Browse the corpus

Consciousness and the Dying Brain. The near-death experience has been reported since antiquity and is often characterized by the perception of light, interactions with other entities, and life recall. Near-death experiences can occur in a variety of situations, but they have been studied systematically after in-hospital cardiac arrest, with an incidence of 10 to 20%. Long attributed to metaphysical or supernatural causes, there have been recent advances in understanding the neurophysiologic basis of this unique category of conscious experience. This article reviews the epidemiology and neurobiology of near-death experiences, with a focus on clinical and laboratory evidence for a surge of neurophysiologic gamma oscillations and cortical connectivity after cardiac and respiratory arrest.

Although many, including physicians and scientists, have asserted that the near-death experience constitutes evidence for an extracorporeal source of consciousness, we reject any non-neurobiological explanation. As a category of experience, we assert that near-death experiences must be considered within the theoretical and empirical framework of the science of consciousness. A comprehensive treatment of consciousness science is beyond the scope of this review (see Seth and Bayne7 for a recent overview), but there are key aspects that provide a necessary context for neuroscientific studies of near-death experiences over the past decade.

d empirical framework of the science of consciousness. A comprehensive treatment of consciousness science is beyond the scope of this review (see Seth and Bayne7 for a recent overview), but there are key aspects that provide a necessary context for neuroscientific studies of near-death experiences over the past decade. Since the 2000s, consciousness has been deconstructed into at least two dimensions: level and content.8,9 The level of consciousness denotes the global state of arousal of an organism; e.g., awake versus somnolent versus sleeping versus anesthetized versus comatose. The content of consciousness denotes the qualitative aspects of experience, e.g., the redness of a rose or the blueness of the sky. Other dimensions such as behavior, connectedness, and sensory organization have been considered and studied.9,10 Unresponsive wakefulness syndrome—formerly known as vegetative state—shows why these dissociable dimensions are necessary, because this condition is characterized by the presence of wakefulness (i.e., on the axis of level of consciousness) in the presumed absence of experience (i.e., on the axis of content of consciousness).8 Classic near-death experiences are often the converse: there is vivid phenomenological content in the absence of observable wakefulness. What is notable about near-death experiences is that they can be mapped to the extremes of this construct; these and related experiences uniquely span a broad state space in a recent three-dimensional framework (fig. 1).6 Because individuals are considered “clinically dead,” they are arguably positioned at the lowest level of consciousness. However, because the experiences themselves are so rich and “realer than real,” they are arguably positioned at the apex of content of consciousness. This is the essence of the paradox of near-death experiences and why many point to these unique experiences as defying neuroscientific explanation.

evel of consciousness. However, because the experiences themselves are so rich and “realer than real,” they are arguably positioned at the apex of content of consciousness. This is the essence of the paradox of near-death experiences and why many point to these unique experiences as defying neuroscientific explanation. Multidimensional framework for consciousness, including near-death or near-death-like experiences. IFT, isolated forearm test; NREM, non–rapid eye movement; REM, rapid eye movement. Used with permission from Elsevier Science & Technology Journals in Martial et al.6; permission conveyed through Copyright Clearance Center, Inc. In the study of consciousness, the so-called explanatory gap is how we bridge the physical properties of the brain with the first-person feelings of experience.11 The explanatory gap exists even assuming a perfectly healthy brain and only a single pixel of subjectivity (e.g., the experience of a single photon in an otherwise dark room). However, this gap widens considerably in the case of near-death experiences, during which the brain is hypofunctional in the context of near-death physiology, while the subjective experience is seemingly among the richest accessible to humans. This is precisely why some posit that there is no scientific bridge that can cross the even wider explanatory gap represented by near-death experiences.

during which the brain is hypofunctional in the context of near-death physiology, while the subjective experience is seemingly among the richest accessible to humans. This is precisely why some posit that there is no scientific bridge that can cross the even wider explanatory gap represented by near-death experiences. The modern science of consciousness emerged in the early 1990s. One of the first formal research programs to be proposed in this new science was the identification of the minimal set of neuronal events and mechanisms sufficient for a specific conscious percept, commonly referred to as the “neural correlates of consciousness.”12 Although there has been considerable progress over the past three decades, major controversies remain unresolved. Most notably, there is still active debate as to whether prefrontal cortex is involved in generating phenomenological content of consciousness or whether it serves only for postperceptual cognitive processing.13,14 Indeed, a number of prominent theoretical frameworks for consciousness hinge on the role of prefrontal cortex.7 Two are of particular relevance for the consideration of near-death experiences: integrated information theory15 and the global neuronal workspace hypothesis.16

y for postperceptual cognitive processing.13,14 Indeed, a number of prominent theoretical frameworks for consciousness hinge on the role of prefrontal cortex.7 Two are of particular relevance for the consideration of near-death experiences: integrated information theory15 and the global neuronal workspace hypothesis.16 Although most neuroscientific approaches to consciousness posit that the integration of neural information is important or necessary for normal conscious experience, integrated information theory moves beyond both necessity and sufficiency to an identity relationship. In other words, consciousness is integrated information, which cannot be trivially reduced to the sum of the neural or informational parts.15,17,18 With respect to near-death experiences, it is important to consider not only what consciousness is for integrated information theory but also where. This theory posits that a confluence of posterior association and sensory cortices functions is the “hot zone” for the content of consciousness19,20; cortex anterior to the central sulcus is not considered relevant for subjective experience.

