HM Medical Clinic

 

Dreaming and rem sleep are controlled by different brain mechanisms

BEHAVIORAL AND BRAIN SCIENCES (2000) 23, 793–1121
Printed in the United States of America
Dreaming and REM sleep are controlled by different brain mechanisms Mark Solms
Academic Department of Neurosurgery, St. Bartholomew's and Royal London
School of Medicine, Royal London Hospital, London E1 1BB, United Kingdom Abstract: The paradigmatic assumption that REM sleep is the physiological equivalent of dreaming is in need of fundamental revision.
A mounting body of evidence suggests that dreaming and REM sleep are dissociable states, and that dreaming is controlled by forebrain
mechanisms. Recent neuropsychological, radiological, and pharmacological findings suggest that the cholinergic brain stem mechanisms
that control the REM state can only generate the psychological phenomena of dreaming through the mediation of a second, probably
dopaminergic, forebrain mechanism. The latter mechanism (and thus dreaming itself) can also be activated by a variety of nonREM trig-
gers. Dreaming can be manipulated by dopamine agonists and antagonists with no concomitant change in REM frequency, duration,
and density. Dreaming can also be induced by focal forebrain stimulation and by complex partial (forebrain) seizures during nonREM
sleep, when the involvement of brainstem REM mechanisms is precluded. Likewise, dreaming is obliterated by focal lesions along a spe-
cific (probably dopaminergic) forebrain pathway, and these lesions do not have any appreciable effects on REM frequency, duration, and
density. These findings suggest that the forebrain mechanism in question is the final common path to dreaming and that the brainstem
oscillator that controls the REM state is just one of the many arousal triggers that can activate this forebrain mechanism. The "REM-on"
mechanism (like its various NREM equivalents) therefore stands outside the dream process itself, which is mediated by an independent,
forebrain "dream-on" mechanism.
Keywords: acetylcholine; brainstem; dopamine; dreaming; forebrain; NREM; REM; sleep
REM. The evidence reviewed here suggests also that thesetwo states are controlled by different brain mechanisms.
It is well established that humans spend approximately 25% REM is controlled by cholinergic brainstem mechanisms of sleeping hours in a state of paradoxical cerebral activa- whereas dreaming seems to be controlled by dopaminergic tion, accompanied by bursts of rapid eye movement (REM) forebrain mechanisms. This unexpected dissociation be- and other characteristic physiological changes (Aserinsky & tween REM sleep and dreaming – and the brain mecha- Kleitman 1953; 1955). This state occurs in roughly 90–100 nisms that regulate them – requires a major paradigm shift minute cycles, alternating with four well-defined stages of in sleep and dream science.
quiescent sleep known as non-REM (NREM) sleep (seeRechtschaffen & Kales 1968 for standardized definitions).
In 70–95% of awakenings from the REM state, normal 2. REM sleep is controlled by pontine
subjects report that they have been dreaming, whereas only brain stem mechanisms
5–10% of NREM awakenings produce equivalent reports(Dement & Kleitman 1957a; 1957b; Hobson 1988b).1 The conclusion that Jouvet (1962) drew from his pioneer- These facts underpin the prevalent belief that the REM ing ablation, stimulation, and recording studies – namely state is "the physiological concomitant of the subjective ex- that REM sleep is controlled by pontine brain stem mech- perience of dreaming" (LaBruzza 1978, p. 1537) and that anisms – remains central to all major contemporary mod- dreaming is merely "an epiphenomenon of REM sleep" els of sleep cycle control (for reviews, see Hobson et al.
(Hobson et al. 1998b, p. R12). The discovery of the brain- 1986; 1998b). The reciprocal interaction model of McCar- stem mechanisms that control REM sleep (Jouvet 1962;McCarley & Hobson 1975) has led to the further inferencethat the same mechanisms control dreaming.2 Mark Solms is a Lecturer in Psychology at University This target article presents a body of evidence that sub- College London and Honorary Lecturer in Neuro- stantially contradicts these prevailing assumptions. This ev- surgery at the St. Bartholomew's and Royal London idence demonstrates that, although there is an important School of Medicine. He is author of The neuropsy- link between REM sleep and dreaming, they are in fact chology of dreams: A clinico-anatomical study (1997) doubly dissociable states (Teuber 1955). That is, REM can Lawrence Erlbaum Associates.
occur without dreaming and dreaming can occur without 2000 Cambridge University Press Solms: Dreaming and REM sleep are controlled by different brain mechanisms ley and Hobson (1975) has dominated the field over the neurons. The same applies to motor cortical events in REM past two decades. According to this model, REM sleep – sleep (Hobson 1988b; Hobson & McCarley 1977).
and therefore dreaming – is triggered by cholinoceptive The brain stem localization of the mechanisms that reg- and/or cholinergic "REM-on" cells, and terminated by ulate REM sleep physiology has become a springboard for aminergic (noradrenergic and serotonergic) inhibitory far-reaching inferences about the mechanisms that regulate "REM-off" cells. The REM-on cells are localized princi- dream neuropsychology. An authoritative model of dream pally in the mesopontine tegmentum and the REM-off cells neuropsychology based on brain stem physiology is the in the nucleus locus coeruleus and dorsal raphe nucleus activation-synthesis model (Hobson 1988b; Hobson & Mc- (Fig. 1). Although it is acknowledged that the complete Carley 1977). According to this model, which has domi- network of nuclei contributing to and giving effect to this nated the field for the past two decades, dreams are actively oscillatory mechanism is more widely distributed than ini- generated by the brain stem and passively synthesized by tial findings indicated (Hobson et al. 1986), executive con- the forebrain. The central tenet of this model is that the trol of the REM/NREM cycle is still localized narrowly causal stimuli for dream imagery arise "from the pontine within the pontine brain stem (Hobson et al. 1998b).3 The brain stem and not in cognitive areas of the cerebrum" assertion therefore remains that "cholinergic brainstem (Hobson & McCarley 1977; p. 1347). The dream process is mechanisms cause REM sleep and dreaming" (Hobson seen as having "no primary ideational, volitional, or emo- 1988b, p. 202).
