From Dreams to Hallucinations: Jean Lhermitte’s Contribution to the Study of Peduncular Hallucinosis and the Dissociation of States

In 1917, Henri Claude and Jean Lhermitte (7) published the observation of a 25-year-old man, a submarine mechanic, admitted in 1916 to a military neurology center in Bourges (a city in central France) for deterioration of his general state, insomnia, polyuria with polydipsia, and bitemporal hemianopia. He had a history of syphilitic chancre. Examination of cerebral spinal fluid (CSF) showed unquantified lymphocytosis and albumin at 0.56 g/L. The initial diagnosis was syphilitic meningitis. After clinical fluctuations, “the subject falls into a deep sleep from which it is impossible to wake him. This ‘narcolepsy’ attack, which lasts about five hours, leaves the subject amnesic and astonished when he wakes up.” Thereafter, his state worsened: “The patient is overcome by a confusional delirium in an oneiroid state [“onirism”]. He says his bed is wet from the rain and fog of the sea; he thinks he is in Toulon [a naval base in the south of France] in springtime.” In the following weeks, the patient had new “attacks of narcolepsy,” and he died 6 months after onset, probably from coincidental pulmonary tuberculosis. The autopsy revealed a retrochiasmatic tumor occupying the interpeduncular fossa and compressing the neighboring structures. The tumor thinned the floor of the third ventricle (in particular, the infundibulum and the lamina terminalis), without reaching the pituitary gland. The histological description suggests a craniopharyngioma (Professor Jacques Poirier, personal communication). Based on this observation, the authors described an “infundibular syndrome.” They observed that, “if one does not confuse them with coma, somnolence and sleep, authentic narcoleptic attacks are not exceptional as manifestations of infundibular tumors.” In his 1910 reports (4, 5), Lhermitte had developed a broader conception of narcolepsy as a syndrome with various causes, particularly tumors, alongside idiopathic or essential forms, as described by Jean-Baptiste Gélineau in 1881 (8), which Lhermitte considered rarer.

Lhermitte and de Saint-Martin described two cases of encephalitis lethargica in 1918 under the name of “primitive poliomesocephalitis with narcolepsy,” stressing the sleep disorder that gave the condition its name (9). In the review he later published, Lhermitte (10) noticed that daytime sleepiness was sometimes associated with insomnia and “a certain degree of agitation,” the latter being “associated with a mild and quiet delirium of a confusional and hallucinatory nature,” which the author attributed to “the close relationship between the dreaming associated with confusion and physiological dreaming. . . . In both cases, it is the same sequence of images that projects its disturbing and capricious phantasmagoria onto the sleeping consciousness, the same lack of insight, the same inconsistency, and the same lack of astonishment at the appearance of the most unexpected and disconcerting phenomena.” These images appeared in the evening, were almost exclusively visual, and could, through a series of intermediaries, lead to an acute delusional syndrome. Lhermitte aligned himself with an opinion previously defended in France by Jean Camus (1872–1924), according to which “lesions of the midbrain and the deep gray-matter nuclei can lead to mental disorders independently of cortical lesions,” suggesting the existence of extracortical regulatory centers of the psychic functions (11).

For Lhermitte, the core features of peduncular hallucinosis were visual hallucinations without delusions, a predominance in the evening, sleep disorders, and an anatomical lesion located in the mesodiencephalic area. However, Lhermitte acknowledged that the syndrome was heterogeneous. Besides visual hallucinations, visual illusions and auditory (including verbal) and tactile hallucinations were reported. Insight into the hallucinatory nature of the experience and the emotional impact varied from one patient to another, or even in the same patient. In most cases, visual hallucinations consisted of people, heads, or animals projected onto an unchanged environment. The patient passively attended the scene (e.g., “she thinks she’s at the theatre attending various performances”) (40). In at least one case (23), the hallucinations were dreamlike, fitting the definitions of onirism (oneiroid syndrome). The patient was acting in a complex and bizarre scenario, moving in an unreal setting along with the hallucinated people. Sleep disorders (e.g., insomnia, daytime sleepiness, or lethargy) were part of the core peduncular hallucinosis syndrome, although van Bogaert (17, 18) had shown that these features could be missing.