t only what consciousness is for integrated information theory but also where. This theory posits that a confluence of posterior association and sensory cortices functions is the “hot zone” for the content of consciousness19,20; cortex anterior to the central sulcus is not considered relevant for subjective experience. By contrast, the global neuronal workspace is argued to span the cortex via long-range projections, with the prefrontal cortex playing a key role as a site of “ignition” for recurrent networks across anterior and posterior cortex.21–23 After ignition, reverberant activity in cortex-wide nodes of the global neuronal workspace sustains and amplifies representations while also allowing accessibility to individual cognitive processors.16,24 This broadcasting effect is argued to constitute conscious experience. Although integrated information theory is considered a perceptual theory of consciousness and global neuronal workspace is considered a cognitive theory of consciousness, both share important features. Neither is strictly localizationist, and both assert the importance of broadly integrating the local and differentiated processing of diverse neural areas. Both theories also affirm the importance of recurrent processing, with the key difference being that the prefrontal cortex is considered a key node for such processing in the global neuronal workspace but not for integrated information theory.

integrating the local and differentiated processing of diverse neural areas. Both theories also affirm the importance of recurrent processing, with the key difference being that the prefrontal cortex is considered a key node for such processing in the global neuronal workspace but not for integrated information theory. Both of these theories might contribute to understanding the neurobiology of near-death experiences and may not be mutually exclusive. In addition to level versus content of consciousness, another key distinction is phenomenal versus access consciousness.25 Phenomenal consciousness represents the purely experiential aspect of consciousness, whereas access consciousness represents a wider set of neural activities that can also include working memory or the ability to report an experience (i.e., what we “do” with conscious experience). Because near-death experiences are—by virtue of their memorability and reportability—a form of access consciousness, we would predict engagement of prefrontal cortex, including recurrent processing, in the dying process. However, given the rich phenomenology, we would also expect enhanced activity in the posterior cortical hot zone that is thought to be the source of conscious content.

Since the 2000s, consciousness has been deconstructed into at least two dimensions: level and content.8,9 The level of consciousness denotes the global state of arousal of an organism; e.g., awake versus somnolent versus sleeping versus anesthetized versus comatose. The content of consciousness denotes the qualitative aspects of experience, e.g., the redness of a rose or the blueness of the sky. Other dimensions such as behavior, connectedness, and sensory organization have been considered and studied.9,10 Unresponsive wakefulness syndrome—formerly known as vegetative state—shows why these dissociable dimensions are necessary, because this condition is characterized by the presence of wakefulness (i.e., on the axis of level of consciousness) in the presumed absence of experience (i.e., on the axis of content of consciousness).8 Classic near-death experiences are often the converse: there is vivid phenomenological content in the absence of observable wakefulness. What is notable about near-death experiences is that they can be mapped to the extremes of this construct; these and related experiences uniquely span a broad state space in a recent three-dimensional framework (fig. 1).6 Because individuals are considered “clinically dead,” they are arguably positioned at the lowest level of consciousness. However, because the experiences themselves are so rich and “realer than real,” they are arguably positioned at the apex of content of consciousness. This is the essence of the paradox of near-death experiences and why many point to these unique experiences as defying neuroscientific explanation.

The modern science of consciousness emerged in the early 1990s. One of the first formal research programs to be proposed in this new science was the identification of the minimal set of neuronal events and mechanisms sufficient for a specific conscious percept, commonly referred to as the “neural correlates of consciousness.”12 Although there has been considerable progress over the past three decades, major controversies remain unresolved. Most notably, there is still active debate as to whether prefrontal cortex is involved in generating phenomenological content of consciousness or whether it serves only for postperceptual cognitive processing.13,14 Indeed, a number of prominent theoretical frameworks for consciousness hinge on the role of prefrontal cortex.7 Two are of particular relevance for the consideration of near-death experiences: integrated information theory15 and the global neuronal workspace hypothesis.16

Near-death experiences are difficult to study because they are not unique to a single cause of death. Thus, establishing an incidence is difficult due to the inability to assess near-death experiences systematically in such a wide array of situations (e.g., drowning, trauma) and furthermore because they can sometimes occur during crises that are unrelated to the actual process of dying (e.g., a near-miss motor vehicle accident).26 Thus, the epidemiology of near-death experiences has focused largely on in-hospital cardiac arrests. The seminal study of near-death experience incidence in survivors of cardiac arrest was conducted in the Netherlands and published in 2001.1 This prospective investigation included 344 patients who underwent 509 successful resuscitations. Using what is considered a more liberal scale for near-death experience, 18% of cardiac arrest survivors reported near-death experiences, with no obvious difference in pharmacologic interventions between those who had them and those who did not. Notably, 2-yr outcomes for cardiac arrest survivors with and without near-death experiences demonstrated a transformative effect that affected social attitude, religious attitude, attitude toward death, and an interest in the meaning of life and appreciation of ordinary things.