tional content" (p. 1347). Accordingly, the forebrain is as-signed an entirely passive role: Its external input and out-put channels are blockaded by brain stem mechanisms, its 3. REM sleep is not controlled
perceptual and motor engrams are activated by brain stem by forebrain mechanisms
mechanisms, and its memory systems merely generate "thebest possible fit of [this] intrinsically inchoate data" (Hob- An important corollary of the hypothesis that REM sleep – son 1988b, p. 204). In this way it makes "the best of a bad and therefore dreaming – is controlled by pontine brain- job in producing even partially coherent dream imagery stem mechanisms is the hypothesis that it is not controlled from the relatively noisy signals sent up from the brain by forebrain mechanisms. Jouvet (1962) classically demon- stem" (Hobson & McCarley 1977, p. 1347).4 strated that the forebrain is both incapable of generating In the latest, admittedly speculative developments of this REM sleep and unnecessary for the generation of REM model (Hobson 1992; 1994; Hobson et al. 1998b), all the sleep: when cortex is separated from brain stem, it no formal characteristics of dream psychology are accounted longer displays the normal cycle of REM activation (which for by the above-described brainstem mechanisms. Dream is preserved in the isolated brainstem). It is still widely ac- hallucinosis, delusion, disorientation, accentuated affect, cepted that the forebrain is a passive participant in the and amnesia are all attributed to the arrest of brain stem REM state. Even the once-popular notion that the eye aminergic (noradrenergic and serotonergic) modulation movements of REM sleep are attributable to forebrain of brainstem-induced cholinergic activation during REM "scanning" of visual dream imagery has been questioned sleep. It is even suggested that similar chemical mecha- (Pivik et al. 1977). The dominant view seems to be that the nisms may underlie major psychotic symptoms that share eye movements, their associated ponto-geniculo-occipital formal features with dreaming (Hobson 1988b; 1992; 1994; (PGO) waves, and the resultant imagery – in short, all the Hobson & McCarley 1977). However, all of these proposi- visual events of REM sleep – are initiated by brain stem tions are questionable on several grounds.
4. Not all dreaming is correlated with REM sleep
Dreaming and REM sleep are incompletely correlated. Be-tween 5 and 30% of REM awakenings do not elicit dreamreports; and at least 5–10% of NREM awakenings do elicitdream reports that are indistinguishable from REM reports(Hobson 1988b). The precise frequency of NREM dream-ing is controversial. However, the principle that REM canoccur in the absence of dreaming and dreaming in the ab-sence of REM is no longer disputed (Hobson 1988b; 1992;cf. Vogel 1978a).
The original source of controversy was Foulkes's (1962) observation that complex mentation can be elicited inmore than 50% of NREM awakenings (Foulkes 1962).
Subsequent studies have confirmed this observation – andsuggested that an average of 43% of NREM awakeningselicit such reports (Nielsen 1999) – but the extent to whichthe reported mentation may legitimately be described as"dreaming" is still disputed (cf. Cavellero et al. 1992). Thisis due to the fact that there are qualitative differences be-tween NREM and REM dreams: In short, the average Figure 1. The major pontine brain stem nuclei implicated in NREM dream is more "thoughtlike" than the average REM REM/NREM sleep cycle control.
dream. This appears to reaffirm the view that the physio- BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Solms: Dreaming and REM sleep are controlled by different brain mechanism logical state differences between NREM and REM sleep den of evidence for the doctrine has thereby shifted from are reflected in cognitive state differences between NREM the phenomenological link between REM sleep and dream- and REM mentation. However, what is crucial for assessing ing to the anatomical link between the pontine brain stem the validity of the claim that dreaming is generated by the and dreaming.
unique physiology of the REM state is not the questionwhether NREM "dreaming" occurs or not, but rather theextent to which NREM dreaming occurs that is indistin- 5. Dreaming is preserved with pontine
guishable from REM dreaming. This takes account of the brain stem lesions
problem of qualitative differences. It is generally acceptedthat NREM mentation that is indistiguishable from REM The assumption of an isomorphism between REM sleep dreaming does indeed occur. Monroe et al.'s (1965) widely and dreaming was important for the reason that the re- cited study suggests that approximately 10–30% of NREM search program that isolated the brain mechanisms un- dreams are indistinguishable from REM dreams (Recht- derlying REM sleep (ablation, stimulation, and recording schaffen 1973). Even Hobson accepts that 5 –10% of studies) was conducted on infrahuman species in which NREM dream reports are "indistinguishable by any crite- concomitant effects on dreaming could not be monitored.
rion from those obtained from post-REM awakenings" The classical method for establishing brain-mind relation- (Hobson 1988b, p. 143). If we adjust this conservative fig- ships in humans is the method of clinicoanatomical corre- ure to account for the fact that NREM sleep occupies ap- lation in cases with naturally occurring lesions. If the as- proximately 75% of total sleep time, this implies that sumption is correct that dreaming (like REM sleep) is roughly one quarter of all REM-like dreams occur outside controlled by brain stem mechanisms, it should be possible of REM sleep. to demonstrate by this method that brainstem lesions in hu- Moreover, REM-like NREM dreams are not randomly mans eliminate both REM sleep and dreaming.
