of all thank the late David Foulkes (1935–2019) for all the help he gave me with several of the key ideas while this article was being prepared. In addition, his innovative dream research and numerous laboratory discoveries over the past 57 years, and his early recognition that dreaming is a form of simulation, led the way to a fully cognitive theory of dreams (Foulkes, 1985), and then aided in the development of a neurocognitive theory of dreams (Foulkes & Domhoff, 2014). I thank a pioneering cognitive researcher in the study of both daydreaming and dreaming, John Antrobus, for insightful suggestions that greatly improved the substance of this article, and an innovative REM sleep expert, Jerry Siegel, for corrections and new additions on issues relating to REM sleep. Correspondence concerning this article should be addressed to G. William Domhoff, Department of Psychology, University of California, Santa Cruz, 1156 High Street, Santa Cruz, CA 95064. E-mail: [email protected] 265 Dreaming © 2019 American Psychological Association 2019, Vol. 29, No. 4, 265–302 ISSN: 1053-0797 http://dx.doi.org/10.1037/drm0000119  The article entitled “A New Neurocognitive Theory of Dreams,” which brought together findings on the neural substrate that subserves dreaming, the development of dreaming, and adult dream content in a new synthesis, was first presented at the meetings of the Association for Psychological Science in 2000, first published in a slightly extended version in 2001, and then included in revised form as a chapter in The Scientific Study of Dreams: Neural Networks, Cognitive Development, and Content Analysis (Domhoff, 2000, 2001, 2003). The theory was significantly extended several years later when research demonstrated that “the neural network for dreaming” is very likely a “subsystem” of the default network, which subserves mind-wandering, daydreaming, and other forms of self-generated spontaneous thought during waking (Domhoff, 2011). Dreaming, like imagination and mind-wandering, can be understood as a form of simulation, “a particular kind or subset of thinking that involves imaginatively placing oneself in a hypothetical scenario and exploring possible outcomes” (Schacter, Addis, & Buckner, 2008, p. 42). In addition, dreams can be characterized as “embodied” simulations in the strict psychological sense of the term as off-line cognition that is body based (Margaret Wilson, 2002, pp. 632–635), which developed from pioneering work by experimental cognitive psychologists studying categories, concepts, and language (Barsalou, 1991, 1999; Gibbs, 1994; Rosch & Mervis, 1975; Rosch, Mervis, Gray, Johnson, & Boyes-Braem, 1976). Building on that past work, embodied simulations are distinctive because they are subjectively experienced as the body in action (Bergen, 2012; Gibbs, 2006; Landau, Meier, & Keefer, 2010; Niedenthal, Winkiel- man, Mondillon, & Vermeulen, 2009). The defining feature of dreaming is therefore the sense of being a participant in (or observer of) an ongoing real event, which sometimes is thought of as a real experience for several seconds or minutes after awakening. As a result of the realistic feelings that often accompany dreaming, dreams have parallels with theatrical plays because the dreamer and the other characters are usually engaged in one or another activity or social interaction within a setting, or series of settings, and sometimes express thoughts or emotions. In a study based on a normative nonlab sample containing 991 dream reports from young women and men, 86.9% of the dream reports included a social interaction or shared social activity, 6.7% included the dreamer seeing, hearing, or thinking about another dream character, 2.2% included only the dreamer and at least one animal, and 4.3% included only the dreamer engaging in an activity (Domhoff & Schneider, 2018, p. 10 Table 3, for a summary of the findings). The theory thereby includes all three of the “distinct but interdependent levels” that are the hallmarks of all theorizing in cognitive neuroscience, as outlined in a general statement on what constitutes an adequate theory in cognitive neuroscience (Ochsner & Kosslyn, 2014, p. 2): an underlying neural substrate (which means a portion of the default network in the case of dreams), the cognitive processes it supports (embodied simulation in this case), and the “output” or “behavioral” level (verbal or written dream reports). The 20th anniversary of the first presentation of the theory seems to be a good time for an assessment. This is because “coherent groups in science generally have only comparatively brief life-spans, typically ten to fifteen years, after which they are either absorbed into mainstream science or die out,” or perhaps persist as “a field on the margins of 266 DOMHOFF  legitimacy,” as in the paradigmatic example of parapsychology (McClenon, 1984, p. 2). In addition to assessing the adequacy of the original and updated versions of the theory on the basis of the past 20 years of dream research, this article also compares the assessments of it with the assessments of four other contemporary theories with a neural dimension: the Freudian, activation-synthesis, memory-consolidation, and threat-simulation theories of dreaming. The Original Theory and the Assessments of It The first version of the theory began with a discussion of lesion studies that provide the likely outlines of the neural substrate that supports dreaming: “The new theory starts with findings from neuropsychological assessments of patients suffer- ing brain injuries, which reveal the areas of the brain that are and are not necessary for dreaming to occur”; however, the analysis then quickly added that “these discoveries are supported by neuroimaging and sleep laboratory studies” (Dom- hoff, 2001, p. 14). In addition to the lesion and neuroimaging studies highlighted in the Introduction to the original article, the importance of findings from what