Dream imagery, rapid eye movement sleep behavior disorder, and hallucinations


Pr Isabelle Arnulf, Pitié-Salpêtrière Hospital, Inserm U975, Pierre and Marie Curie University, Paris 75013, France. Email: isabelle.arnulf@psl.aphp.fr


Rapid eye movement (REM) sleep behavior disorder (RBD) and hypnagogic hallucinations are salient symptoms of abnormal and dissociated REM sleep that are frequently associated in serious neurological diseases. RBD is a strong, independent risk factor for hallucinations in narcolepsy (odds ratio: 4.3) and in Parkinson's disease (odds ratio: 2.7). In Parkinson's disease, RBD also predicts incident hallucinations and psychosis in prospective cohorts. Status dissociatus (a mixture of hallucinations, RBD, and dissociated sleep–wake states) is observed in patients with Guillain-Barré when hallucinating, but also in Lewy bodies dementia, delirium tremens, fatal familial insomnia, and Morvan's chorea. This co-occurrence of RBD and visual hallucinations suggests a common, extensive lesion within REM sleep executive systems.


In normal subjects, dream imagery (with sensory hallucinations) is more salient during N1 and rapid eye movement (REM) sleep than during N2–N3 sleep. The occasional leak of dream imagery into wake at sleep onset or offset is viewed as the mechanism of physiological sleep-related hallucinations. Physiological hallucinations are reported before entry into stage N1, and at sleep offset. Visual hallucinations (VHs) are reported in numerous neurological diseases.1 A non-exhaustive list includes neurodegenerative diseases (Parkinson's disease, Lewy body dementia), probably auto-immune diseases (narcolepsy, Guillain-Barré, Morvan's chorea), prion disease (fatal familial insomnia), diencephalic lesion (peduncular hallucinosis), toxic disease (delirium tremens), migraine (hallucinations are more geometrical than complex however), seizures (micropsia), and Charles Bonnet syndrome (complex hallucinations in subjects with major visual disturbances). They share common features, including the frequent vision of human figures or faces (sometimes torsos without heads or the opposite), animals (real or bizarre), sometimes in miniature, or scenery of outstanding beauty. The images are static or moving, vivid, and often silent, projected over the normal background of the room. The observer is usually immobile, and placed in darkness or poorly lighted environment. A common model of these hallucinations includes three main mechanisms: (i) direct irritation of the sensorial or integrative cortex, via epilepsy; (ii) lesions of peripheral and central visual pathway, causing defective visual input and hallucinations as a cortex release phenomenon, as in Charles Bonnet phenomenon; and (iii) brainstem lesions/dysfunctions of ascending serotonergic and cholinergic pathways, that modify the thalamocortical connections, as in peduncular hallucinosis and narcolepsy.1 The hallucinations would be, in all these cases, released or produced by the visual association cortex, hence their strikingly similar phenomenology. In addition, mental confusion, dementia or altered consciousness (that are frequent in neurological diseases) may promote the hallucinations, decrease the insight into their unreality, and favor their transformations into hallucinations-based delusions and psychoses. The consciousness is transiently altered at sleep onset. Indeed, the thalamic deactivation occurring at sleep onset most often precedes that of the cortex by several minutes in normal subjects, possibly promoting the hypnagogic hallucinations.2 In this direction, hallucinations and REM sleep behavior disorder (RBD) may represent manifestations of sleep–wake state dissociation (specifically admixtures of wakefulness and REM sleep), of close but different origins. We will review here the evidences supporting that hallucinations in several of these diseases are hypnagogic and are associated to RBD and to other REM sleep phenomena (especially status dissociatus).


Narcolepsy, a disorder of excessive sleepiness with dissociated REM-sleep phenomena is associated with frequent and multi-sensorial sleep-related hallucinations (Fig. 1), to the point that patients may experience transient difficulties to distinguish dreaming from reality. Hallucinations also occur in 40% of patients with narcolepsy during wide-awake state (i.e. when fully awake and not only in association with sleep onset/offset or with a drowsy stage),3 hence they may be phasic elements of sleep (mostly but not exclusively of REM sleep) occurring during wake state, exactly as REM-sleep associated muscle atonia may occur both at sleep onset/offset (sleep paralysis) and in wide-awake state (cataplexy). Abnormal ponto-geniculo-occipital wave activity could be the neural substrates for narcoleptic hallucinations. We recently studied the risk factors for hallucinations in 100 consecutive patients with primary narcolepsy. RBD was more frequent in narcoleptics with (37%) than without (11%) hallucinations, a link persisting in multivariate analysis, as only sleep paralysis (odds ratio [OR]: 5.7) and RBD (OR: 4.3) were significantly associated with hallucinations.3 As this is a cross-sectional and not a longitudinal study, whether RBD occurs before, at time of, or after the onset of hallucinations cannot be addressed.