distributed through the sleep cycle; they cluster around Large lesions of the pontine brainstem eliminate all specific NREM phases. As many as 50–70% of awakenings manifestations of REM sleep in domestic cats ( Jones from sleep onset (descending NREM Stage I) yield reports 1979), and this correlation has been confirmed in 26 hu- that are not significantly different from REM dreams in all man cases with naturally occurring lesions (Adey et al.
respects except for length (Foulkes et al. 1966; Foulkes & 1968; Chase et al. 1968; Cummings & Greenberg 1977; Vogel 1965; Vogel et al. 1972). Also, vivid REM-like reports Feldman 1971; Lavie et al. 1984; Markand & Dyken 1976; are obtained with increasing frequency during the late Osorio & Daroff 1980). However, elimination of REM (or NREM stages, in the rising morning phase of the diurnal near-elimination of REM) due to brainstem lesions was ac- rhythm (Kondo & Antrobus 1989).5 This suggests that these companied by cessation of dreaming in only one of these REM-like dreams are generated by specific NREM mecha- cases (Feldman 1971).6 In the other 25 cases, the investi- nisms. In fact, within the reciprocal-interaction paradigm – gators either could not establish this correlation or they did where wakefulness and REM sleep are seen as terminal not consider it (Adey et al. 1968; Chase et al. 1968; Cum- points on a continuum of aminergic demodulation – sleep mings & Greenberg 1977; Lavie et al. 1984; Markand & onset and the rising morning phase have the opposite phys- iological characteristics to the REM state (Hobson 1992; Although cessation of dreaming has not been demon- strated in cases with elimination of REM due to brain- This is just one strand of the body of evidence that makes stem lesions, the converse is also true: the preservation of it difficult to retain the assumption that dreaming is gener- dreaming in such cases has not been satisfactorily demon- ated by the unique physiological mechanism of the REM strated (Solms [1997a] reported preserved dreaming in four patients with large pontine lesions, but polygraphic In modifying the activation-synthesis model to accom- data was lacking). The paucity of evidence in this respect modate these facts, the claim that all dreams are generated is at least partly due to the fact that pontine brain stem by the brain stem mechanisms that produce the REM state lesions large enough to obliterate REM usually render has recently been abandoned (Hobson 1992). This impor- the patient unconscious (Hobson et al. 1998b).8 More- tant shift in the dominant theory has passed almost unno- over, according to the revised version of the activation- ticed, however, because the closely related claim that all synthesis model (the AIM model), dreaming is generated dreams are generated by pontine brainstem mechanisms by both the REM and NREM components of the sleep- has been retained (Hobson 1992; 1994). In the revised cycle control oscillator (Hobson 1992; 1994). This implies version of the activation-synthesis model (the Activation- that dreaming can only be eliminated by very extensive Input-Mode [AIM] model), both REM and NREM dreams brain stem lesions that obliterate both the REM and the are attributed to reciprocal interactions between aminergic NREM components of the oscillator. Such large lesions and cholinergic brainstem neurons (Hobson 1992; 1994).
are almost certainly incompatible with the preservation The formal characteristics of both REM and NREM men- of consciousness. It is therefore difficult to imagine how tation are therefore still described as "a function of the the assumption that dreaming is controlled by brainstem physiological condition of the reciprocally interacting brain mechanisms can ever be refuted directly by lesion data.
stem neuronal populations that constitute the sleep-cycle It can, however, be refuted indirectly via the corollary control oscillator" (Hobson 1992, p. 228). Thus the doctrine hypothesis that dreaming is not controlled by forebrain of pontine brain stem control of dreaming has been re- mechanisms. That is, the brain stem hypothesis would tained, despite the fact that the assumption upon which it be falsified by clinicoanatomical methods if it could be was explicitly based – the assumption of an isomorphism demonstrated unequivocally that dreaming is eliminated between REM sleep and dreaming (Hobson 1988b; 1992; by forebrain lesions that completely spare the brain Hobson & McCarley 1977) – has been disproved. The bur- BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Solms: Dreaming and REM sleep are controlled by different brain mechanisms 6. Dreaming is eliminated by forebrain lesions
16 cases, the lesion was situated in the white matter sur- which completely spare the brain stem
rounding the frontal horns of the lateral ventricles. In thesecases the damage was invariably bilateral. Of special inter- Subjective loss of dreaming due to a focal forebrain lesion est is the fact that this lesion site coincides exactly with the was first reported more than 100 years ago. Wilbrand (1887; region that was targeted in modified (orbitomesial) pre- 1892) described a patient who dreamed "almost not at all frontal leukotomy (Bradley et al. 1958). This association is anymore" (1887, p. 91) after suffering a bilateral occipital- confirmed by the fact that a 70–90% incidence of complete temporal thrombosis. Müller (1892) documented a similar or nearly complete loss of dreaming was recorded in several patient with bilateral occipital hemorrhages who "had no large series of prefrontal leukotomy (Frank 1946; 1950; Jus further dreams since her illness, whereas previously she not et al. 1973; Partridge 1950; Piehler 1950; Schindler 1953).
infrequently had vivid dreams and saw all sorts of things The many cases included in the latter series increases to in them" (p. 868). Following these classical reports, 108 almost 1,000 the number of reported cases of cessation of further cases with complete (or nearly complete) loss of dreaming caused by focal forebrain lesions.
dreaming in association with focal forebrain lesions havebeen published (Basso et al. 1980; Boyle & Nielsen 1954;Epstein 1979; Epstein & Simmons 1983; Ettlinger et al.
7. Dreaming is actively generated
1957; Farah et al. 1988; Farrell 1969; Gloning & Sternbach by forebrain mechanisms
1953; Grunstein 1924; Habib & Sirigu 1987; Humphrey &Zangwill 1951; Lyman et al. 1938; Michel & Sieroff 1981; It is not surprising that dreaming is lost with lesions in the Moss 1972; Neal 1988; Nielsen 1955; Pena-Casanova et al.
