Psychogenic, Nonepileptic Seizures Associated with Video-EEG–Verified Sleep


Address correspondence and reprint requests to Dr. O. Devinsky at Rivergate, 4th Floor, NYU Medical Center, Comprehensive Epilepsy Center, 560 First Avenue, New York, NY 10016, U.S.A. E-mail:


Summary:  Purpose: Nonepileptic seizures (NES) are expressions of a psychiatric disease state, usually conversion disorder, that mimic epileptic seizures (ES) but are not associated with the neurophysiologic changes of epilepsy. Conversion has not been demonstrated to emerge from the sleeping state. Emergence out of sleep is usually considered a virtual exclusion criterion for NES, signifying the presence instead of ES. We sought to test this hypothesis.

Methods: We retrospectively reviewed the video-EEG of all patients admitted to our epilepsy unit over a 3-year period, who were suspected of manifesting NES. We examined the relation between NES and the patients' sleep/wake state in this cohort. Two epileptologists blindly reviewed an intermixture of cases suspected to represent NES emerging out of sleep, together with control cases. Classification of each case was made independently.

Results: We found that in a small minority of patients (<1%), NES began either while the patient was sleeping, or within several seconds of arousal, well before a plan to simulate a seizure could likely have been formulated in the wakeful state.

Conclusions: In some cases, NES are not the product of the awake mind, but rather represent a psychiatric condition that can be manifest in sleep.

Nonepileptic seizures are paroxysmal behavioral alterations that mimic epileptic seizures. NES episodes are not accompanied by any pathologic EEG changes and represent psychogenic phenomena. However, NES patients are not viewed as malingerers, who consciously feign illness for some secondary gain, but are rather generally thought to have one of two psychiatric illnesses: conversion disorder or dissociative states (1). NES episodes may consist of motor behaviors or of internal sensory or cognitive changes. They can mimic either generalized or partial seizures, and the patient can either remain communicative or become unresponsive.

NES are not evenly distributed throughout the circadian rhythm. Psychogenic symptoms are assumed to occur only during wakefulness, and we found no prior studies documenting NES or other conversion symptoms in sleep. With the possible exception of nocturnal panic attacks (as discussed later), we are not aware of prior reports of any nonconscious psychiatric disease explicitly documented to be manifest in sleep. Nevertheless, assuming that it is conscious volition that is inaccessible during sleep and that other psychological functions continue to operate, if NES patients are not conscious malingerers, episodes might, at least in principle, occur during sleep. The few studies that have objectively examined this question have not documented NES during sleep; conversely, when NES seem to occur during sleep, these studies suggest that the patient is actually awake, with eyes closed (i.e., “pseudosleep”) (2,3). In contrast, ES frequently occur during sleep. Historically, seizures occurring during true sleep have been interpreted as ES as a sleep disorder [e.g., rapid-eye-movement (REM) behavior disorder].

We report several patients with NES that began during EEG-verified sleep and several additional patients with NES that began within several seconds of arousal. We thus demonstrate that, at least in some cases, NES are not under the control of the patient's conscious mind.

NES are common, with up to 10% of epilepsy patients having some NES episodes; conversely, 10–40% of NES patients have epilepsy (1,4). In a recent study of 387 consecutive admissions to a video-EEG (v-EEG) unit (5), 64% of the patients had only ES, 25% had NES, and 10% had neither. Diagnosing NES is important, because psychiatric therapy hinges on the patient's awareness of the problem. Further, NES patients frequently appear in the emergency department with intractable seizures, leading to aggressive pharmacologic management, and occasionally to intubation and intensive care unit (ICU) admission, with attendant potential complications including pneumothorax, infections, hemorrhage, and even death.

Several clinical features help distinguish NES from ES (2,5–11). However, the clinical distinction may at times be impossible, requiring v-EEG recording to capture and correctly diagnose episodes. V-EEG also allows rigorously defined electrophysiologic determination of the patient's sleep–wake state (12).