those who had them and those who did not. Notably, 2-yr outcomes for cardiac arrest survivors with and without near-death experiences demonstrated a transformative effect that affected social attitude, religious attitude, attitude toward death, and an interest in the meaning of life and appreciation of ordinary things. In the same year, Parnia et al.27 published a smaller study of 63 patients (number of successful resuscitations unknown) using a more conservative near-death experience scale and found that 6.3% reported them. Again, there was no obvious difference in the pharmacology of resuscitation efforts. More recently, Parnia et al.28,29 conducted the AWAreness during REsuscitation (AWARE) studies (AWARE I in 2014 and AWARE II in 2023). AWARE I was a prospective, multicenter observational investigation of cardiac arrest survivors, with 140 patients studied for a stage 1 interview and 101 of those patients who completed stage 2 interviews.28 Near-death experiences were reported by 9%, with 2% reporting actual awareness of resuscitation events; a much broader group (46%) reported recollection of some experience (e.g., déjà vu, bright lights). AWARE II was a 25-site prospective study that included more objective measures such as audiovisual testing using a computer and headphones, continuous electroencephalography, and cerebral oximetry.29 There were 567 in-hospital cardiac arrests, of which 53 survived, of which 28 were available for interviews. Of those available for interviews, 21.4% had what was categorized as a transcendent recalled experience of death (a term proposed to replace near-death experience). In summary, near-death experiences and other experiences are common around the time of cardiac arrest and pose a challenge in reconciling apparent behavioral quiescence and a dysfunctional brain with what can be vivid and transformative phenomenology.

How do we bridge the gap between the known neurobiology of consciousness and the consistent reports of experiences that occur around the time of death? Here, we discuss the neurophysiology—from bedside to bench and back—that might account for the emergence of consciousness in the dying brain.

How do we bridge the gap between the known neurobiology of consciousness and the consistent reports of experiences that occur around the time of death? Here, we discuss the neurophysiology—from bedside to bench and back—that might account for the emergence of consciousness in the dying brain. One of the most provocative clinical findings that gestured toward a neural basis for near-death experiences was the observation of the so-called end-of-life electrical surges. Using processed electroencephalogram (EEG; Bispectral Index [Medtronic; Dublin, Ireland], SedLine [Masimo; USA]), a surge of activity occurred just before death in seven critically ill patients who were otherwise neurologically intact (i.e., there was no brain trauma or cerebral infarction).30 Processed EEG indices declined after withdrawal of support as patients became increasingly hypoxic, but thereafter surged to levels consistent with consciousness around the time of death, and then sharply decreased thereafter (see fig. 2 for an example). A follow-up study of 35 critically ill patients showed no end-of-life electrical surge in 7 patients who met criteria for brain death but, of the remaining 28 patients, 13 (46.4%) evidenced a surge.31 A related case series in the setting of organ donation showed that four patients had a surge in processed EEG (Bispectral Index) after withdrawal of support and before cardiac death3; major electromyographic and electrocardiographic artifacts did not appear to account for the finding. However, other case series have reported that end-of-life electrical surges are not observed in critically ill patients after withdrawal of life-sustaining therapy.32

er withdrawal of support and before cardiac death3; major electromyographic and electrocardiographic artifacts did not appear to account for the finding. However, other case series have reported that end-of-life electrical surges are not observed in critically ill patients after withdrawal of life-sustaining therapy.32 End-of-life electrical surge observed with processed electroencephalographic monitoring. This Bispectral Index tracing started in a range consistent with unconsciousness and then surged to values associated with consciousness just before death and isoelectricity. Used with permission from Mary Ann Liebert Inc. in Chawla et al.30; permission conveyed through Copyright Clearance Center, Inc.

phic monitoring. This Bispectral Index tracing started in a range consistent with unconsciousness and then surged to values associated with consciousness just before death and isoelectricity. Used with permission from Mary Ann Liebert Inc. in Chawla et al.30; permission conveyed through Copyright Clearance Center, Inc. Although provocative, there are critical caveats that limit the interpretation of these clinical findings. First, commercially available processed EEG modules cannot reliably distinguish between nonspecific EEG activation and organized neurophysiologic activity associated with conscious processing. Thus, end-of-life electrical surges could merely reflect activity derived from excitatory impulses after hypoxic neurons lyse and release neurotransmitters such as glutamate. Second, they are not able to reliably distinguish between wakefulness and seizure activity, which can also occur in the dying brain. Third, these modules only have a frontal montage and thus provide limited insight into the cortex-wide events related to consciousness, such as functional, directed, or effective connectivity. Fourth, although electromyography is measured by such monitors, it is difficult to exclude entirely. Fifth, large-scale clinical trials have brought into question the reliability of the Bispectral Index in detecting awareness.33,34 What was needed after these clinical reports was a systematic approach in an animal model that could leverage intracranial EEG electrodes and more advanced methods of analyzing the neurophysiology of consciousness.