PTO junction – a region that supports various cognitive 1985; Piehler 1950; Ritchie 1959; Solms 1997a; Wapner et processes that are vital for mental imagery (Kosslyn 1994).
al. 1978). This clinicoanatomical correlation between sub- But why should it be lost with lesions in the ventromesial jective loss of dreaming and forebrain lesions has been con- quadrant of the frontal lobes? firmed repeatedly by the REM awakening method (Benson This region contains substantial numbers of fibers con- & Greenberg 1969; Brown 1972; Cathala et al. 1983; Efron necting frontal and limbic structures with dopaminergic 1968; Jus et al. 1973; Kerr et al.1978; Michel & Sieroff 1981; cells in the ventral tegmentum (Fig. 3). These circuits arise Murri et al. 1985) and by morning-recall questionnaires from cell groups situated in the ventral tegmental area of (Arena et al. 1984; Murri et al. 1984; 1985).9 Tsai, where the source cells for the mesolimbic and In short, of the 111 published cases in the human neuro- mesocortical dopamine systems are situated. They ascend logical literature in which focal cerebral lesions caused ces- through the forebrain bundles of the lateral hypothalamus sation or near cessation of dreaming, the lesion was local- via basal forebrain areas (synapsing on many structures ized to the forebrain – and the pontine brain stem was along the way, including nucleus basalis, bed nucleus of the completely spared – in all but one case (Feldman 1971).
stria terminalis, and shell of the nucleus accumbens) and Critically, the REM state was entirely preserved in all of the they terminate in the amygdala, anterior cingulate gyrus, forebrain cases in which the sleep cycle was evaluated (Ben- and frontal cortex. Descending components of this system son & Greenberg 1969; Efron 1968; Jus et al. 1973; Kerr et probably arise from the latter brain areas, and there is rea- al. 1978; Michel & Sieroff 1981). In view of the wide ac- son to believe that they are influenced strongly by cholin- ceptance of the assumption that REM sleep is the physio- ergic circuits (Panksepp 1985).
logical equivalent of dreaming, this lack of clinicoanatomi- This system is thought to have been the primary target of cal evidence correlating loss of REM sleep with loss of modified prefrontal leukotomy (Panksepp 1985). Its cir- dreaming is striking.
cuits instigate goal-seeking behaviors and appetitive inter- The 110 published cases of loss of dreaming due to focal actions with the world (Panksepp 1985; 1998a). It is accord- forebrain pathology fall into two anatomical groups (Fig.
ingly described as the "SEEKING" or "wanting" command 2).10 In 94 cases the lesion was situated in the posteriorconvexity of the hemispheres, in or near the region of theparieto-temporo-occipital (PTO) junction. The lesion wasunilateral in 83 cases (48 left, 35 right) and bilateral in 11cases. This localization has been confirmed repeatedly insubstantial group studies (Arena et al. 1984; Cathala et al.
1983; Murri et al. 1984; 1985; Solms 1997a). In the other Figure 2. Lesion sites associated with loss of dreaming and pre-served REM sleep.
Figure 3. The mescortical/mesolimbic dopamine system.
BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Solms: Dreaming and REM sleep are controlled by different brain mechanism system of the brain (Berridge, in press; Panksepp 1998a). It Solms 1997a; Thomayer 1897). In 22 of the 24 published is considered to be the primary site of action of many stim- cases of this type, the recurring nightmares were caused by ulants (e.g., amphetamine and cocaine; see Role & Kelly epileptiform activity in the temporal lobe, that is, by an un- 1991). The positive symptoms of schizophrenia – some of equivocally forebrain mechanism. (In the other two cases, which can be artificially induced by l-dopa, amphetamines, the nightmares were associated with epileptiform activity in and cocaine intoxication – are widely thought to result from another part of the forebrain: the parietal lobe.) The causal overactivity of this system (Bird 1990; Kandel 1991; Pank- link between the epileptic activity and the recurring night- sepp 1998a). This system is also considered to be the pri- mares in such cases was demonstrated by Penfield and his mary site of action of antipsychotic medications (Role & coworkers (Penfield 1938; Penfield & Erickson 1941; Pen- Kelly 1991). A major psychological effect of antipsychotic field & Rasmussen 1955), who were able to reproduce the therapy is loss of interactive interest in the world (Lehmann same anxious experiences artificially (in the form of waking & Hanrahan 1954; Panksepp 1985). This underpins the "dreamy state" seizures) by stimulating the temporal lobe popular view that antipsychotic medications – which block focus. This causal link between the forebrain seizures and mesocortical-mesolimbic dopaminergic activity – yield the recurring nightmares was confirmed (in Penfield's and "chemical leukotomies" (Breggin 1980; Panksepp 1985).
other cases) by the fact that both the underlying seizure dis- Damage along this system produces disorders character- order and the nightmares responded to anticonvulsant ized by reduced interest, reduced initiative, reduced imag- therapy and/or anterior temporal lobectomy (Boller et al.
ination, and reduced ability to plan ahead (Panksepp 1985).
1975; Epstein 1964; 1967; 1979; Epstein & Ervin 1956; Ep- Lack of initiative or adynamia – where the patient does stein & Freeman 1981; Epstein & Hill 1966; Solms 1997a).
nothing unless instructed (Stuss & Benson 1983) – was a These observations demonstrate conclusively that dream- commonly observed side effect of orbitomesial prefrontal ing can be initiated by forebrain mechanisms (which are leukotomy (Brown 1985).
unrelated to REM sleep) and terminated by forebrain le- The following facts suggest that dreaming is generated by sions (which spare the REM cycle).
this dopamine circuit. First, dreaming ceases completelyfollowing transection of the forebrain component of thiscircuit (Frank 1946; 1950; Gloning & Sternbach 1953; Jus 8. Dreams are generated by a specific
et al. 1973; Partridge 1950; Piehler 1950; Schindler 1953; network of forebrain mechanisms
Solms 1997a). These lesions have no effect on REM sleep.