is now called electrical brain stimulation were added later in the article in a more detailed discussion of this neural substrate: “Then, too, studies using stereotaxic electrodes to locate the sites causing seizures in epileptic patients show that the ‘dreamy state’ sometimes experienced as part of the diagnostic process is related to the temporal- limbic region” (Domhoff, 2001, p. 24). The neural dimension of the theory is therefore based on three independent lines of research, each of which uses a different methodology. Based on these three different types of studies, it was concluded that, Dreaming depends on the normal functioning of a relatively specific neural network located primarily in the limbic, paralimbic, and associational areas of the forebrain. If there are defects in this network, dreaming can be lost temporarily or permanently, or be impaired in some way, such as loss of visual dream imagery. (Domhoff, 2001, p. 14) Because this substrate can be activated to varying degrees in both REM and non-REM (NREM) sleep, this theory encompasses dreaming in any stage of sleep. In terms of the frequency of dreaming, this generalization primarily means REM sleep and the increasing amount of NREM 2 dreaming in the last 2 hr of sleep (Antrobus, Kondo, Reinsel, & Fein 1995; Domhoff & Schneider, 1999, pp. 149–150; Herman, Ellman, & Roffwarg, 1978; Pivik & Foulkes, 1968; Wamsley, Hirota, Tucker, Smith, & Antrobus, 2007), to the point that there are no significant differences between REM and NREM 2 recall and content at the time of awakening (Cicogna, Natale, Occhionero, & Bosinelli, 1998). Soon thereafter, the theory was expanded to include dreaming at sleep onset (Foulkes, Spear, & Symonds, 1966; Foulkes & Vogel, 1965), and also brief episodes of dreaming during long periods of drifting waking thought, when individual participants are left alone in a room for upward of 30 min, with their waking state monitored by an electroencephalogram (EEG; Domhoff, 2003, pp. 20, 31–32; Foulkes & Fleisher, 1975; Foulkes & Scott, 1973). The article then added a developmental dimension to the theory based on the unexpected and counterintuitive findings on the dreams of children ages 3–15, NEUROCOGNITIVE THEORY OF DREAMS 267  which were discovered in longitudinal and cross-sectional studies in a laboratory setting (Foulkes, 1982, 1999, for a summary and synthesis; Foulkes, Hollifield, Sullivan, Bradley, & Terry, 1990). These findings were included and defended as a crucial building block in a neurocognitive theory of dreaming at a time when they were often still ignored or rejected, usually on the basis of what proved to be unfounded methodological criticisms concerning the comfort level of young children in a sleep-laboratory setting or the inability of the children to adequately convey their dream experiences after an abrupt night awakening (Hunt, 1989; Resnick, Stickgold, Rittenhouse, & Hobson, 1994; Weinstein, Schwartz, & Arkin, 1991). The theory then incorporated the large body of findings on dream content, mostly studied with the Hall and Van de Castle (1966) coding system, leading to the conclusion that the “output” of the neural network for dream generation is “generally continuous with waking conceptions and contains a great deal of previously unrealized repetition in characters, social interactions, misfortunes, negative emotions, and themes” (Domhoff, 2001, p. 14). Based on the evidence that cognitive dream theorists had accumulated to suggest that dreaming did not have any adaptive function, or at least any adaptive function that had been proposed up to that point (Antrobus, 1993; Blagrove, 1992, 1996, 2000; Foulkes, 1985, 1993), the theory claimed that dreams are most likely “the accidental by-product of two great evolutionary adaptations, sleep and consciousness,” but noted that “many societies have invented cultural uses for dreams, usually in conjunction with religious ceremonies and medicinal practices” (Domhoff, 2001, p. 15). As might be expected, the reactions of cognitively oriented dream researchers to the new theory were mostly positive (Foulkes, 2017; Wamsley & Antrobus, 2006). However, a Freudian-oriented research psychologist, who focused on personality studies, found the theory wanting in the level of detail on neuroimaging, “premature” in placing an emphasis on Foulkes’s (1982, 1999) laboratory findings on children in the light of what he considered to be good evidence from home-collected reports and clinical case studies, and further stated that the presentation was too readily dismissive concerning “the effects on dreams of subliminal informational inputs” (Holt, 2004, pp. 405, 408). Still, he did write that the theory builds on “convincing evidence that relatively specific areas of the brain are necessary for dreaming” and added that it is surprising how much can be learned from content analysis (Holt, 2004, p. 404). The most critical reviewer, although agreeing that “forebrain activation is essential to dreaming,” that dream content can be studied scientifically, that the Hall/Van de Castle coding system is “good at catching” dream content, and that dreams “reflect an individual’s personality, concerns, feelings, and conflicts,” also argued that “the theory neglects the voluminous work emerging from both the new discipline of cognitive neuroscience and its traditional counterpart, neuropsychol- ogy,” downplays the role of “REM sleep physiology” in explaining key aspects of dreaming and dream content, and thus overstates the similarities of dreaming and waking thought while ignoring the unusual formal features of dreaming (Hobson, 2003, pp. 188–189). In his view, it also neglects the issue of why so few dreams are recalled and underestimates the frequency of bizarreness and emotions in dream content (Hobson, 2003, pp. 190–191). 