Anatomically, Lhermitte postulated that the lesion disrupted a regulatory sleep center located in the midbrain tegmentum or the third ventricle floor, possibly in the pons, but he repeatedly ruled out the possibility of the lesion being localized in the lateral geniculate nucleus or the thalamus. However, the patient reported by van Boagert did have a thalamic lesion (of the pulvinar) associated with those of the midbrain. Interestingly, Lhermitte and Julian de Ajuriaguerra, in an article devoted to hallucinations associated with lesions of the visual pathways (46), reported on patients with ophthalmopathy and visual hallucinations resembling those of peduncular hallucinosis in whom a neuropathological study revealed a thalamic lesion. In Lhermitte’s time, anatomical data were rare, and thalamic lesions could be missed. The possible involvement of thalamic nuclei is now recognized. Galetta et al. (12) collected data on 86 patients with peduncular hallucinosis, among whom 69 had a computerized tomography or MRI brain scan; of these 69 patients, 21 (30%) had a thalamic lesion, isolated or associated with other lesions. In another series of 23 published cases of peduncular hallucinosis, selected on stringent criteria including the exclusion of coexisting lesions of the cortex or visual pathways, Boes et al. (47) found that 16 patients (70%) had a thalamic lesion. It must also be emphasized that peduncular hallucinosis is a rare condition. Peduncular hallucinosis was not recorded in a series of 22 infarcts limited to the midbrain (48), and it was not mentioned (although it is not stated whether it was systematically looked for) in several series totaling 102 patients of infarcts of the midbrain, either isolated (49, 50) or associated with other lesions (51).

In summary, it currently seems relevant to apply the name of peduncular hallucinosis to hallucinations predominating in the visual modality and associated with a demonstrated lesion located in the midbrain, pons, or thalamus, the last of these being commonly involved.

The first way to address this question is to compare the phenomenological characteristics of dreams and peduncular hallucinosis. Although dreams can occur in all stages of sleep, “it is clear that rapid eye movement (REM) sleep represents the sleep state most conducive to sensorily vivid, motoric and emotionally salient dreams with the most elaborated narrative structure” (52). Lhermitte was not aware of the existence of REM sleep. REMs in sleep were described in 1953 (53), but REM sleep as a specific sleep stage was characterized by Michel Jouvet under the name “paradoxical sleep” after Lhermitte’s death, in the 1960s (54). In his descriptions of dreams, Lhermitte highlighted the features of REM-associated dreams (20, 55).

Hallucinations associated with neurological conditions, including peduncular hallucinosis, are usually quite different from dreams (Table 2). In most cases of complex visual hallucinations of neurological or ophthalmological origin, the hallucinated features, most often people or animals, are projected onto an unchanged environment and do not result in a complex and bizarre scenario (56). There are, however, hallucinatory states close to dreaming within the framework of what Emmanuel Régis (1855–1918), a Bordeaux psychiatrist, called délire de rêve or délire onirique (oneiroid delirium) (57, 58). For Régis, this state fell within the category of mental confusion (i.e., delirium) and was associated with intoxications or infections. The author emphasized the similarities with dream images, as well as the nocturnal predominance and the association with sleep or daytime sleepiness: “They are active, moving sleepers” (58). This gray zone between hallucinations and dreams had already been emphasized by Charles Lasègue in 1881 in the context of delirium tremens in an article with a self-explanatory title: “Alcoholic delirium is not a delirium but a dream” (59). The term onirism (oneiroid state), specific to French nosography, later freed itself from its original link with mental confusion, with, for example, Raoul Benon: “Onirism is characterized by the development in the waking state of illusions and hallucinations which resemble those observed in the normal dream, and which are not accompanied, in pure cases, by delirium or mental confusion” (60). Among the historical cases of peduncular hallucinosis, the term onirism only applies to the patient described by Lhermitte and Lévy (23, 24); more precisely, to the phenomena immediately following his awakening from intoxication, before the appearance of hallucinations of a completely different type. In this case, the two types of hallucinations probably were of different origin: dream-like hallucinations of toxic origin first, then typical peduncular hallucinosis hallucinations secondary to a focal lesion.

Clearly, then, peduncular hallucinosis is phenomenologically distinct from dreaming, a fact that Lhermitte, probably influenced by his pathophysiological conceptions, was reluctant to accept. He constantly insisted on the similarities between dreams and peduncular hallucinosis, from his first review paper (39) (“Hallucinatory images [of peduncular hallucinosis] possess the same attributes as those of dreams”) until his last article on this topic (61) (“This type of hallucinosis corresponds to a real waking dream”). Even in the posthumous eighth edition of his book about dreams (55), Lhermitte wrote that the images of peduncular hallucinosis “only differ from the images of the physiological dream in that the former are internalized, while the latter are externalized.” At most, Lhermitte referred to “dream fragments,” (34) “dreamlike fragmentation” (45), or images “similar” to those of dreams (3).