Transection or chemical inhibition of the same circuit re- In the activation-synthesis model, dream imagery was at- duces the positive symptoms of schizophrenia (Breggin tributed to nonspecific forebrain synthesis of chaotic brain- 1980; Panksepp 1985), some formal features of which have stem impulses. This conception of the neuropsychological long been equated with dreaming (Freud 1900; Hobson mechanisms underlying the formal characteristics of dream 1992; 1988b; Hobson & McCarley 1977). Second, ady- imagery is incompatible with recent clinicoanatomical and namia (a common side effect of the surgical transection of functional imagery findings (Braun et al. 1997; 1998; Solms this circuit) is a typical correlate of loss of dreaming follow- 1997a). Data derived from these two methods have pro- ing deep bifrontal lesions, and it statistically discriminates duced a remarkably consistent picture of the dreaming between dreaming and nondreaming patients with such brain (Hobson et al. 1998b). Both the clinicoanatomical lesions (Solms 1997a). Third, chemical activation of this studies (Solms 1997a) and the functional imagery studies circuit (e.g., through l-dopa) stimulates not only positive (Braun et al. 1997; 1998; Franck et al 1987; Franzini 1992; psychotic symptoms but also excessive, unusually vivid Heiss et al. 1985; Hong et al. 1995; Maquet et al. 1990; dreaming and nightmares (Nausieda et al. 1982; Scharf et 1996; Madsen 1993; Madsen & Vorstrup 1991; Madsen et al. 1978),11 in the absence of any concomitant effect on the al. 1991a; 1991b; Nofzinger et al. 1997) suggest that dream- intensity, duration or frequency of REM sleep (Hartmann ing involves concerted activity in a highly specific group of et al. 1980).12 Fourth, drugs that block activity in this cir- forebrain structures. These structures include anterior and cuit (e.g., haloperidol) inhibit excessive, unusually fre- lateral hypothalamic areas, amygdaloid complex, septal-ven- quent, and vivid dreaming (Sacks 1985; 1990; 1991) and tral striatal areas; and infralimbic, prelimbic, orbitofrontal, other psychotic symptoms.
anterior cingulate, entorhinal, insular, and occipitotempo- These facts suggest that the mesocortical-mesolimbic ral cortical areas (Braun et al. 1997; Maquet et al. 1996; dopamine system plays a causal role in the generation of Nofzinger et al. 1997). Primary visual cortex and dorsolat- dreams. The relationship between this putative dopamin- eral prefrontal cortex are deactivated during REM dream- ergic "dream-on" mechanism and the cholinergic REM-on ing (Braun et al. 1998). The role of the parietal operculum mechanism of the reciprocal interaction model is discussed is uncertain (Heiss et al. 1985; Hong et al. 1995; Maquet et in the final section of this paper.
A further body of evidence strongly supports the view This differentiated pattern of regional activation and in- that dreaming can be initiated by forebrain mechanisms in- activation mirrors some striking neuropsychological disso- dependently of the REM state. It is well established that ciations that have been reported in the clinicoanatomical nocturnal seizures – which typically occur during NREM literature. For example, unimodal abnormalities of visual sleep (Janz 1974; Kellaway & Frost 1983) – can present in dream imagery occur only with lesions in visual association the form of recurring nightmares13 (Boller et al. 1975; cortex (Solms 1997a), but lesions in primary visual cortex Clarke 1915; De Sanctis 1896; Epstein 1964; 1967; 1979; have no effect on dreams. That is, visual dream imagery Epstein & Ervin 1956; Epstein & Freeman 1981; Epstein is intact in cortically blind patients (with V1/V2 lesions) & Hill 1966; Kardiner 1932; Naville & Brantmay 1935; Os- whereas patients with irreminiscence who are unable to tow 1954; Penfield 1938; Penfield & Erickson 1941; Pen- generate facial and color imagery in waking life (due to V4 field & Rasmussen 1955; Rodin et al. 1955; Snyder 1958; lesions) also cannot generate faces or colors in their dreams BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Solms: Dreaming and REM sleep are controlled by different brain mechanisms (Adler 1944; 1950; Botez et al. 1985; Brain 1950; 1954; vation. Lesions in the same region of the left hemisphere Charcot 1883; Grunstein 1924; Kerr et al. 1978; Macrae & convexity also cause cessation of dreaming in association Trolle 1956; Sacks 1985; 1990; 1991; Sacks & Wasserman with disorders of quasi-spatial (symbolic) operations. This 1987; Solms 1997a; Tzavaras 1967). Dream imagery is sim- suggests that quasi-spatial cognition is equally essential for ilarly unaffected by primary cortical lesions in the other dreaming, and that this aspect of dreaming is contributed modalities. Hemiplegic patients (with unilateral periro- by left PTO activation. Lesions in ventromesial occipito- landic lesions) experience normal somatosensory and so- temporal (visual association) cortex cause unimodal deficits matomotor imagery in their dreams (Brown 1972; 1989; of dream imagery, in association with identical deficits of Grünstein 1924; Mach 1906; Solms 1997a). Similarly, apha- waking imagery. This suggests that the visual imagery of sic patients with left perisylvian lesions experience normal dreams is produced by activation during sleep of the same audioverbal and motor speech imagery in their dreams (Ca- structures that generate complex visual imagery in waking thala et al. 1983; Schanfald et al. 1985; Solms 1997a). These perception. It also suggests that these structures are acti- findings suggest that somatosensory, somatomotor, audio- vated in dreams by heteromodal structures that are down- verbal, and motor speech imagery in dreams are generated stream of these unimodal visual processes during waking outside of the respective unimodal cortices for these classes perception. Lesions in other unimodal cortices have no of perceptual and motor imagery (probably in heteromodal effect on dream imagery, notwithstanding their marked ef- paralimbic or PTO cortex). This implies that perceptual fects on waking perceptual and motor functions. This ac- and motor dream imagery does not isomorphically reflect the counts for the predominantly visual quality of dream hal- simple activation of perceptual and motor cortex during lucinosis. It also suggests that the "backward projection" sleep, as was claimed by the authors of the activation-syn- process which presumably generates visual dream imagery thesis model (Hobson 1988b; Hobson & McCarley 1977).