al trials have brought into question the reliability of the Bispectral Index in detecting awareness.33,34 What was needed after these clinical reports was a systematic approach in an animal model that could leverage intracranial EEG electrodes and more advanced methods of analyzing the neurophysiology of consciousness. To address some of the limitations of the original clinical observations of end-of-life electrical surges, Borjigin et al.35 conducted a study of nine rats that were instrumented with electrodes for EEG (intracranial), electrocardiography, and electromyography. The animals were anesthetized with ketamine and xylazine, a commonly used anesthetic regimen in animal experimentation. After anesthesia, the rats received an intracardiac injection of potassium to induce cardiac arrest and death. Multiple and characteristic neurophysiologic stages were identified after cardiac arrest, with the third including a surge of power and coherence in the theta and gamma bandwidth. The surge of gamma activity was consistent with what was observed in a case series of critical care patients31 but could still relate to nonspecific neuronal activity. More compelling was the fact that directed functional connectivity, measured via symbolic transfer entropy, surged in feedforward (posterior-to-anterior cortex) and even more in feedback (anterior-to-posterior cortex) directions (fig. 3). Although there are limitations in the inferences that can be made, symbolic transfer entropy is an information-theoretic measure that is a surrogate for communication across the brain and a marker for conscious states.10,36 In other words, this suggests that the neurophysiologic activity observed after cardiac arrest was organized and informative. Importantly, additional experiments in this study demonstrated that (1) EEG was not artifactual or derived from electrocardiographic or electromyographic contamination and (2) the surge of power and coherence in gamma and theta oscillations observed after potassium-induced cardiac arrest was also observed in the same time frame after carbon dioxide–induced hypoxic arrest. This study arguably represents the first systematic approach that identified a putative neural correlate of consciousness (i.e., corticocortical coherence and recurrent processing) in the near-death state.

ssium-induced cardiac arrest was also observed in the same time frame after carbon dioxide–induced hypoxic arrest. This study arguably represents the first systematic approach that identified a putative neural correlate of consciousness (i.e., corticocortical coherence and recurrent processing) in the near-death state. Surge of feedforward and feedback connectivity after cardiac arrest in a rodent model. Panel A depicts time course of feedforward (blue) and feedback (red) directed connectivity during anesthesia (A) and cardiac arrest (CA). Panel B shows averages of directed connectivity across six frequency bands. Error bars indicate standard deviation. *** denotes P < 0.001.

ity after cardiac arrest in a rodent model. Panel A depicts time course of feedforward (blue) and feedback (red) directed connectivity during anesthesia (A) and cardiac arrest (CA). Panel B shows averages of directed connectivity across six frequency bands. Error bars indicate standard deviation. *** denotes P < 0.001. A follow-up study by Li et al.37 further investigated the neurophysiologic changes around the time of death. In rats undergoing experimental asphyxia, the inhalation of carbon dioxide stimulated a robust and sustained increase of global cortical coherence and directed connectivity between frontal and posterior cortices in theta and gamma bands, in parallel with an orderly progression of cardiac arrhythmias until the onset of ventricular fibrillation.37 Notably, two sequential clusters of increased coherence at 65 to 115 Hz and at 25 to 55 Hz coincided with distinct changes in heart rate, suggesting interactions between the brain and the heart during the dying process.37 In a rat model in which asphyxia was induced by discontinuing mechanical ventilation, it was found that the electrocorticographic and blood pressure signals progressed through four distinct stages between the onset of asphyxia and circulatory arrest. The third stage was characterized by near-electrocerebral silence but exhibited increased phase coherence of beta and low gamma (13 to 39 Hz) between bilateral motor cortices, accompanied by a posterior shift of higher-frequency (30 to 100 Hz) power to the visual cortex, which were interpreted as possible correlates of conscious processing.38 A high-frequency surge near death has also been observed in the rat hippocampus.39 Local field recordings in dying rats revealed that a lethal dose of chloral hydrate induced a transient surge in beta and low-gamma (30 to 55 Hz) oscillations and increased coherence within and between the hippocampi, which was speculated to facilitate episodic memory retrieval and/or distort perception. The surge was associated with a reduced heart rate but occurred well before cardiac arrest.39

rate induced a transient surge in beta and low-gamma (30 to 55 Hz) oscillations and increased coherence within and between the hippocampi, which was speculated to facilitate episodic memory retrieval and/or distort perception. The surge was associated with a reduced heart rate but occurred well before cardiac arrest.39 To summarize, systematic basic science studies have identified a surge of synchronized high-frequency neurophysiologic activity in diverse rodent models of abrupt asystolic cardiac arrest,35 asphyxia,37,38 and drug intoxication.39 The first seminal study in rodents identified a transient surge of coordinated gamma activity within the first 30 s after cardiac arrest.35 Subsequent studies characterized the complex time-dependent changes in electrophysiologic recordings during the dying process and revealed a surge of activity before the time of cardiac or circulatory arrest,37–39 a finding more aligned with recent findings in dying human patients discussed in the next section.40,41

