Prevalence of near-death experiences in people with and without REM sleep intrusion

Hypotheses and research questions

The objectives of this study were two-fold.

• Primary objective: To estimate the frequency of near-death experiences and REM sleep intrusions reported in a large sample of adult humans collected from an online crowdsourcing service.

• Secondary objective: To test the hypothesis that people who report a near-death experience have a greater frequency of REM sleep intrusions.

Study design

An online platform, Prolific Academic (https://prolific.ac/), was used to recruit a large global sample of lay people. Prolific Academic is a crowdsourcing online platform to recruit human subjects for research that compares favorably in terms of honesty and diversity of participants and data quality (Woods et al., 2015; Peer et al., 2017). Participants were recruited without any filters except for English language and age ≥18 years.

The survey was announced under the headline “Survey on Near-Death Experiences and Related Experiences”. Prior to the start of the survey, participants were instructed that their monetary reward was fixed, regardless of whether they would report having had a near-death experience or not. No further information was given. We then asked participants to complete a questionnaire comprising demographic information on age, gender, employment status and place of residence, followed by four questions about REM sleep intrusion (Table 1). The questions about REM sleep intrusion were identical to those used by Nelson et al. in their case-control study (Nelson et al., 2006). Participants were then asked if they ever had had a near-death experience (regardless if this experience had occurred in life-threatening or non-life-threatening situations). If not, the survey ended here; if yes, we continued to inquire about this experience in detail, including all 16 items of the Greyson Near-Death Experience Scale (GNDES), the most widely used standardized tool to identify, confirm and characterize near-death experiences in research (Greyson, 1983). In addition, we inquired about unpleasant feelings (Table 1), which is not covered by the GNDES. Further, participants were given the possibility to describe their experience in their own words. See Table 1 for details.

Statistics

We estimated the number of participants required to be 384, using a very high population size (300,000,000), a confidence level of 95% and a margin of error of 5%. However, as previous studies have estimated that 4–8% of the population have had a near-death experience (Knoblauch, Schmied & Schnettler, 2001; Perera & Padmasekara, 2005; Facco & Agrillo, 2012), we decided to enroll approximately 1,000 participants to identify an estimated 50–100 individuals with near-death experiences.

Data were analyzed using R (R Core Team, 2018). Categorical variables were analyzed using a chi-squared test. Continuous variables were compared using Student’s t-tests. We calculated odds ratios for having near-death experiences with or without co-occurrence of REM sleep intrusion and performed multivariate logistic regression analysis to correct for age, gender, employment status, place of residence, and whether the situation in which an experience was made was perceived as life-threatening or not. To adjust for multiple testing, we used Bonferroni correction and set the alpha level at 0.01.

To prevent HARKing (Hypothesizing After the Results are Known) (Fraser et al., 2018), we pre-registered the study, including all objectives, with the Open Science Framework (https://osf.io/ykr3g).

Ethics

Participants gave consent for publication of their (anonymous) data. Participation was anonymous, voluntary and restricted to those ≥18 years. Participants received a monetary reward upon completion of the survey, following the platform’s ethical rewards principle (≥ $6.50/h). The Ethics Committee of the Capital Region of Denmark waives approval for online surveys (section 14 (1) of the Committee Act. 2; http://www.nvk.dk/english).

Data availability statement

All de-identified raw data are provided in the Supplemental Information. The data analysis plan can be accessed via the Open Science Framework (https://osf.io/ykr3g).

Near-death experiences: prevalence and semiology

Two-hundred eighty-nine participants (28%; 95% CI [25–31%]) claimed to have had a near-death experience, 106 of which reached the threshold of 7 points or more on the GNDES (37%; 95% CI [31–43%]). Thus, confirmed near-death experiences were reported by 106/1,034 participants (10%; 95% CI [8.5–12%]) (Fig. 2).

Participants perceived the situation in which they had their experience slightly more often as truly life-threatening (n = 158; 55%) than not truly life-threatening (n = 131; 45%), and this was irrespective of whether their experience met criteria for a near-death experience according to the GNDES or not (p = 0.55; Table 3).