268 DOMHOFF  The early versions of the new neurocognitive theory received attention in part due to the favorable treatment of them in a comparison with Freudian theory (Freud, 1900; Solms, 1997, 2000a) and activation-synthesis theory (Hobson, 1988; Hobson & McCarley, 1977). This article, written by coauthors who studied consciousness, with little or no previous involvement in dream studies, concluded that “dream consciousness is remarkably similar to waking consciousness, even allowing for the differences in ‘volition, self-awareness and reflection, affect, and memory . . . ,’” an assertion that runs contrary to the other two theories they discussed; they went on to say that “Converging evidence from multiple fields” suggests that dreaming may be “closely related to imagination, where brain activity presumably flows in a ‘top-down’ manner” (Nir & Tononi, 2010, p. 97). This assessment also included a detailed table, which provided very useful comparisons of the three theories, and made clear how different the cognitive view is from the other two theories (Nir & Tononi, 2010, Table 1, p. 93). Although the original article from 2001 had been cited over 50 times by late 2019 by other authors, and the updated article from 2011 almost as often, most of those mentions were made in passing, and seldom supported, criticized, or made use of the theory. However, two different groups of researchers noted that their neuroimaging findings supported the theory’s claims concerning the neural sub- strate that enables dreaming (Eichenlaub et al., 2014; Fox, Nijeboer, Solomonova, Domhoff, & Christoff, 2013). The theory has not died out, but it has not been “absorbed into mainstream science” either (McClenon, 1984, p. 2). The Fate of Predictions Based on the Theory Drawing on several suggestive findings in earlier research studies, the new neurocognitive theory presented two hypotheses that were later supported, one concerning likely future findings on lucid dreaming, the other concerning the degree and valence of emotionality in dreams. In the case of lucid dreaming, it was suggested that the neural substrate that enables dreaming is very likely augmented when this rare phenomenon occurs: “lucid dreaming may be a product of a dream state in which the higher-order neural patterns that give us ‘core consciousness’ and an ‘autobiographical self’ are more active than usual” (Domhoff, 2001, p. 18). It did so on the basis of a study reporting that “higher levels of alpha activity during REM are related to lucid dream reports” (Tyson, Ogilvie, & Hunt, 1984, p. 442) and a second one (Shapiro et al., 1995), reporting, a greater sense of control in an exploratory PET-scan study of 12 male participants when the medial frontal cortex and rectal orbital gyrus were more active, and a greater sense of things being out of control when the amygdala was most active. (Domhoff, 2001, p. 18) The hypothesis drawn from these two studies receives preliminary support in two separate studies that need to be replicated due to their small sample sizes, which together produced only five possible instances of self-aware dreaming in four out of 10 practiced lucid dreamers during 21 nights in a sleep lab. The first study reported, based on one instance each from three of six student participants, that the results showed a “hybrid state” with “wake-like inter-scalp networking, including high- frequency bands,” which were “most pronounced in frontal and frontolateral NEUROCOGNITIVE THEORY OF DREAMS 269  coherences” (Voss, Holzmann, Tuin, & Hobson, 2009, pp. 1191–1192, 1195, 1196). In the second study, four adult males between ages 27 and 32, who had been training themselves to achieve self-reflectiveness during dreaming for 4 or more years, spent a collective total of 15 nights in which they were scanned with a functional MRI; one of them reported two instances of lucidity, which were accompanied by “a reactivation of several areas normally deactivated during REM sleep,” which were for the most part regions in the frontoparietal control network, such as the frontopolar cortex and the dorsolateral prefrontal cortex (Dresler et al., 2012, p. 1020). Research on patients who had suffered damage to the amygdala (Adolphs, Tranel, & Damasio, 1998; Damasio, 1999) led to the hypothesis that such patients “might be ideal candidates for future defect studies because they have lost their capacity for fear in waking life and express predominantly positive emotions” (Domhoff, 2001, p. 25). More specifically, it was claimed that they would report less emotion and “their negative emotions percent would be far lower than the 80% figure that has been found in several different studies” (Domhoff, 2001, p. 25). These predictions received strong support several years later in a study of 23 Most Recent Dreams from eight patients with damage to the basolateral amygdala (Blake, Terburg, Balchin, van Honk, & Solms, 2019). The original statement of the theory also suggested a tentative hypothesis that did not prove to be fruitful. Although the article warned that “very little progress has been made” in the study of symbolism in dreams, and later added that 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 in waking life to most individuals through a combination of developmental experiences and cultural heritage,” it also said that cognitive linguistics “presents new ideas for studying metaphors in dreams that provide additional starting points” (Domhoff, 2001, p. 26; Lakoff, 1993, 1997). This possibility was soon downplayed and then rejected on the basis of subsequent work (Domhoff, 2003, pp. 33–36, 128–133; 2015, pp. 33–36, 128–133). It now seems likely that the neural substrate that subserves dreaming provides insufficient cognitive capacity to support figurative thought (Domhoff, 2018a), which is a new hypothesis based on recent neuroimaging studies that reveal the brain regions involved in the comprehension and production of metaphors (Beaty, Silvia, & Benedek, 2017; Benedek et al., 2014; Holyoak & Stamenkovi´c, 2018). Mistakes and Shortcomings