Figure 1.

Hypnogram of a patient with Parkinson's disease presenting visual hallucinations (VHs) during the night. The first VH was an old woman sitting in a corner of the hospital bedroom, the second VH was a little boy, the grand son of her cousin, sitting on her bed and then disappearing under the bed. Each VH lasted 24 s. They timely appeared at REM sleep offset (arrow). The 20 s extract of the sleep recording at time of the second VH is shown in the lower figure, with normal rapid eye movement (REM) sleep (left hand), an REM, then a startle, with the patient awakening from REM sleep (muscle tone is restored, the EEG goes from theta rhythm in REM sleep to alpha rhythm during the VH) then opening the eyes and repeatedly blinking at the VH.


In neurodegenerative diseases, VHs and RBD are two markers of synuclein rather than tau pathology. They are frequent in Parkinson's disease and Lewy body dementia (but not in multiple system atrophy) and infrequent in Alzheimer disease and progressive supranuclear palsy (which are tauopathies). Hallucinations are experienced by a mean 33% of patients with Parkinson's disease, and increase in frequency (up to 75%) with increasing age and longer disease course.4 They are notably difficult to treat, because neuroleptics (even when atypical) worsen parkinsonism. The hallucinations are mostly visual, with silent animals and figures. They are graded from illusions, benign (presence or passage) hallucinations, severe hallucinations, and psychosis. In this last case, the patients have no insight into the unreality of their hallucinations and may interact with them (e.g. speaking to them, inviting them to dinner or trying to shoot them). In addition to older age and longer disease course, several risk factors have been associated to the occurrence of hallucinations: dopaminergic and anticholinergic therapy, cognitive impairment, visual deficit, and sleep disturbance. First, it was noticed that hallucinations developed after a period of vivid, distinctly different dreams, and that psychosis develop after hallucinations, suggesting a kindling phenomenon. Later, two groups found that sleep was more disrupted and shortened in patients with than without hallucinations.5,6 Eventually, we examined the time association between sleep and severe hallucinations.7 In video-polysomnography, VHs timely occurred at the end of night-time and daytime REM sleep (narcolepsy-like) episodes (example in Fig. 2). We hypothesized that VHs corresponded to irruption of REM sleep dreaming into wakefulness. There were 7/10 patients with RBD in the hallucinator group, versus 5/10 patients with RBD in the non-hallucination groups. Eventually, one patient with VHs plus RBD died and had a brain examination, showing numerous Lewy bodies in the subcoeruleus nucleus, but not in the pedunculo-pontine nucleus nor in the cortex. These results supported a brainstem origin for VHs in Parkinson's disease, rather than in a cortical origin (and beginning of Lewy body dementia) as thought before.20 Soon after, it was shown in ambulatory polysomnography at home that 33% of VHs were hypnopompic (emerging from REM sleep but also from N2/N3 sleep), and in 66% of VHs were hypnagogic, occurring in awake or drowsy state prior to N1 brief episodes.21

Figure 2.

Systematic analysis of 12 published studies on the risk of having visual hallucinations when sleep behavior disorder is present in Parkinson's disease. The odd ratios (blues squares) and 95% confidence interval (dotted lines on the x-axis) are displayed in function of the studies (on the y-axis).


There was a higher amount of REM sleep time without atonia (52% vs 13%) in 14 parkinsonian patients with versus 8 patients without hallucinations, suggesting that increased muscle tone and possibly RBD was associated to VHs in Parkinson's disease.22 Following this idea, the association between RBD (a disorder of REM muscle tone) and VHs was examined in 12 cross-sectional series of patients with Parkinson's disease to date, from all countries.8–19 The number of patients per study ranged from 36 to 457, and the definition of RBD ranged from simple dream enacting behavior assessed by interview,8–10,13–17 to polysomnography-defined RBD11,12 and RBD solely defined by movements in REM sleep on the video.18 We calculated the risk for experiencing VHs when RBD was present (OR and 95% confidence interval) in each study in Figure 2. When pooling all data together, the presence of RBD predicted 2.7 more frequent hallucinations. Two longitudinal studies have examined the factors predicting concomitant (present at time of study) and incident (not present at time of study, but appearing later) VHs and psychosis in prospective cohorts of patients with Parkinson's disease. In the US hospital-based cohort of 89 patients with Parkinson's disease followed over 10 years, the rate of hallucinations raised from 33% at baseline to 63% after 10 years, while the rate of enacted dream behavior (clinical RBD, defined by interview in absence of polysomnography) increased from 12% to 33%.16 Clinical RBD predicted twice more frequent concomitant VHs during the whole study, with higher risks for daytime (OR: 5) than for night-time (OR: 2) VHs. Vivid dreaming and disrupted sleep did not predict more concomitant and incident VHs, while clinical RBD did not predict more incident VHs. In the Norway population-based cohort of 230 patients with Parkinson's disease followed over 12 years, clinical RBD predicted a higher risk of a concomitant and incident VHs/psychosis.15 The OR for incident VHs was the highest and the most significant for clinical RBD (3.52), an OR even higher than “classical” risk factors for VHs such as dementia (2.52) or levodopa-equivalent dose of treatment (1.26). All in all, these various results show that clinical RBD predicts more frequent concomitant and incident VHs across the Parkinson's disease course.