(Kosslyn 1994; Zeki 1993) does not extend further back It also suggests that dream imagery is not generated by than visual association cortex (V3).14 chaotic activation of the forebrain. Rather, it appears that These evidence-based clinicoanatomical inferences (which specific forebrain mechanisms are involved in the genera- tally very closely with the available functional imagery data) tion of dream imagery and that this imagery is actively con- place the neuropsychology of dreaming on an equivalent structed through complex cognitive processes. footing with that of other cognitive functions. This finally In addition, a detailed analysis of the known forebrain paves the way for a testable theory of the brain mechanisms mechanisms implicated in dreaming accounts empirically underlying the complex psychology of dreaming (Solms (Solms 1997a) for the formal characteristics of dreams – such as hallucination, delusion, disorientation, negative af- A noteworthy disparity between the clinicoanatomical fect, attenuated volition, and confabulatory paramnesia – and functional imagery data is the involvement of the pon- which were previously attributed speculatively (Hobson tine brain stem in dreaming sleep in some of the functional 1992; 1994) to the arrest of brain stem aminergic modula- imaging studies (Braun et al. 1997; Maquet et al. 1996) but tion during REM sleep. Lesions in anterior thalamus, basal not the clinicoanatomical studies (Solms 1997a). This dis- forebrain, anterior cingulate, and mesial frontal cortex parity is readily attributable to the fact that dreaming sleep cause excessively vivid and frequent dreaming, a break- was equated with REM sleep in the relevant imaging stud- down of the distinction between dreaming and waking cog- ies, which precluded the possibility of comparing dreaming nition, and other reality-monitoring deficits. This suggests with nondreaming NREM epochs (cf. Heiss et al. 1985).
that the hallucinated, delusional, disoriented, and param- Imaging studies of the dreaming brain at sleep onset, or nestic quality of dream cognition may be associated with during the rising morning phase of the diurnal rhythm inhibition of these structures during sleep. Discharging le- (when the brainstem mechanisms that generate REM are sions in medial and anterior temporal cortex cause recur- uncoupled from the putative forebrain mechanisms that ring nightmares during sleep and unpleasant hallucinatory generate dreaming), would be enlightening on this point.15 experiences during waking life. This suggests that the typi-cal emotional and complex episodic qualities of dreams areproduced through activation of these structures during 9. The relationship between dreaming
sleep. It also suggests that these structures participate and REM sleep reconsidered
causally in the generation of at least some dreams. Bilaterallesions in the ventromesial frontal white matter cause com- The high correlation between the REM state and dreaming plete cessation of dreaming in association with adynamia has traditionally been interpreted as indicating that the and other disorders of volitional interest. This suggests that brain stem mechanisms that generate REM simultaneously these motivational mechanisms are essential for the gener- generate dreaming (i.e., that the REM state is intrinsic to and ation of dreams. Lesions in dorsolateral prefrontal cortex isomorphic with dreaming). However, the data reviewed cause disorders of volitional control, self-monitoring, and above suggest that REM and dreaming are in fact doubly other executive deficits, but they have no effect on dreaming.
dissociable states, in both normal and pathological condi- This suggests that dorsolateral prefrontal cortex is inessential tions, and that they are controlled by different brain mech- for dreaming sleep, which might explain the attenuated voli- anisms. The high correlation between REM and dreaming tion and other executive deficiencies of dream cognition (and therefore requires an alternative explanation.
further account for the defective self-monitoring). Right- Perhaps the most reasonable possibility is suggested by sided lesions in the PTO junction cause complete cessation the observation that the various brain states that correlate of dreaming in association with disorders of spatial cogni- with vivid dream reports all involve cerebral activation dur- tion. This suggests that normal spatial cognition is essential ing sleep. The most common of these is the "paradoxical" for dreaming. It also suggests that the concrete spatial qual- state of REM, in which the brain is simultaneously asleep ity of dreams is supported by right hemispheric PTO acti- and highly activated. Dream reports are also correlated BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Solms: Dreaming and REM sleep are controlled by different brain mechanism with specific NREM states: descending Stage I (sleep on- were based has therefore lapsed. Progress in this area will set) and the rising morning phase of the diurnal rhythm.
now be hampered if we do not acknowledge our initial er- These states are situated at polar ends of the sleep cycle, in ror, and resist the temptation to compress our expanding the transitional phases between sleep and waking. The cor- knowledge of the dreaming forebrain into the initial REM- relations between these states and dreaming have accord- based theoretical framework.
ingly been interpreted as cerebral activation effects (Antro-bus 1991; Hobson 1992). The same interpretation has been applied to the inverse correlation that exists between depth 1. Reported dream recall rates vary, depending not only on the
of NREM sleep (as measured by the sensory arousal thresh- method of awakening and interview but also on the investigator's old) and dreamlike mentation (Zimmerman 1970). Another definition of "dreaming" (Foulkes 1966). The figures cited hereare conservative (they are discussed in more detail in sect. 4).