The original presentation of the theory was not without mistakes and shortcomings in terms of the presentation strategy, one or two substantive issues, and a serious omission that would have widened the scope and possible impact of the theory. First, it seemed sensible at the time to begin the discussion of the neural basis for dreaming with lesion studies, which had played a central role in neuropsychology and cognitive neuroscience, and there was more evidence con- cerning lesions and dreaming (Jus, Jus, Gautier, et al., 1973; Kerr & Foulkes, 1981; Kerr, Foulkes, & Jurkovic, 1978; Solms, 1997) than there was from imaging studies of REM sleep (Braun et al., 1997; Maquet et al., 1996; Nofzinger, Mintun, Wiseman, Kupfer, & Moore, 1997). However, it soon became clear that neuroimaging studies 270 DOMHOFF  were gaining more legitimacy and wider usage, thanks in part to the replacement of PET-scans, which require the use of radioisotopes, with the safer alternative of functional MRI (Andrews-Hanna, 2012, p. 253, Figure 1). Neuroimaging studies of the default network therefore became the starting point for the presentation of the theory when the first major update occurred (Domhoff, 2011). Still, even though lesion studies now play second fiddle, they remain important from a neurocognitive point of view because they provide the crucial first-person testimony as to the presence, alteration, or absence of dreaming, which has led to the useful and solidly established conclusion that only lesions inside the default network have any impact on dreaming (Domhoff, 2018b, Chapter 5). This update also made it possible to state the four specific conditions under which dreaming very likely occurs, namely, (1) an intact and fully mature neural substrate for dreaming . . . (2) an adequate level of cortical activation, which can be provided by the REM mechanism and/or generally higher brain activation at sleep onset and late in the sleep period; (3) an occlusion of external stimuli, most likely through gates in the thalamus; and (4) the loss of conscious self-control, i.e., a shutting down of the prefrontal executive systems that connect us to the external world. . . . (Domhoff, 2011, p. 1172) Second, it proved to be both a strategic and substantive mistake to try to put the posttraumatic stress disorder (PTSD) dreams of PTSD victims, recurrent dreams reported by college students, dreams that a majority of people say via question- naires they have experienced, and repeated themes in individual dream series along one hypothetical “repetition dimension.” This term also had unfortunate and regrettable echoes of Freud’s (1920) claim that a “repetition compulsion” was a key element in psychic life. The phrase was changed to “repetition principle” in 2003 and the idea was abandoned thereafter as a case of overreach. On a strictly substantive level, the first presentation of the theory erred in that it too quickly concluded, based on one neuroimaging study (Maquet et al., 1996) and PTSD dreams (Hartmann, 1984; Kramer, Schoen, & Kinney, 1987), that the amygdala and other parts of the support system for emotions were regularly active during dreaming. Since that time, a meta-analysis comparing neuroimaging studies of mind-wandering and dreaming has shown that the amygdala is generally not active in either of those states (Fox et al., 2013), and a lesion study has shown that the amygdala is not necessary for dreaming to occur (Blake et al., 2019). It is now also known that parts of the amygdala are included in a perception network that detects and interprets social signals from other people and an affiliation network that promotes prosocial behavior and concern for others, as well as an aversion network, so activation of the amygdala per se is not necessarily indicative of emotions (Bickart, Dickerson, & Barrett, 2014, p. 238). Moreover, it is now known that the prefrontal cortex in general (Dixon, Thiruchselvam, Todd, & Christoff, 2017; Lindquist, Wager, Kober, Bliss-Moreau, & Barrett, 2012), and perhaps the dorsolateral prefrontal cortex in particular (Le- Doux, 2012, 2015, 2019), are necessary for experiencing emotions in the waking state. Even more generally, the frontoparietal control, dorsal attentional, default, and salience networks combine in slightly different ways to construct emotion states, with the salience network involved in all of the emotional states (Tourou- toglou, Lindquist, Dickerson, & Barrett, 2015). Because all these networks except NEUROCOGNITIVE THEORY OF DREAMS 271  for the default network are relatively deactivated during both REM and NREM 2 sleep, these new findings may account for the lower levels of emotions in most dream reports than is generally assumed, a replicated finding that is discussed further at later points in this article. The major shortcoming of the article was a failure to include the literature on daydreaming, which developed in tandem with dream research from the 1960s through the 1990s, and always paid attention to and incorporated new findings from sleep-dream laboratories (Antrobus, Antrobus, & Singer, 1964; Antrobus, Singer, Goldstein, & Fortgang, 1970; Singer, 1966, 1975; Singer & McCraven, 1961). A similar strand developed in the extensive work by Eric Klinger (1971, 1990, 1999). It included a large-scale field study using random contact with the participants through the use of a pager, and in the process found 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, p. 124). Still another study, by other investigators, included a 2-week collection of daydreams from high-achieving teenagers; it documented that daydreams relate overwhelmingly to personal concerns, which is consistent with one of the major replicated findings on dream content (Domhoff, 2003; Gold & Reilly, 1985). Then, too, there is a relationship between the degree of bizarreness in daydreams and nonlab dreams in a study of the same participants in both aspects of the