35 Subsequent studies characterized the complex time-dependent changes in electrophysiologic recordings during the dying process and revealed a surge of activity before the time of cardiac or circulatory arrest,37–39 a finding more aligned with recent findings in dying human patients discussed in the next section.40,41 Informed by the findings of basic science studies, recent investigation has leveraged whole-scalp electroencephalographic recordings and advanced signal analysis methods to characterize neurophysiologic processes in the dying human brain.42 Vicente et al.40 measured continuous, whole-scalp electroencephalography in an 87-yr-old patient undergoing cardiac arrest after traumatic subdural hematoma. A surge in absolute power of low (30 to 60 Hz) and high (80 to 150 Hz) gamma activity was identified after suppression of electroencephalographic activity in both hemispheres, accompanied by the modulation of left-hemispheric gamma power by the phase of the alpha band, similar to the cross-frequency interactions involved in cognitive processes and memory recall in healthy subjects.40

Hz) gamma activity was identified after suppression of electroencephalographic activity in both hemispheres, accompanied by the modulation of left-hemispheric gamma power by the phase of the alpha band, similar to the cross-frequency interactions involved in cognitive processes and memory recall in healthy subjects.40 More recently, Xu et al.41 analyzed electroencephalographic signals in four comatose dying patients.43 In two of four patients, the withdrawal of ventilatory support markedly increased the absolute power of beta and gamma (25 to 150 Hz) and the cross-frequency coupling between the high-gamma (80 to 150 Hz) power in central channels and the beta phase in frontal channels, which is consistent with (although does not demonstrate directly) neural communication between these regions, critical for cognitive processing.43,44 The same two subjects showed a surge of cortical coherence and directed connectivity in the gamma band (25 to 55 Hz) between the regions within the temporal–parietal–occipital junctions, a part of the so-called posterior cortical hot zone for the neural corelates of consciousness.20 Independent component analysis of the multichannel electroencephalographic signals demonstrated that muscle artifact did not fully account for the findings.20 It is possible, however, that high-frequency oscillations could reflect seizure activity, which could induce nonordinary states of consciousness. Indeed, the two patients who evidenced a surge of gamma activity in the study by Xu et al.41 had histories of seizures, whereas there was no such history in the two patients without a surge.20 Although obvious seizures were excluded based on a neurologist’s review of the EEG, it is still possible that small or deep brain areas could have generated seizure activity that was not detected by scalp EEG.

u et al.41 had histories of seizures, whereas there was no such history in the two patients without a surge.20 Although obvious seizures were excluded based on a neurologist’s review of the EEG, it is still possible that small or deep brain areas could have generated seizure activity that was not detected by scalp EEG. These data provide evidence that the human brain can generate organized activity near death, with consistent findings across species. These coordinated activities were observed before the onset of ventricular fibrillation or asystole41,45 (ambiguous in Vicente et al.,40 but see their fig. 2A and the commentary), with timing consistent with that observed in rodent models of asphyxia37,38 and drug intoxication39 but incompatible with the end-of-life electrical surges that appeared within 3 to 6 min of the complete loss of measurable blood pressure.31

1,45 (ambiguous in Vicente et al.,40 but see their fig. 2A and the commentary), with timing consistent with that observed in rodent models of asphyxia37,38 and drug intoxication39 but incompatible with the end-of-life electrical surges that appeared within 3 to 6 min of the complete loss of measurable blood pressure.31 A recent large-scale retrospective study identified transient resumption of cardiac activity after circulatory arrest in 14% of 480 patients during withdrawal of life-sustaining measures, of which no patients regained consciousness or survived.46 A related substudy in eight patients further found that electroencephalographic activity stopped at a median of 78 s before circulatory arrest. Electroencephalographic amplitude, spectral power, and coherence demonstrated a progressive reduction, although one patient had brief resumptions of cardiac activity after more than 60 s of circulatory arrest.32 The study also found a surge in electroencephalographic power in four of eight patients after withdrawal of life-sustaining measures and before circulatory arrest, consistent with past studies.40,41 However, there is no evidence to correlate the observed neurophysiologic changes with any conscious experience or any sign of returning consciousness in the patients who died after terminal cardiac arrest.32,40,41

r withdrawal of life-sustaining measures and before circulatory arrest, consistent with past studies.40,41 However, there is no evidence to correlate the observed neurophysiologic changes with any conscious experience or any sign of returning consciousness in the patients who died after terminal cardiac arrest.32,40,41 Moving forward, new insights into the neural correlates of consciousness around the time of death may be gained from real-time brain monitoring coupled with awareness testing during cardiopulmonary resuscitation, such as in the AWARE II study.29 In a pilot substudy of 85 subjects,29 interpretable electroencephalographic signals (SedLine) were obtained from 53 patients during ongoing cardiopulmonary resuscitation until termination or sustained return of spontaneous circulation. Although detailed analysis has not yet been reported, the presence of near-normal electroencephalographic activity up to 35 to 60 min into cardiopulmonary resuscitation may reflect conscious activity around the time of death or resumption of consciousness in patients who survived.29