One of us (D.O.) retrospectively reviewed all v-EEG studies between 1998 and 2000 and identified 76 patients diagnosed with NES. All patients in whom the diagnosis of NES versus surface EEG–negative ES was uncertain were excluded, as were all purely subjective episodes. Only episodes with motor changes observable on video (44 of 76 patients, 33 women and 11 men) were included. From this group, only events occurring when the patient appeared to be asleep were selected. Eight hundred six NES events in the 44 patients were screened; the patient was clearly awake in 688, and behaviorally appeared to be sleeping in 118 episodes that occurred in 27 patients.

The 118 episodes of interest were then reviewed by the investigator (D.O.) to determine whether any of the following exclusion criteria were met: (a) whether the EEG showed any change consistent with ES, (b) whether the EEG before and during the episode was consistent with arousal rather than sleep (non-REM or REM), and (c) whether the clinical episode was consistent with an explanation other than NES (as elaborated in Discussion).

Events that were not excluded by these criteria (14 of 118 events, in eight patients) were then presented independently and in a blinded fashion to two epileptologists, board certified in clinical neurophysiology (A.R. and O.D.), interspersed with 15 episodes that were excluded by one or more criteria, used as controls. The two epileptologists agreed in their classification of each of the 14 events as either fulfilling all criteria for an episode of NES in sleep (eight of 14) or as being indeterminate (six of 14). Similarly, there was perfect concordance between the epileptologists in classifying all control events (which had been taken from the same cohort of NES patients) as having failed to fulfill selection criteria.


Of the 118 events, 104 had posterior-dominant (PD) alpha EEG, consistent with arousal. Thus most NES patients who appeared to be asleep during their episodes were actually awake. However, in eight events in five patients, the NES episode either began during EEG sleep, without intervening alpha rhythm (six episodes), or occurred within 7 s after the onset of alpha rhythm (i.e., NES was the first behavioral manifestation of arousal; two episodes). In all cases, the patient was in stage II or III sleep at event onset or several seconds before onset. See Fig. 1 for an example of the EEG tracings in relation to the onset of the NES episode. Table 1 describes the motor manifestations of each of the events, the manifestations of each patient's NES episodes while clearly awake, and characteristics of the interictal EEG for each patient.

Figure 1.

A nonepileptic seizure episode (with initial motoric manifestation indicated by the stippled arrow) emerging from sleep. Note the EEG sleep architecture preceding the episode, in contrast to the muscle-activity artifact that dominates the EEG after the episode begins. There is no intervening epoch of wakeful EEG.

Table 1. 
Total no. of
episodes from
sleep/no. from
Motor manifestations
of NES episodes
  • a

     Wakefulness: symmetric, well-organized background, reactive to eye opening and closure.

  • PD rhythm, as listed above. No generalized background or focal slowing; drowsiness, attenuation and fragmentation of the posterior rhythm with an increase in the low- to medium-amplitude theta components centrally; stage II non-REM sleep, central, symmetric vertex sharp transients, K-complexes, and sleep spindles at 12–14 Hz; stages III and IV non-REM sleep, diffuse medium- to high-amplitude semirhythmic delta activity intermixed with low-amplitude theta and beta components. No epileptiform activity.

12/25Right-to-left and up–down uncoordinated swinging movements of the trunk and torso, and affirmative head bobbing. Movement frequency and episode duration vary from seconds up to 3 min. Episodes occurring out of sleep indistinguishable from those out of wakefulnessaSee below. PD = 11–12 Hz
22/18Shaking movements of either the right hand or of both hands, with arms fully extended forward. Varying frequency of movements and episode duration, from seconds to >2 min. Episodes occurring out of sleep indistinguishable from those out of wakefulnessaSee below. PD = 10.5–11.5 Hz
32/11Vigorous bilateral upper extremity shaking, with nonstereotypic amplitude, frequency, and duration, lasting from 15 to 110 s. Episodes occurring out of sleep indistinguishable from those out of wakefulnessaSee below. PD = 9–10 Hz
41/5Sustained arching of the back, lasting from 25 to 55 s, followed by combativeness and unresponsiveness lasting from seconds up to several minutes. Occasional repeated episodes of back arching during the period of unresponsiveness. Episodes occurring out of sleep indistinguishable from those out of wakefulnessaSee below. PD = 8.5–9.5 Hz
51/4Jerking and shaking movements of various limbs (single or multiple). Movements are uncoordinated, and episode duration lasts from 20 s to 5 min. Episodes occurring out of sleep indistinguishable from those out of wakefulnessaSee below. PD = 9.5–10.5 Hz