study (Kunzendorf, Hartmann, Cohen, & Cutler, 1997). This oversight in the first presentation of the theory occurred even though the promulgator of the neurocognitive theory of dreams had included work on daydreaming in a book published 5 years earlier that brought together the accumulated findings based on the quantitative study of dream content (Domhoff, 1996, pp. 6–7, 211). All of this work on daydreaming fits well with the later findings on the role of the default network in mind-wandering, daydreaming, dreaming, and other forms of self-generated thought (Andrews-Hanna, Smallwood, & Spreng, 2014). New Findings That Support the Theory For all the initial mistakes and shortcomings, the original and updated versions of the theory have been largely supported by the accumulating evidence on all three of the theory’s dimensions. To begin with, numerous new waking neuroimaging studies provide further evidence that can be used to argue that dreaming is supported by a subsystem of the default network (Andrews-Hanna, Irving, Fox, Spreng, & Christoff, 2018; Fox, 2018; Fox et al., 2018; Fox, Foster, Kucyi, Daitch, & Parvizi, 2018; Meyer, 2019; Meyer, Hershfield, Waytz, Mildner, & Tamir, 2019; Zabelina & Andrews-Hanna, 2016). So, too, did a meta-analysis comparing neuroimaging studies of mind-wandering and REM sleep (Fox et al., 2013), and a study of neuroimaging patterns during dreaming (Eichenlaub et al., 2014). A study of the influence of psychedelics on the default network revealed that sensory substrates were augmented and the default network was relatively deactivated, which provides an important corrective to the anecdotal and clinical claim that dreaming has important commonalities with states induced by psychedelics (Fox, Girn, Parro, & Christoff, 2018). 272 DOMHOFF  Two further case studies provided additional evidence that lesions in posterior regions of the default network can lead to the loss of dreaming (Bischof & Bassetti, 2004; Poza & Martí Massó, 2006), and a study of the effects of electrical brain stimulation in the temporal lobe (Vignal, Maillard, McGonigal, & Chauvel, 2007) showed its importance in generating spontaneous dreamlike thought, and thereby lent further support to past studies in this tradition (Bancaud, Brunet-Bourgin, Chauvel, & Halgren, 1994; Penfield & Rasmussen, 1950, pp. 162–181). Most recently, the largest electrical brain stimulation study to date, which focused exclusively on the issue of dreaming, demonstrated that all 77 instances of a dream-like state, or a feeling of dream recall, were evoked by electrical brain stimulation in regions in the temporal lobe, and most frequently in the medial temporal lobe (Curot et al., 2018, pp. 9–10). These results are in turn consistent with waking electrical brain stimulation evidence suggesting that the medial temporal lobe is important in the initiation of any form of spontaneous thought during waking (Fox, 2018, pp. 170, 175). Still other work, based on high-density EEG studies, provides new evidence that the brain patterns are very similar when the EEG results from NREM 2 and REM awakenings are only compared when dreams are also reported (Perogamvros et al., 2017; Siclari et al., 2017). Although this work notes that it provides some evidence that temporal and medial prefrontal areas are sometimes active, especially during REM periods (Perogamvros et al., 2017, p. 1773; Siclari et al., 2017, pp. 873, 875), it nonetheless emphasizes the importance of posterior regions. Therefore, as one of these two new contributions also states (Perogamvros et al., 2017, p. 1773), this work is limited to some extent by the absence of neuroimaging data. Based on the evidence from the neuroimaging, lesion, and electrical brain stimulation studies cited in the previous two paragraphs, this useful new work understates the importance of temporal and frontal areas of the neural substrate that subserves dreaming. This emphasis on neural networks in constructing a neurocognitive theory of dreaming is consistent with the current general focus in the cognitive neurosciences: “Network science, combined with non-invasive functional imaging, has generated unprecedented insights regarding the adult brain’s functional orga- nization, and promises to help elucidate the development of functional architec- tures supporting complex behavior” (Grayson & Fair, 2017, p. 15). There also has been new work that supports and extends the developmental dimension of the theory (Domhoff, 2018b, pp. 175–180, for a summary of the evidence that follows in this paragraph). At the neural level, there is now cross-sectional evidence that the default network does not become somewhat adultlike until ages 9–11 (Fair et al., 2008, 2009; Gordon et al., 2011; Sato et al., 2014; Supekar et al., 2010). These findings are supported in an equally important longitudinal neuroimaging study of children ages 10 and 13 (Sherman et al., 2014). These new findings on the maturation of the default network are comple- mented by new work concerning the frequency of dream reporting and the substance of dream content in preadolescents and adolescents from a sleep lab at the University of Zurich (Strauch, 2004, 2005; Strauch & Lederbogen, 1999). This 5-year longitudinal study of dreaming in 24 children (12 boys, 12 girls) from ages 9–15 supports the earlier finding that dreaming is adultlike in its scope and complexity at ages 9–11; it also shows that emotion comes into dreams gradually and that most of the dream elements present in adult dreams also appear in the NEUROCOGNITIVE THEORY OF DREAMS 273  dreams of preteens and teenagers (Foulkes, 1982; Strauch, 2005, pp. 160–161, 163, 167). In terms of new findings, this study discovered that teenagers dream more frequently of their peers as they mature, are more likely to be criticized or punished by adults, especially men, and are more likely to be aggressors than victims in interacting with their peers, all of