One of the most provocative clinical findings that gestured toward a neural basis for near-death experiences was the observation of the so-called end-of-life electrical surges. Using processed electroencephalogram (EEG; Bispectral Index [Medtronic; Dublin, Ireland], SedLine [Masimo; USA]), a surge of activity occurred just before death in seven critically ill patients who were otherwise neurologically intact (i.e., there was no brain trauma or cerebral infarction).30 Processed EEG indices declined after withdrawal of support as patients became increasingly hypoxic, but thereafter surged to levels consistent with consciousness around the time of death, and then sharply decreased thereafter (see fig. 2 for an example). A follow-up study of 35 critically ill patients showed no end-of-life electrical surge in 7 patients who met criteria for brain death but, of the remaining 28 patients, 13 (46.4%) evidenced a surge.31 A related case series in the setting of organ donation showed that four patients had a surge in processed EEG (Bispectral Index) after withdrawal of support and before cardiac death3; major electromyographic and electrocardiographic artifacts did not appear to account for the finding. However, other case series have reported that end-of-life electrical surges are not observed in critically ill patients after withdrawal of life-sustaining therapy.32

To address some of the limitations of the original clinical observations of end-of-life electrical surges, Borjigin et al.35 conducted a study of nine rats that were instrumented with electrodes for EEG (intracranial), electrocardiography, and electromyography. The animals were anesthetized with ketamine and xylazine, a commonly used anesthetic regimen in animal experimentation. After anesthesia, the rats received an intracardiac injection of potassium to induce cardiac arrest and death. Multiple and characteristic neurophysiologic stages were identified after cardiac arrest, with the third including a surge of power and coherence in the theta and gamma bandwidth. The surge of gamma activity was consistent with what was observed in a case series of critical care patients31 but could still relate to nonspecific neuronal activity. More compelling was the fact that directed functional connectivity, measured via symbolic transfer entropy, surged in feedforward (posterior-to-anterior cortex) and even more in feedback (anterior-to-posterior cortex) directions (fig. 3). Although there are limitations in the inferences that can be made, symbolic transfer entropy is an information-theoretic measure that is a surrogate for communication across the brain and a marker for conscious states.10,36 In other words, this suggests that the neurophysiologic activity observed after cardiac arrest was organized and informative. Importantly, additional experiments in this study demonstrated that (1) EEG was not artifactual or derived from electrocardiographic or electromyographic contamination and (2) the surge of power and coherence in gamma and theta oscillations observed after potassium-induced cardiac arrest was also observed in the same time frame after carbon dioxide–induced hypoxic arrest. This study arguably represents the first systematic approach that identified a putative neural correlate of consciousness (i.e., corticocortical coherence and recurrent processing) in the near-death state.

Informed by the findings of basic science studies, recent investigation has leveraged whole-scalp electroencephalographic recordings and advanced signal analysis methods to characterize neurophysiologic processes in the dying human brain.42 Vicente et al.40 measured continuous, whole-scalp electroencephalography in an 87-yr-old patient undergoing cardiac arrest after traumatic subdural hematoma. A surge in absolute power of low (30 to 60 Hz) and high (80 to 150 Hz) gamma activity was identified after suppression of electroencephalographic activity in both hemispheres, accompanied by the modulation of left-hemispheric gamma power by the phase of the alpha band, similar to the cross-frequency interactions involved in cognitive processes and memory recall in healthy subjects.40

Data obtained from rodent models and critically ill patients undergoing withdrawal of care have led to the hypothesis that near-death experiences could be caused by the coordinated high-frequency activities around death.31,35,41 A recent study has investigated the neural origins of the high-frequency activity by simultaneous electrocorticographic and intracellular recordings of neocortical neurons in a rat asphyxia model.47 After the interruption of oxygen supply, the surge of beta–gamma (10 to 50 Hz) electrocorticographic activities was associated with regular firing at 6 to 7 Hz in pyramidal neurons driven by rhythmic membrane oscillations, which were shaped by the summation of high-frequency (100 to 140 Hz) depolarizing and hyperpolarizing synaptic potentials.47 Similar changes in synaptic activities have been previously found in the rat hippocampus in vivo48 and in vitro during cerebral ischemia,49 which were suggested to result from γ-aminobutyric acid– and adenosine-dependent inhibition,49 as well as after hypoxia in mouse neocortical layer 5 pyramidal neurons in vitro, which were suggested to be triggered by activation of glutamate-gated channels.50,51

ocampus in vivo48 and in vitro during cerebral ischemia,49 which were suggested to result from γ-aminobutyric acid– and adenosine-dependent inhibition,49 as well as after hypoxia in mouse neocortical layer 5 pyramidal neurons in vitro, which were suggested to be triggered by activation of glutamate-gated channels.50,51 In parallel with increased high-frequency oscillations, Li et al.37 found significant increases in the cortical release of a wide array of neurotransmitters within the first 2 min of asphyxia, including adenosine, dopamine, norepinephrine, serotonin, γ-aminobutyric acid, glutamate, and aspartate in both the frontal and occipital cortices. The increase in serotonin release was especially marked, with more than a 20-fold surge within the first 2 min and around 250-fold after up to 20 min of asphyxia.37 Serotonin plays diverse roles via different receptors, and activation of serotonergic 5-hydroxytryptamine 2A receptors is considered the primary mechanism by which psychedelics induce visual hallucinations and mystical experiences in humans, some of which resemble near-death experiences.52–54 In addition, a significant increase in dimethyltryptamine release was detected in the rat visual cortex after cardiac arrest,55 although it is unclear whether the resultant concentrations are sufficient to generate near-death experiences.56 It has been posited that the massive release of several transmitters provides a neurochemical basis for the transient surge of high-frequency activity that possibly mediates experience around the time of death,37 but this has not been demonstrated empirically.