The relationship between peduncular hallucinosis and the hallucinations accompanying states of transition between wakefulness and sleep is a more difficult topic. These phenomena were described by Jules Baillarger in 1845 (62), then by Alfred Maury in 1848, who defined them as hypnagogic (when falling asleep) or hypnopompic (when waking up) (63). The very nature of the sensory experience remains debated; notably, the editors of DSM-5 dismissed these phenomena from the realm of hallucinations (64). These phenomena, which for convenience I will call hallucinations, are very common in the general population (about 25%) and even more frequent in younger subjects (65). In primary narcolepsy, these hallucinations are more frequent, more vivid, and have greater emotional repercussions than in healthy subjects. They are also frequently accompanied by sleep paralysis, which makes them more frightening (66). Lhermitte, in his 1927 report (67), emphasized their usually visual character and described them as “images,” or “visions,” differentiating them from hallucinations and dreams. He briefly mentioned that in narcoleptic syndromes, “sometimes, before falling into deep sleep, the patient was subject to auditory or visual phenomena of a hypnagogic nature.” Phenomenologically, the hallucinations of peduncular hallucinosis are closer to those described during sleep–wake transitions than to the dreams associated with REM sleep (Table 2). However, the distinction between “half-sleep visions,” “hallucinosis,” and “authentic hallucinations” was of limited relevance for Lhermitte: “If from the clinical and phenomenological point of view the distinction must be maintained, this no longer holds true from the psycho-physiological point of view” (68).

Recently, it was proposed that in some conditions such as Parkinson’s disease, hallucinations could result from REM dream intrusions into the waking state (69), but this remains controversial (70). It is unknown whether narcoleptic-like sleep-onset REM periods occur in patients with peduncular hallucinosis, which would be the first step in linking hallucinations to REM dream intrusions. The rarity and transient nature of peduncular hallucinosis probably made it impossible to perform polysomnographic studies during the syndrome. However, in two cases, it was possible to show by polysomnography that a REM sleep behavior disorder could be associated with peduncular hallucinosis. The first case involved a hypothetical mechanism (compression of the left subthalamic region by a dilated carotid termination) (71). The second followed a small ischemic lesion of the rostro-dorsal region of the pons (72). The patient reported visual hallucinations and illusions, which persisted in an attenuated form during the years of follow-up. He also developed a REM sleep behavior disorder. This case demonstrates that the same circumscribed pontine lesion can lead to both types of disorders, suggesting an overlap in the mechanisms.

Lhermitte stated that a lesion of a sleep regulating center would allow the “release” of dreams in a waking subject, through mechanisms involving “the entire extent of the brain, and especially the cerebral cortex” (3). This is not the place to elaborate on the complex mechanisms regulating wakefulness, REM and non-REM sleep, and dreams. Briefly, these phenomena depend on the interaction of a number of neurotransmitter systems in the brainstem, hypothalamus, and basal forebrain, which act by modulating the activity of thalamic nuclei and cortical areas (73). Importantly, mechanisms regulating dreams are at least partially distinct from those regulating sleep stages (74). This complexity precludes the possibility of putting forward a mechanistic model linking a lesion of the upper brainstem to the occurrence of transient episodes of peduncular hallucinosis.