which can be understood in terms of the neurocognitive theory of dreams as likely enactments of their view of themselves and their roles in waking life (Strauch, 2005, p. 160). Then, too, the original developmental studies of dreaming are also supported by the accumulating evidence showing that preschool children are largely lacking in the waking cognitive capacities that are also very likely necessary for dreaming: generating mental imagery, organizing experience in a narrative form, imagining past and future scenarios, and developing an autobiographical self (Domhoff, 2018b, pp. 154–162 for a literature review and synthesis; Foulkes, 2017). To begin with, one large-scale study of mental imagery during waking showed that the preschool group did only half as well as the 8-year-olds in generating a simple visual image, and the 8-year-olds did only half as well as the 14-year-olds; in the case of mental rotation tests, the gap became increasingly large as the difficulty of the task increased (Kosslyn, Margolis, Barrett, Goldknopf, & Daly, 1990, p. 1000, Figure 1, p. 1007, Figure 4). These findings are consistent with the mental imagery tests used in conjunction with the cross-sectional dream study of children 5 through 8, which did not detect sufficient capacity to create mental imagery at age 5, and led to the conclusion that “the possibility of kinematic imaging emerges somewhere between 5 and 8 years of age” (Foulkes, Sullivan, Hollifield, & Bradley, 1989, p. 450). Other studies reveal that only half of young children’s waking statements about an event are narratives by age 3, but by age 5 or 6 many children can tell a story that contains a beginning, middle, and end (Reese, 2013, pp. 197–198; Taylor, 2013, p. 803), although this ability is not fully developed until early adolescence (Bauer, 2013, p. 522; Bauer, Burch, Scholin, & Güler, 2007). Similarly, participants in the cross-sectional dream study were able to produce only simple narrative scenes, without chronology or sequence, at ages 5–7, but at age 8 they were able to generate a narrative with continuity in two temporal units, along with evidence of causality (Foulkes et al., 1990, pp. 456, 461). Waking studies suggest that children do not have the ability to engage in “pretend dramatic play” until they are age 4 or 5, even in stimulating preschool environments (Nelson, 2007, p. 170). In the cross-sectional laboratory dream study of children 5 through 8, the ability to produce complex imaginative stories in response to story prompts significantly correlated with the participants’ overall rate of dream recall when age was held constant (Foulkes et al., 1990, p. 458). Finally, preschool children do not seem to have much awareness of a spontaneous inner mental life (Eisbach, 2013, for a review and synthesis), and personal (autobiographical, autonoetic) memories only gradually develop and become organized into an autobiographical self around age 6 (Bauer, 2013, pp. 521–522; Gopnik, 2009, Chapter 5), all of which suggests that children do not have “the basics of autobiographical memory” and “a roughly adult understanding of consciousness” until they are around the age of 6 (Gopnik, 2009, p. 156). Similarly, in dream studies, “self-involvement in dream scenarios reliably appeared only later (age 7�) than a first stage (age 5�) in which simple dream actions were performed by others” (Foulkes, 2017, p. 4). Generally speaking, then, the various 21st century 274 DOMHOFF  waking findings summarized in this and the previous four paragraphs support an assertion in the first version of the neurocognitive theory; “Dreaming is a cognitive achievement that develops gradually over the first 8 or 9 years of life” (Domhoff, 2001, p. 14). Returning to the adult level, the relatively few new relevant studies of dream content have continued to yield results supporting earlier findings that dream reports most frequently include the people of greatest personal concern to the dreamer in waking life, along with the avocations and issues that are of the most interest and concern in waking life (Bulkeley, 2018; Dale, Lortie-Lussier, Wong, & De Koninck, 2016; Domhoff, 2003, Chapter 5). These new findings also support earlier content-analysis findings with adult dream reports showing that there is relatively little emphasis in dreams on a person’s routine daily events, such as school or work, or on the economic and political events that are of concern to many people in waking life (Domhoff, 1996; Foulkes, 1985; Hall, 1951; Hartmann, 2000), which in turn is consistent with the activation of key hubs in the self-system during dreaming. The most important new adult content findings, reported by mathematical psychologists, demonstrate that the social networks in dreams are very similar to those in waking life. They are small-world networks that have the same properties that are also found in studies of brain networks, memory networks, and many aspects of the social and natural worlds, which is evidence that dreams are more lawful than has previously been thought (Han, 2014; Han, Schweickert, Xi, & Viau-Quesnel, 2016). Then, too, the frequency of appearance of familiar characters in dreams is consistent with Zipf’s power law, which is further evidence for the lawfulness of dreams (Schweickert, 2007a, 2007b). All in all, aside from the misplaced emphasis on the limbic system in the original version of the theory, its sins of omission and commission did not prove to be serious, and the theory has continued to be updated on the basis of new evidence. But if and to what extent it is viable in comparison with other contemporary theories is a separate question. Assessments of Other Dream Theories This section discusses the main assumptions, hypotheses, and findings with regard to four other contemporary theories of dreaming—Freudian theory, activation-synthesis theory, memory-consolidation theory, and threat-simulation theory. All of them