ations are sufficient to generate near-death experiences.56 It has been posited that the massive release of several transmitters provides a neurochemical basis for the transient surge of high-frequency activity that possibly mediates experience around the time of death,37 but this has not been demonstrated empirically. It has been argued that the duration of the high-frequency surge in the early phase of the dying process may be too short to explain the richness and the temporal extent of the near-death experience.47 Considered more broadly, the asphyxia-induced dying process involves successive changes in neuronal activity, including the early surge of high-frequency activity followed by low-frequency activities that progressively decline toward isoelectricity. Crucially, the isoelectric state is associated with a block of action potentials and a loss of cell integrative properties, which can lead to death but also could be reversed with successful resuscitation.47,57 For successful resuscitations, near-death experience may be generated during the relatively slow process leading to recovery of consciousness, which was characterized by overwhelming 5- to 10-Hz oscillations and attenuated high-frequency activities that resemble the patterns observed in hallucinatory diseases such as schizophrenia, as proposed by Schramm et al.47 in a rodent model of asphyxia and resuscitation. Although provocative, it should be noted that the observed neurophysiologic patterns throughout the dying-to-recovery process may not be sufficient to constitute the near-death experience, just as Gidon et al.58 questioned in a thought experiment: does the exact replay of neural activity, which was initially recorded during a conscious experience, reconstitute the conscious experience in a disconnected or disorganized brain? It is possible that other neurophysiologic processes, such as the intrinsic cause–effect structure of the brain network, could be essential or complementary neural events generating consciousness. In this context, we therefore cannot exclude the possibility that the near-death experience could also emerge from the period of low-frequency activities or even the isoelectric period, during which the capacity of neurons to engage in cause–effect interactions is not fully disabled but only suppressed and recoverable if brain reoxygenation was rapidly restored.47,57

possibility that the near-death experience could also emerge from the period of low-frequency activities or even the isoelectric period, during which the capacity of neurons to engage in cause–effect interactions is not fully disabled but only suppressed and recoverable if brain reoxygenation was rapidly restored.47,57 Bursts of gamma activity during cortical hypoxia have been studied at the cellular and physiologic levels,59,60 but it is still unclear how gamma activity explosively surges and interacts globally across distant brain regions.35 In particular, how can gamma precipitously increase right before all electroencephalographic waves collapse? It is crucial to understand the underlying mechanism because the integration of distributed neural information appears necessary for the emergence of higher-order brain functions, such as cognition and consciousness,61 of central relevance to near-death experiences.

ht before all electroencephalographic waves collapse? It is crucial to understand the underlying mechanism because the integration of distributed neural information appears necessary for the emergence of higher-order brain functions, such as cognition and consciousness,61 of central relevance to near-death experiences. We propose two potential network mechanisms for the paradoxical surge in global brain connectivity followed by a sudden transition to isoelectric electroencephalogram: explosive synchronization and the percolation process. Explosive synchronization is a universal mechanism of abrupt transitions from an incoherent state to a synchronized state in complex dynamic networks and is characterized by hysteresis (i.e., asymmetric forward and backward pathways).62 Explosive synchronization has been studied in many physical and biologic systems, including power grid failure, epileptic seizure, highly sensitive acoustical signal transduction in the cochlea, and explosive contagion in social networks.62 In recent studies, our research team has demonstrated evidence of explosive synchronization in the hypersensitive brain of chronic pain (fibromyalgia),63 in arousal from light anesthesia,64 and in the anesthetic hysteresis phenomenon (i.e., the difference in anesthetic doses associated with loss and recovery of consciousness).65 Recent human near-death experience data suggest that the dying brain is in a state of competition between two opposing forces, brain network breakdown and homeostasis, two core mechanisms of explosive synchronization. Thus, the dying brain may be poised for an explosive surge in global connectivity just before isoelectricity.

human near-death experience data suggest that the dying brain is in a state of competition between two opposing forces, brain network breakdown and homeostasis, two core mechanisms of explosive synchronization. Thus, the dying brain may be poised for an explosive surge in global connectivity just before isoelectricity. Another potential mechanism for the surge in global gamma coherence and communication is percolation, a random probabilistic process that exhibits a phase transition. This possibility assumes that the percolation process in the dying brain randomly disrupts functional connections one by one. As the process reaches a tipping point, the brain may end up with a mere skeleton network consisting of a few hubs (highly connected brain regions) and pathway shortcuts. At this stage, the dramatically elevated local brain activities due to cortical hypoxia might mutually interact only through the few surviving shortcuts. This could lead to a massive information flow, especially in links between largely segregated brain regions in an unconscious state.66 The temporal–parietal–occipital junctions, which showed a 132-fold increase in surrogates of information flow compared to the resting state,41 might play the role of such a bridge in the dying brain. Finally, when the last surviving shortcut is deleted, the massive information flow will suddenly drop to near zero, typically along with complete brain network failure. This mechanism based on phase transition is only theoretical but, importantly, is well described in nonbiologic networks and thus could be a generalized explanation. Further computational and empirical research is needed to confirm whether explosive synchronization or the percolation process actually plays a role in the dying brain.