In their seminal description of four patients with RBD in 1986, Schenck et al. mentioned a patient with Guillain-Barré syndrome.23 Despite Guillain-Barré syndrome affects mostly the peripheral nervous system, signs of central nervous dysfunction such as sleepiness, hallucinations, abnormal antidiuretic hormone secretion and abnormally low CSF levels of hypocretin are noted.24 In a consecutive series of 139 patients with Guillain-Barré syndrome without any psychotropic and opiates, we found that patients frequently reported vivid dreams (19%), illusions (30%, including an illusory body tilt), hallucinations (60%, mainly visual) and delusions (70%, mostly paranoid).25 In 13 patients monitored during the hallucinatory period, the polysomnography showed a status dissociatus characterized by major insomnia, REM sleep without atonia during 92% of REM sleep time in all hallucinators (RBD could not be assessed as patients were totally paralyzed by the Guillain-Barré syndrome process), REMs during stage N2, sleep onset in REM periods and continuous, rapid switching between wake, N1 and REM sleep stages across day and night. In addition, there were signs of dysautonomia. The status progressively disappeared when hallucinations subsided. In this case, status dissociatus (rather than “isolated” RBD) was timely concomitant of hallucinations.


Peduncular hallucinations are prominent, formed, and colored VHs immediately perceived as unreal by the patient. They are associated with focal lesions of the pons, midbrain or diencephalon. A single patient with peduncular hallucinations (caused by a vascular subthalamus lesion) underwent a polysomnography and functional imaging in symptomatic period.26 The hallucinations occurred in the night-time, mostly in recumbent position, during an awake period with numerous blinking on the EOG. In addition, this patient had RBD. The concomitant PET scan showed a hypermetabolism in the occipital and right opercular area.


Notably, some forms of status dissociatus include continuous dreaming and hallucinations, motor activity and a total loss of wake–sleep boundaries, with patients fluctuating continuously between N1/REM and wake.27 Such status dissociatus is observed in Parkinson's disease, Lewy bodies dementia, alcohol withdrawal syndrome (delirium tremens),28 in a case of frontal dementia of unknown origin,29 in fatal familial insomnia, and in Morvan's chorea. It has been named “agrypnia excitata” by the Bologna group, and sometimes “oneirism” in French groups, but these names recover the same observation of severe VHs and enacted dreams (resembling a continuous RBD, but with open eyes).


It is likely that hallucinations in Parkinson's disease (and possibly in Lewy body dementia) are better explained by state dissociation and intrusion of dreamlike phenomena into wakefulness than by “psychosis” of cortical origin. One of the evidences for this mechanism is the frequent association of hallucinations with RBD. The VHs are strikingly similar among these diseases (mostly colored, silent, brief, moving human, and animals), suggesting a common mechanism. The spectrum of hallucinations however varies, with more passage and presence hallucinations, and less frequent VHs and multimodal hallucinations in patients with Parkinson's disease compared to those with narcolepsy.3 Similarly, although RBD occurs in neurodegenerative and probably autoimmune neurological diseases with VHs, its frequency and severity varies (from less than one-third in narcolepsy to 2/3 in Parkinson's disease), and its phenomenology varies from classical RBD to status dissociatus. The RBD research was boosted when an electrophysiological marker of RBD (namely enhanced muscle tone during REM sleep) was identified. One needs to find an equivalent electrophysiological marker of VHs (such as monitoring the ponto-geniculo-occipital waves, not yet possible in humans) to clarify the research in this fascinating field.


The author declares no conflict of interests.