have been subject to strong criticism by a variety of dream researchers with varying perspectives. Freudian Dream Theory By the 1990s, studies inside and outside of a lab setting had created strong doubts that very many, if any, aspects of Freudian dream theory had any substance, whether the issue concerns the importance of the day residue in initiating the specific contents of a dream, the origins of speech acts in dreams, the pervasiveness of wishes (which are said to be physiological in origin) as the basis for every dream, NEUROCOGNITIVE THEORY OF DREAMS 275  the role of the dream-work in disguising the wishes in dreams, the pervasive role of repression in shaping dreams and causing them to be forgotten, or the functional role assigned to dreams as the guardians of sleep (Domhoff, 2003, pp. 136–143; 2018b, Chapter 7; Fisher & Greenberg, 1977, 1996, for detailed summaries of all past findings related to Freudian dream theory; Goodenough, 1991, for studies on the recall and forgetting of dreams; Loftus, Joslyn, & Polage, 1998; Loftus & Ketcham, 1994, for the experimental evidence leading to the abandonment of the concept of repression by academic research psychologists). The usefulness of free association as a method of understanding the meaning of dreams, which provides the basic foundations for the Freudian theory of dreams, is called into question by the lack of evidence that it leads to the understanding of dream content (Fisher & Greenberg, 1977, p. 66; Foulkes, 1978), which led Foulkes (1996a, p. 617) to conclude that “extensive experience in association gathering” had convinced him “of its inherent arbitrariness.” Moreover, and contrary to Freudian claims that the method is free of any suggestive influence by the psychotherapist, there is experimental evidence that subtle suggestions from an experimenter- therapist can falsely convince many people on the basis of dream interpretations that they were once lost or abandoned as young children (Mazzoni & Loftus, 1998; Mazzoni, Loftus, Seitz, & Lynn, 1999). These findings on the power of suggestion in a therapeutic context take on greater importance when Freud’s (1900, pp. 114–119) several reports of arguments with patients concerning the wishful basis of their dreams are taken into consider- ation. What Freud saw as overcoming the patients’ resistance to therapists’ insights may involve the social psychology of persuasion and self-persuasion. This does not imply that all psychoanalytic sessions are exercises in suggestion, but it does mean that the burden of proof is on Freudian researchers working in a clinical setting to demonstrate that any therapeutic data they use to make claims about dreams are not confounded by suggestion and persuasion. In the wake of the wide-ranging critique of Freudian dream theory that had been fashioned by the late 1990s, the theory has been defended by two researchers on two different grounds. First, one research psychologist (Erdelyi, 1996, 2004) claims that at least some of the results of studies based on subliminal perception, including his own, demonstrate that unconscious processes can be influenced significantly through the presentation of psychodynamically relevant stimuli, in- cluding in the case of dreams. However, numerous carefully controlled experiments led most research psychologists to conclude that subliminal stimuli are limited to small priming effects for one or two words, with no ability to influence concepts (Avneon & Lamy, 2018; Biderman & Mudrik, 2018; Greenwald, Draine, & Abrams, 1996; Kihlstrom, 2004). The main proponent of Freudian dream theory in the 21st century, who recast it as “neuro-psychoanalysis,” bases his claims primarily on findings on the loss or alteration of dreaming in neurological patients (Solms, 1997). Freud’s wish fulfill- ment theory is said to be supported by the fact that dopamine, which he assumes to be the primary neurochemical active during REM sleep, also supports the “appetitive interests” that he believes are akin to Freud’s concept of the libido (Solms, 2000a, 2002; Solms & Turnbull, 2002, pp. 116, 312). However, the highly complex neurochemistry of REM sleep primarily involves cholinergic, glutamater- gic, and GABAergic neurons (Boucetta, Cissé, Mainville, Morales, & Jones, 2014), 276 DOMHOFF  with the role of dopamine still uncertain at best, and perhaps nonexistent (Siegel, 2017, pp. 9–10, 12). Solms (1997, 2000a) also defends the hypothesis that dreams are the guardian of sleep, based on the claim that dreaming may involve the “backward projection” of the impulses arising in the dopaminergic system (located in the basal forebrain) to the inferior parietal lobes and visual association cortex, thereby preserving sleep. Other dream researchers quickly noted there is little or no evidence that such a mechanism is responsible for dreaming (Antrobus, 2000; Doricchi & Violani, 2000), and since then it has been refuted by the evidence that a subsystem of the default network is the neural substrate that enables dreaming (Domhoff, 2011; Eichenlaub et al., 2014; Fox et al., 2013). In addition, Solms (1997) defended the guardian-of-sleep hypothesis based on his finding that patients who reported the cessation of dreaming more often said they had disrupted sleep than the control sample. However, those findings are not impressive in that 51% of the 101 patients with global loss of dreaming indicated that their sleep was not disrupted (Solms, 1997, pp. 164–165; Solms & Turnbull, 2002, p. 214). If dreaming is necessary to preserve sleep, then it might be expected that virtually all patients reporting the complete loss of dreaming would report an inability to sleep at all. Moreover, two studies of over a dozen lobotomized patients in a sleep lab found that they almost never recalled a dream after REM awakenings, but they had normal REM/NREM cycles and reported in the morning that they had slept well, a claim supported by the EEG records