Bursts of gamma activity during cortical hypoxia have been studied at the cellular and physiologic levels,59,60 but it is still unclear how gamma activity explosively surges and interacts globally across distant brain regions.35 In particular, how can gamma precipitously increase right before all electroencephalographic waves collapse? It is crucial to understand the underlying mechanism because the integration of distributed neural information appears necessary for the emergence of higher-order brain functions, such as cognition and consciousness,61 of central relevance to near-death experiences.

There has been substantial progress over the past 15 yr toward creating a scientific framework for near-death experiences. It is now known that there can be surges of high-frequency oscillations in the mammalian brain around the time of death, with evidence of corticocortical coherence and communication just before cessation of measurable neurophysiologic activity. This progress has traversed the translational spectrum, from clinical observations in critical care and operative settings, to rigorous study in animal models, and to more recent and more neurobiologically informed investigations in dying patients. But what does it all mean? The surge of gamma activity in the mammalian brain around the time of death has been reproducible and, in human studies, surrogates of corticocortical communication have been correlated with conscious experience. What is lacking is a correlation with experiential content, which is critically important to verify because it is possible that these neurophysiologic surges are not associated with any conscious experience at all. Animal studies preclude verbal report, and the extant human studies have not met the critical conditions to establish a neural correlate of the near-death experience, which would require the combination of (1) “clinical death,” (2) successful resuscitation and recovery, (3) whole-scalp neurophysiology with analyzable signals, (4) near-death experience or other endogenous conscious experience, and (5) memory and verbal report of the near-death experience that would enable the correlation of clinical conditions, neurophysiology, and conscious experience. Although it is possible that these conditions might one day be met for a patient that, as an example, is undergoing an in-hospital cardiac arrest with successful restoration of spontaneous circulation and accompanying whole-scalp neurophysiologic monitoring that is not compromised by the resuscitation efforts, it is unlikely that this would be an efficient or reproducible approach to studying near-death experiences in humans. What is needed is a well-controlled model.

uccessful restoration of spontaneous circulation and accompanying whole-scalp neurophysiologic monitoring that is not compromised by the resuscitation efforts, it is unlikely that this would be an efficient or reproducible approach to studying near-death experiences in humans. What is needed is a well-controlled model. Deep hypothermic circulatory arrest has been proposed as a model, but one clinical study showed that near-death experiences are not reported after this clinical intervention.67 Psychedelic drugs provide an opportunity to study near-death experience–like phenomenology and neurobiology in a controlled, reproducible setting. Dimethyltryptamine, a potent psychedelic that is endogenously produced in the brain and (as noted) released during the near-death state, is one promising technique. Administration of the drug to healthy volunteers recapitulates phenomenological content of near-death experiences, as assessed by a validated measure as well as comparison to actual near-death experience reports.54

ly produced in the brain and (as noted) released during the near-death state, is one promising technique. Administration of the drug to healthy volunteers recapitulates phenomenological content of near-death experiences, as assessed by a validated measure as well as comparison to actual near-death experience reports.54 Of direct relevance to anesthesiology, one large-scale study comparing semantic similarity of (1) approximately 15,000 reports of psychoactive drug events (from 165 psychoactive substances) and (2) 625 near-death experience narratives found that ketamine experiences were most similar to near-death experience reports.53 Of relevance to the neurophysiology of near-death states, ketamine induces increases in gamma and theta activity in humans, as was observed in rodent models of experimental cardiac arrest.68 However, there is evidence of disrupted coherence and/or anterior-to-posterior directed functional connectivity in the cortex after administration of ketamine in rodents,69 monkeys,70 and humans.36,68,71 This is distinct from what was observed in rodents and humans during the near-death state and requires further consideration. Furthermore, psilocybin causes decreased activity in medial prefrontal cortex,72 and both classical (lysergic acid diethylamide) and nonclassical (nitrous oxide, ketamine) psychedelics induce common functional connectivity changes in the posterior cortical hot zone and the temporal parietal junction but not the prefrontal cortex.73 Once true correlates of near-death or near-death–like experiences are established, leveraging computational modeling to understand the network conditions or events that mediate the neurophysiologic changes could facilitate further mechanistic understanding.

Near-death experiences have been reported since antiquity and have profound clinical, scientific, philosophical, and existential implications. The neurobiology of the near-death state in the mammalian brain is characterized by surges of gamma activity, as well as enhanced coherence and communication across the cortex. However, correlating these neurophysiologic findings with experience has been elusive. Future approaches to understanding near-death experience mechanisms might involve psychedelic drugs and computational modeling. Clinicians and scientists in anesthesiology have contributed to the science of near-death experiences and are well positioned to advance the field through systematic investigation and team science approaches. Supported by grant No. R01GM11293 from the National Institutes of Health (Bethesda, Maryland). The authors declare no competing interests.