state which triggers NREM dreaming is complex partial There is no generally accepted definition of dreaming. For our seizure activity, which could be described as a pathological purposes, dreaming may be defined as the subjective experience of form of cerebral activation during sleep. The fact that a complex hallucinatory episode during sleep. However, what is dreaming can be artificially generated by the administra- more important than an absolute definition of dreaming in the tion of a variety of stimulant drugs, including both cholin- present context is the relative frequency with which dream reports ergic16 and dopaminergic agents, is open to a similar inter- obtained from REM and NREM sleep are considered indistin- pretation. Of crucial theoretical importance is the fact that guishable by blind raters.
dopaminergic agents increase the frequency, vivacity, and 2. Control in this context implies activate, generate, sustain,
duration of dreaming without similarly affecting the fre- and terminate.
quency, intensity and duration of REM sleep (Hartmann 3. The concept of "executive control" (Hobson & McCarley
1977, p. 1338; Hobson et al. 1998b, p. R7) implies that the distrib- et al. 1980). This observation, together with the equally uted network of structures that contribute to and give effect to the important fact that damage to ventromesial frontal fibres various physiological manifestations of the REM state are recruited obliterates dreaming but spares the REM cycle (Jus et al.
and coordinated by a cholinergic/aminergic oscillator that is "cen- 1973), suggests a specific dopaminergic dream-on mecha- tered" in the mesopontine tegmentum (Hobson 1988b, p. 185).
nism that is dissociable from the cholinergic REM-on mech- Accordingly, Hobson proposes that "the on-off switch is the reciprocal-interacting neuronal populations comprising the am- These observations show that dreaming is not an intrin- inergic neurons and the reticular neurons of the brain stem" (p. 205).
sic function of REM sleep (or the brain stem mechanisms 4. "If we assume that the physiological substrate of conscious-
that control it). Rather, dreaming appears to be a conse- ness is in the forebrain, these facts completely eliminate any pos- quence of various forms of cerebral activation during sleep.
sible contribution of ideas (of their neural substrate) to the pri-mary driving force of the dream process" (Hobson & McCarley This implies a two-stage process, involving (1) cerebral ac- 1977, p. 1338).
tivation during sleep and (2) dreaming. The first stage can 5. These dreams are difficult to distinguish from REM dreams.
take various forms, none of which is specific to dreaming it- The following are illustrative examples. The first is a sleep-onset self, since reliable dissociations can be demonstrated be- dream (descending Stage I): tween dreaming and all of these states (including REM).
The second stage (dreaming itself) occurs only if and when [It] had something to do with a garden plot, and I was planting seed in it.
the initial activation stage engages the dopaminergic cir- I could see some guy standing in this field, and it was kind of filled and cul- cuits of the ventromesial forebrain. It is reasonable to hy- tivated, and he was talking about this to me. I can't quite remember what pothesize on this basis that these forebrain circuits are the it was he did say, it seems to me as if it had to do with growing, whether final common path leading from various forms of cerebral these things were going to grow (Foulkes 1966, pp. 129–30).
activation during sleep (both REM and NREM) to dream- The second example is a later NREM dream (25 minutes after the ing per se. In this view, the high correlation between last REM episode): dreaming and the REM state merely reflects the fact that itis a regular and persistent source of cerebral activation dur- I was with my mother in a public library. I wanted her to steal something ing sleep. It is also possible that specific aspects of the REM for me. I've got to try and remember what it was, because it was something state (e.g., noradrenergic and serotonergic demodulation) extraordinary, something like a buffalo head that was in this museum. I had facilitate the primary dopaminergic effects. However, such told my mother previously that I wanted this head and she said, all right, facilitatory factors, which vary across the different sleep you know, we'll see what we can do about it. And she met me in the library, states associated with dreaming are not intrinsic to the part of which was a museum. And I remember telling my mother to please dream process itself.
lower her voice and she insisted on talking even more loudly. And I said,if you don't, of course, you'll never be able to take the buffalo head. Every- The biological function of dreaming remains unknown.
one will turn around and look at you. Well, when we got to the place where This is at least partly attributable to the fact that the func- the buffalo head was, it was surrounded by other strange things. There was tion of dreaming and the (equally unknown) function of a little sort of smock that little boys used to wear at the beginning of the REM sleep have been conflated for more than 40 years of century. And one of the women who worked at the library came up to me research. Future studies of these functions should be un- and said, dear, I haven't been able to sell this smock. And I remember say- coupled from one another. The statistical correlation be- ing to her, well, why don't you wear it then? For some reason or other I tween dreaming and REM sleep led early investigators to had to leave my mother alone, and she had to continue with the buffalo the understandable conclusion that they shared a single head project all by herself. Then I left the library and went outside, and underlying mechanism. Subsequent research has demon- there were groups of people just sitting on the grass listening to music strated that this conclusion was erroneous: Dreaming and (Foulkes 1996, pp. 110–11).
REM sleep are in fact doubly dissociable states, they have 6. This was a case of closed head injury with traumatic occlu-
different physiological mechanisms, and in all likelihood sion of the basilar artery. Autopsy and relevant radiological data they serve different functional purposes. The premise upon were lacking. The distinct possibility of forebrain damage in this which the prevailing neuroscientific theories of dreaming case cannot be excluded.
BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6 Solms: Dreaming and REM sleep are controlled by different brain mechanisms 7. In one report (Osorio & Daroff 1980) two patients recalled
hallucinations in awake subjects (Perry & Perry 1995). These ef- no dreams when awoken during atypical NREM epochs; this is not fects are enhanced by eye closure. Therefore, if the REM state is unexpected and does not constitute evidence of loss of dreaming.
indeed partly mediated by basal forebrain cholinergic mechanisms, 8. However, this is not always the case. At least eight patients
as has recently been suggested by proponents of the reciprocal- with cessation or near-cessation of REM have been reported who interaction hypothesis (Hobson et al. 1998b), then something else were sufficiently conscious to communicate meaningfully with an must be added to the cholinergic activation in order to account for examiner (Feldman 1971; Lavie et al. 1984; Markand & Dyken the occurence and formal characteristics of dreamlike mentation 1976; Osorio & Daroff 1979).
during this state. What is proposed here is that this "something 9. The possibility that the reported loss of dreaming in these
else" is provided by the putative dopaminergic mechanism dis- patients is attributable to amnesia for dreams rather than true loss cussed above, the stimulation of which correlates positively with of dreams has been excluded not only by REM awakening but the generation of complex hallucinations, delusions, and other also by neuropsychological examination of memory functions in dreaming versus nondreaming patients (Solms 1997a).
12. In view of the importance of these findings in the present
10. This analysis excludes the "several" cases of cessation of
context, Hartmann et al.'s (1980) study is briefly summarized here: dreaming after cerebral commissurotomy reported by Bogen 13 subjects slept in the laboratory on four occasions each. They (1969), whose findings have never been replicated (Greenwood et were awakened at the end of the first and second REM periods al. 1977; Hoppe 1977).
and either l-dopa (500 mg) or placebo were administered, so that 11. Excessive, unusually frequent, and vivid dreaming (of the
the action of the l-dopa would coincide with the third REM pe- type stimulated by dopamine agonists) has also been described in riod. A study lasting 52 nights yielded 128 dreams, of which 90 association with lesions of the anterior cingulate gyrus, basal fore- were postmedication (42 l-dopa and 48 placebo). Each dream was brain nuclei and closely related structures (Gallassi et al. 1992; scored by four blind raters on five dream content scales: dream- Gloning & Sternbach 1953; Lugaresi et al. 1986; Morris 1992; Sacks likeness, nightmarelikeness, vividness, emotionality, and detail.
1995; Solms 1997a; Whitty & Lewin 1957). Similar phenomena The l-dopa condition dreams were significantly more dreamlike have been linked with central visual deafferentation (Brown 1972; (p , 0.01), vivid (p , 0.01), detailed (p , 0.01), and emotional 1989; Grünstein 1924; Hécean & Albert 1978; Solms 1997a). In (p , 0.05; t-test for correlated samples) than the placebo condi- some of these cases, dreaming occurs continuously throughout tion dreams. The two treatment conditions did not differ signifi- sleep (Gallassi et al. 1992; Gloning & Sternbach 1953; Lugaresi et cantly on any polygraphic measures, including REM frequency, al. 1986; Morris et al. 1992; Sacks 1995; Solms 1997a; Whitty & duration, and density.
Lewin 1957). These patients are unable to distinguish between 13. These are subjective experiences of complex hallucinatory
dreams and real experiences, and reality monitoring in general is episodes, not night terrors. Here is an example: disturbed (Solms 1997a). Most striking are cases in which wakingthoughts spontaneously transform into complex hallucinatory the patient [35 year old woman with idiopathic complex-partial seizures] experiences, resulting in confabulatory delusional states (Solms reported a recurrent dream about her [dead] brother . . which has reap- 1997a; Whitty & Lewin 1957). This disorder has been interpreted peared several times. The dream is as follows: "I am walking down the (Solms 1997a) as indicating that basal forebrain nuclei and closely street. I meet him. He is with a group of people whom I know now. I feel related structures – which are known to participate in discrimina- that I will be so happy to see him. I say to him, ‘I'm glad you're alive,' but tive cognitive processes – play a critical role in distinguishing he'll deny that he is my brother and he'll say so, and I'll wake up crying and between thoughts and perceptions (i.e., inhibiting hallucinosis).
trying to convince him." (Epstein & Ervin 1956, p. 45) Accordingly, damage to these mechanisms results in excessivedreaming during sleep (when the visual system is deafferented) and Electroencephalography revealed a poorly defined right anterior the intrusion of dreamlike mentation into waking thought.
temporal/right temporal spike focus, which appeared with the on- It is reasonable to assume that the normal alternations between set of drowsiness and light sleep.
thoughtlike and dreamlike mentation that occur throughout the 14. This backward projection mechanism is apparently medi-
sleep cycle are somehow related to these (largely cholinergic) ated in part by the cholinergic basal forebrain mechanism dis- forebrain mechanisms. However, they appear to exert this influ- cussed previously.
ence in the opposite direction to that predicted by the activation- 15. The uncertain role of the parietal operculum in REM and
synthesis hypothesis. The fact that damage to cholinergic fore- NREM dreaming also awaits further investigation, but this ques- brain structures (i.e., reduction in cortical acetylcholine) produces tion is unrelated to the main topic of the present paper.
excessive dreaming and dreamlike mentation is consistent with the 16. Interesting to note, if cholinergic agents are administered
widely held view that cortical acetylcholine enhances discrimina- prior to sleep onset they cause insomnia, if they are administered tive cognitive mechanisms (Perry & Perry 1995). Likewise, it is during NREM sleep they induce REM, and if they are adminis- well known that anticholinergic agents (e.g., scopolamine or at- tered during REM they provoke awakening (Sitaram et al. 1978b; ropine), acting on the muscarinic receptors which predominate in Sitaram et al. 1976). This suggests a nonspecific activation- the basal forebrain, produce dreamlike mentation and complex arousal effect.
BEHAVIORAL AND BRAIN SCIENCES (2000) 23:6

Source: http://ftp.bstu.by/ai/To-dom/My_research/Papers-2.1-done/Cognitive-S/Consciousness/For-icnnai-2010/2/Ref/download.pdf