One way to approach the neural basis of peduncular hallucinosis without a prior theoretical framework is to identify the regions functionally affected by lesions producing this syndrome. Indeed, the classical clinico-anatomical method has given way to connectivity studies, where the focus is on the circuitry involved in the symptoms and not the location of lesions in individual cases (47, 75). Boes et al. (47) collected 23 cases of peduncular hallucinosis with identifiable causative brain lesions, mapped these lesions onto a standard template brain, and used normative connectome data and overlapping lesion-associated networks to identify regions common to peduncular hallucinosis syndrome. Although the lesions were heterogeneously distributed, 22 of 23 lesions were negatively correlated with extrastriate visual cortex, a region implicated in visual hallucinations, based on prior functional imaging in patients (75, 76). The authors suggested that sites of anticorrelation (i.e., extrastriate visual cortex) predict sites of postlesion hyperactivity (47). Moreover, the peduncular hallucinosis lesions were positively correlated with the lateral geniculate nucleus. Interestingly, using a similar method, Kim et al. (75) showed that lesions causing visual hallucinations were connected to the lateral geniculate nucleus and that all subcortical lesions were again negatively correlated to extrastriate visual cortex. Thus, a plausible final common pathway of peduncular hallucinosis is the activation of extrastriate visual cortex, possibly secondary to an activation of lateral geniculate nuclei. The missing links between the brainstem lesion and visual cortex hyperactivity have been discussed by Müri (77), who summarized two models. The first one derives from the anatomical description in monkeys of a loop linking the area TE of the inferotemporal cortex (an area involved in object, face, and scene perception) to the basal ganglia (78). The proposed loop starts from TE, then projects successively on the tail of the caudate nucleus, the substantia nigra pars reticulata (SNr), the nigral territory of the thalamus (VAmc in Olszewski’s nomenclature), and finally, back to TE. In this model, a brainstem lesion interrupting the excitatory output from the subthalamic nucleus to the SNr would remove the inhibitory input of SNr to VAmc, leading to increased thalamic input to TE. One major limitation of this model is that SNr lesions are not constantly present in cases with peduncular hallucinosis. The second model proposes that a brainstem lesion produces an imbalance in several transmitter systems, including serotoninergic and cholinergic systems (79). This imbalance would impair the control of inputs to the thalamus and modulate the gating and filtering mechanisms of the thalamus to the visual and other sensory cortices. The key thalamic nuclei in this model are the dorsal lateral geniculate nucleus and the lateral pulvinar. Schematically, “lesions of the raphe nuclei may produce disinhibition of the dorsal lateral geniculate nucleus and impair the fidelity of retino-geniculo-cortical transmission” (79). In line with Lhermitte, Manford and Andermann suggested that hallucinations in narcoleptic syndromes and peduncular hallucinosis were likely to “share a similar pathophysiological basis” (79). Finally, the concept that, in various settings, visual hallucinations are linked to aberrant activity within thalamo-cortical networks remains relevant today (80, 81).

Lhermitte explicitly described a dissociation between normal elements of sleep, waking, and dreaming. Throughout his work, he insisted on this concept with regard to peduncular hallucinosis, writing for example, “Pathological processes can carry out real dissociations of the hypnic function and generate somatic sleep in some cases, psychic sleep in others” (39). Lhermitte also used the term “dissociation” to describe the origin of cataplexy: “We are justified in considering cataplexy as the striking image of a dissociation of the function of sleep, a dissociation which results in inhibition, that is to say the deactivation of muscle tone which contrasts with the conservation of the functions of the mind” (3). He also described sleep paralysis under the term of “awakening cataplexy,” a phenomenon that would seem to “border the physiological state”: “Immediately after waking up in the morning, while his consciousness is perfectly awake, the subject, although he has the desire and the will, finds himself paralyzed. . . . This state is accompanied by an impression of diffuse and extremely unpleasant anguish.” Here again, the phenomenon originates with “the dissociation between psychic awakening and somatic awakening” (67). In short, “the mind watches over a sleeping body in cataplexy” (82). Lhermitte later came back (45) to what Kinnier Wilson had called in the meantime “sleep paralysis” in the context of narcolepsy (83, p. 90). He noticed that sleep paralysis could occur not only on waking, but also on falling asleep or in the middle of the night and that it could be associated with various sensory hallucinations, “psychic hallucinations,” and a “feeling of presence.” In peduncular hallucinosis and narcolepsy, “morbid sleep . . . can . . . manifest a positive side which, contrasting with the inhibition of consciousness or motor activity, allows the deployment of a complete or fragmentary dreamlike activity” (45).

Although described and discussed by Lhermitte, the concept of state dissociation was only recently formalized by Mark Mahowald and Carlos Schenck, two prominent sleep disorder specialists (84, 85). Healthy individuals have three states of being: wakefulness, REM sleep, and non-REM sleep. Either physiologically or in pathological circumstances, some features usually related to one state may occur in another state (Figure 1). For instance, sleepwalking results from a dissociation between wakefulness and non-REM sleep. In summary, “none of the changes occurring with sleep are invariably coupled to sleep” (86). The concept of dissociation of states, or state dissociation, is now widely accepted (87). Lhermitte was interested in and wrote about most of these states, including hypnagogic hallucinations, cataplexy, sleep paralysis, confusional arousals, and somnambulism (20, 61). For him, peduncular hallucinosis hallucinations belonged to this category of phenomena. Interestingly, Mahowald and Schenck (85), in line with Lhermitte, mentioned peduncular hallucinosis and other types of hallucinations as examples of dissociation between wakefulness and REM sleep, an interpretation that is still under discussion. Finally, one of the most studied dissociation of states at present, REM behavior disorder, was not identified by Lhermitte or contemporary authors but was first described by Carlos Schenck in 1986 (88).