In each case, within seconds of the onset of the event, the patient's EEG showed that he or she had wakened, likely because of the movements associated with the NES episode. In almost all cases, the nurse entered the room within 1 min, turned on the lights, and attempted to communicate with the patient. In seven of the eight episodes, the patients continued manifesting the same NES behavior for at least several seconds with no disruption, whereas in one case, the episode was terminated instantly as the patient interacted with the nurse. In the one case in which the nurse did not enter the room, the patient's EEG transformed from stage II sleep to alpha, followed within 6 s by an NES episode that lasted for 53 s, after which he fell back asleep, without any other manifestation of arousal.

In two cases, the NES episode emerged within 7–10 s after a moderately loud background noise. The patients' EEGs showed that they continued sleeping after the noise until the NES episode was under way, but the NES may nevertheless represent a type of startle response to the noise. Six of the 118 events, occurring in three patients, were associated with indeterminate EEG, probably indicating drowsiness, and were excluded.

The five patients with NES in sleep or immediately on arousal (three men and two women) had an average age of 33.8 years (range, 21–47 years). Their psychiatric comorbidities did not notably differ from those of the rest of the cohort. There was no documented sleep disorder among the patients, and their baseline EEGs were consistent with normal arousal and sleep architecture. Only one of the five patients had rare interictal epileptiform EEG activity and a history of ES. However, his seizures, dating back to childhood, were all tonic–clonic, and bore no similarity to the recorded NES episodes.


NES are conversion symptoms or, less often, dissociative states (for the nonmotor, unresponsive subtype of NES, which we excluded from consideration in this study). DSM IV-R (13) classifies NES as a type of conversion disorder, “with seizures or convulsions” (code 300.11), and the actual clinical manifestations are interpreted as a symbolic communication of the patient's psychological stress (1). Conversion disorder is frequently associated with severe prior emotional trauma (often sexual abuse); this has been reported in 25–43% of NES patients (8,14). There is a female-to-male ratio of ∼3:1 in NESs and other conversion disorder patients, as we also found in our 76 patients. Other psychiatric comorbidity in NES patients is very frequent (5). Alper et al. (15) suggested that patients with “nonconversion nonepileptic seizures,” in whom the most common psychiatric diagnoses are anxiety disorders, psychotic disorders, and impulse-control problems, represent a subgroup without female predominance and with low rates of sexual or physical abuse. Our group of five patients with sleep-related NESs, with its male majority and lack of abuse history, may better fit into this category.

The differentiation of NES from ES can be challenging, despite v-EEG data. Only 20% of simple partial ES (i.e., sensory, motor, or psychic events involving no alteration in consciousness) and ∼90% of complex partial ES are associated with EEG changes (11). Clinical features typical of ES and atypical of NES (2,5–11) are needed for proper diagnosis in such surface EEG–negative cases. Based on these criteria (including episode onset, duration, content, postictal state, and suggestibility), our patients' episodes are readily classified as NES, as described in Table 1.

Our patients' NES episodes can similarly be distinguished from seizures that originate in the mesial or orbital frontal cortex (16,17). These surface-negative episodes have features not seen in other forms of ES, including bilateral, complex movements with irregular frequencies, such as pelvic thrusting, genital manipulation, limb thrashing, or bicycling movements. However, as with all ES, they are characterized by stereotypy, with individual episodes typically brief, and occurring in bursts of multiple recurrences followed by prolonged quiescence (7). This is atypical for NES (18,19) and was not characteristic of our patients' episodes, as described in Table 1.

Parasomnias might possibly account for our patients' sleep-related events. However, this is unlikely because whereas parasomnias represent a disorder of sleep physiology itself, most of our patients' episodes continued unabated after the patient wakened. Further, each event, including the “preictal” and “postictal” state, was similar to the events that the patients had while awake. In several patients, nearly identical NES episodes occurred while fully awake, within 1 h after the sleep-related episode. One woman had an analogous episode the following night, while she appeared to be asleep but was awake by EEG criteria (pseudosleep). When both NES and parasomnias were described in one patient (20), they were clinically very different, and the parasomnias stopped with a benzodiazepine (BZD), while daytime NES continued unabated. Additionally, that patient had abnormal EEG sleep architecture, whereas our patients did not.