of their sleep (Jus, Jus, Gautier, et al., 1973; Jus, Jus, Villeneuve, et al., 1973). At the same time that Solms (2000a, 2000b, 2002; Solms & Turnbull, 2002) defended and amended Freud’s theory at the neural level, he ignored the equally important findings on the cognitive process of dreaming and the content of dreams that are overviewed and referenced in an earlier section of this article, including the laboratory studies on the development of dreaming (Foulkes, 1982; Foulkes et al., 1990). These omissions are significant in terms of assessing Freudian dream theory because these findings raise serious questions about the adequacy of the theory, which lacks a development dimension and does not encompasses the findings showing that most dream content collected in both lab and nonlab settings does not fit well with its claims about dream content (Domhoff, 1996, 2018b, Chapters 1–3; Dorus, Dorus, & Rechtschaffen, 1971; Snyder, 1970). More recently, new challenges for the theory have arisen concerning its emphasis on symbolism (Freud, 1916, Chapter 10). As noted in an earlier section, there is now evidence that suggests the neural substrate that subserves dreaming may not be able to support the cognitive capacities that have been shown to be necessary to comprehend or generate metaphors in waking neuroimaging studies (Beaty et al., 2017; Benedek et al., 2014; Holyoak & Stamenkovi´c, 2018). There is also new evidence suggesting that potentially symbolic elements in dream content are very infrequent (Domhoff, 2003, pp. 33–36, 128–133; 2015). When the shortcomings of Solms’ (2000a, 2000b, 2002; Solms & Turnbull, 2002) claims about the Freudian implications of neurological findings are combined with the lack of systematic attention to developmental and content studies, and with the new evidence suggesting that symbolism may not be an important feature of dreams, it does not bode well for Freudian dream theory finding acceptance outside of the therapeutic community. NEUROCOGNITIVE THEORY OF DREAMS 277  Activation-Synthesis Theory Activation-synthesis theory stresses the brain basis of dreaming, with cognitive processes a secondary issue. It states that dreaming is initiated by putatively random and chaotic firings that arise from the brain stem during REM sleep, and claims that dreaming is a cortical attempt to make sense of these brain stem events (Hobson, 2000; Hobson, Pace-Schott, & Stickgold, 2000a, 2000b). As a result, activation-synthesis theory focuses on seemingly unusual features that are said to be unique to dreaming, such as frequent occurrences of flying under one’s own power; more specifically, the theory claims that “the individual historical compo- nents of dream plot construction” are “diluted” by “chaotic cerebral activation processes” that lead to “visuomotor hallucinations, delusional beliefs, thought impairments, emotional storms, and memory defects” (Hobson & Kahn, 2007, p. 857). The theory originally began by locating the origins of REM periods in giant neurons in the pontine gigantocellular tegmental field (Hobson & McCarley, 1977; Hobson, McCarley, & Wyzinski, 1975; McCarley & Hobson, 1975). It further claimed that the neurochemical initiation of the REM-on system is cholinergic in nature and concluded that neurons in the locus coeruleus are responsible for turning off REM sleep (Hobson, Lydic, & Baghdoyan, 1986, pp. 378–379; Hobson & McCarley, 1977). Although the fact is not often remarked upon in the literature focused exclusively on dreaming and dream content, all of these claims were refuted within the space of a few years by sleep researchers and physiologists that specialize in the study of REM sleep. Experimental lesion studies soon showed that the activation of the giant cells in the pons was not specific to REM sleep; instead, these neurons were related to movement, as indicated by their very high levels of activity in waking, a relationship the activation-synthesis missed because they used a head restraint while making their recordings (Siegel & McGinty, 1977; Siegel, McGinty, & Breedlove, 1977). Nor did the destruction of the entire gigantocellular tegmental region have any effect on REM sleep (Drucker-Colin & Pedraza, 1983; Friedman & Jones, 1984a, 1984b; Jones, 1979; Sakai, Petitjean, & Jouvet, 1976; Sastre, Sakai, & Jouvet, 1981; Vertes, 1977, 1979). Moreover, “histochemical and pharmacological data” showed that the neurochemistry of REM was not cholinergic in nature (Jones, 1986, p. 410). Still another lesion study demonstrated that the neurons in the locus coeruleus do not play any essential role in the cyclic appearance of REM and NREM sleep (Jones, Harper, & Halaris, 1977). The new evidence was succinctly synthesized as follows: The lateral pons is the brain region critical for REM sleep. Medial pontine regions, including the gigantocellular tegmental field neurons, are not critical for REM sleep generation. . . . the neurons whose interaction is critical for REM sleep constitute only a small percentage of the cells within the lateral pontine region. (Siegel & McGinty, 1986, p. 421, italics in the original) Physiologist Barbara Jones (2000, p. 956) later concluded that she did “not know of any physiological evidence that the cortex has no control over the brain stem or over the central activity of dreams,” and added that “corticofugal outputs reach the entire brain stem as well as the spinal cord, influencing the very neurons shown to be critical for the initiation and maintenance of REM sleep in the pontine 278 DOMHOFF  reticular formation.” Subsequent evidence confirms the large degree of forebrain control of the REM generator, especially through the hypothalamus (Luppi, Clément, & Fort, 2013). The highly complex nature of this system was demon- strated in the aforementioned study using new methods of histochemical neuronal identification, which reveal that cholinergic, glutamatergic, and GABAergic types of neurons a...