A possible limitation of our study is that an alpha EEG rhythm may not be evident in all wakeful situations for all patients. However, every patient in the cohort had robust alpha architecture during daytime epochs of definite arousal, over the course of the admission to the v-EEG unit. Moreover, during the onset of all episodes considered here, the patients appeared to be sleeping on video observation. Thus the conditions that do elicit alpha EEG—a relaxed patient with eyes closed—were clearly fulfilled. As a next step, polysomnography, together with full v-EEG monitoring, would aid in further characterization of the relation between sleep architecture, the transition to arousal, and the precise onset of the NES episodes.

This is the largest study to date of NESs during real or apparent sleep. We likely underestimated the prevalence of NES by excluding many probable NES cases (e.g., those without major motor manifestations). Thus the prevalence of NES in sleep we report (0.99%, eight of 806 NES episodes screened) is probably conservative. In contrast to the low prevalence of NES in true sleep, we found that NES in pseudosleep were fairly common (12.9%, 104 of 806). Benbadis (21) suggested that preictal pseudosleep is 100% specific for NES, but the question of whether the patients were feigning sleep or were merely attempting to fall asleep remains unanswered.

Bazil et al. (2), in their retrospective review of seizures and sleep, including 280 NES episodes, found no NES in sleep, and found that only 6% of NES episodes occurred between midnight and 6 a.m. The failure to document a single case may reflect the low prevalence and sample size. If a significant number of their 280 NES cases were subjective events, the expected number of motor NES in sleep might be less than one. Green (22) described a patient with surface EEG–negative seizure-like episodes, occurring only when the patient appeared to be asleep. However, it is unknown whether the episodes occurred during sleep or pseudosleep, and surface-negative frontal seizures or parasomnias cannot be excluded as possible etiologies.

As for the episodes in our cohort that occurred within several seconds after EEG arousal, it is extremely unlikely that these patients could have developed a concrete plan (consciously or unconsciously) to stage a pseudoseizure during this very brief interval. Sleep inertia is the term used to describe the time that begins immediately with arousal, which is characterized by confusion and poor cognitive function, with a return to baseline levels over 15–90 min (23–26). Although our cases involved the first few seconds of arousal, a period that has not been as carefully subject to study as the first few minutes of arousal, one may nevertheless conclude a fortiori that rapid decision making and plan formulation are unlikely to occur. Possibly recurrent daytime NES may facilitate development of “automatized” NES that can be triggered in an unconscious or transitional state of consciousness.

Along similar lines, recent work has begun to characterize nocturnal panic attacks, some of which seem to emerge from true sleep (27,28). Patients with such attacks may differ from others in their baseline autonomic regulation and in having a higher prevalence of recall of psychological trauma. It is possible that resulting neurochemical sleep-wakefulness differences or a dream-related neurophysiologic cascade triggers these episodes from sleep. Analogous work in characterizing NES from sleep would be of value.

Devinsky and Gordon (4) described four cases of NES that immediately followed partial complex or absence ES, with a lag of <3 s. Given the close temporal proximity to a state of altered consciousness, these cases may also exemplify NES that are not the products of the conscious mind. In these cases, electrical seizures may have triggered a particular psychological state leading to the expression of NES.

We conclude that, in a small minority of cases, NES occur in sleep or in the immediate transition out of sleep, and thus emergence from sleep does not inherently exclude the diagnosis of NES (although, given the rarity of such emergence, the possibility of ES must be exhaustively investigated in these cases). This observation supports the theory that conversion disorder is not the product of consciousness. This report offers one of the few documented examples in which a psychological disorder was demonstrably manifest during sleep. The question of whether and how other psychiatric conditions (such as panic attacks) might present during sleep should be pursued. Such a study would have the benefit of circumventing the complications of arousal and conscious volition and might allow a clearer view of the disease processes themselves, opening the way to new insights.

Acknowledgment: We thank Karen Starner for assistance in collecting and organizing patient data, especially video-EEG reports.