Near-death experiences and experiences with ≤ 6 points on the GNDES occurred in the following situations, listed with decreasing frequency: motor accident 27% (n = 77), near-drowning 19% (n = 56); intense grief or anxiety 18% (n = 51), substance abuse 11% (n = 33); psychological distress without organic disease 9.7% (n = 28); physical violence other than combat 8.3% (n = 24), critical illness 8.0% (n = 23); childbirth complication 8.0% (n = 23); suicide attempt 6.9% (n = 20); anesthesia/medical procedure 6.9% (n = 20); cardiac arrest/heart attack 5.5% (n = 16); meditation or prayer 5.2% (n = 15); anaphylactic reaction 4.8% (n = 14); combat situation 3.8% (n = 11); syncope 1.7% (n = 5); epileptic seizure 1.4% (n = 4); and other 19% (n = 56).

The most often reported symptoms were abnormal time perception (faster or slower than normal; reported by 252 participants; 87%); exceptional speed of thoughts (n = 189; 65%); exceptional vivid senses (n = 182; 63%); and feeling separated from one’s body, including out-of-body experiences (n = 152; 53%).

Experiences that qualified as a true near-death experience according to the GNDES were perceived much more often as pleasant (n = 41; 53%) than experiences that did not (n = 21; 14%; p < 0.0001; chi-squared test; neutral experience excluded; Table 3).

Around one third of participants reporting a near-death experience or near-death-like experience stated having had two or three such experiences (n = 92, 32%); and some even claimed to have had more than three (n = 10; 3.5%).

Near-death experiences and evidence for REM sleep intrusion

Evidence for REM sleep intrusion (i.e., ≥3 criteria fulfilled) was much more common in people with experiences above the cut-off point of the GNDES (n = 50∕106; 47%) than in people with experiences below this threshold (n = 47∕183; 26%) or in those without any such experience (n = 107∕744; 14%; p =  < 0.0001; chi-squared test; Tables 2 and 3). Following multivariate regression analysis to adjust for age, gender, place of origin, employment status and perceived danger, this association remained highly significant; i.e., people with REM sleep intrusion were more likely to exhibit near-death experiences than those without REM sleep abnormalities (odds ratio 2.85; 95% CI [1.68–4.88]; p = 0.0001; Table 4).

Selected written reports from participants can be found in Tables 5 and 6. Raw data are provided in the online supplemental files.

Prevalence and semiology of near-death experiences

Using crowdsourcing methods, we found that one out of 10 people from a large sample of 35 countries had a confirmed near-death experience (10%; 95% CI [8.5–12%]). This estimate is slightly higher than what was reported in previous studies using traditional interview-based surveys in Australia (8%) (Perera & Padmasekara, 2005) and Germany (4%) (Knoblauch, Schmied & Schnettler, 2001); but, of note, none of those studies validated reports with the GNDES. Experiences that did not fulfill criteria for a near-death experience were roughly twice as common in our survey.

Similar to previous reports, we found that near-death experiences occur in various cultures and nationalities and irrespective of employment status, age and gender (Thonnard et al., 2013; Charland-Verville et al., 2015; Martial et al., 2017; Martial et al., 2018; Cassol et al., 2018). However, unlike most previous reports in which near-death experiences were almost always associated with peacefulness and well-being (Thonnard et al., 2013; Charland-Verville et al., 2015; Martial et al., 2017; Martial et al., 2018; Cassol et al., 2018), we found a much higher rate of people stating that their experience was unpleasant. Although experiences with a cut-off score of at least 7 points on the GNDES were more often pleasant (53%) than experiences with a lower score (14%; p < 0.0001), almost half of all near-death experiences were labelled as stressful. It should be noted that (Thonnard et al., 2013; Charland-Verville et al., 2015; Cassol et al., 2018) included participants who had experienced near-death experiences in life-threatening situations only, i.e., excluding experiences made in non-life-threatening circumstances, which might have led to different frequencies of reported positive and negative emotions. In addition, however, we think the discrepancy between the present and many previous studies is likely because the GNDES addresses only pleasant feelings but does not include negative emotions, in contrast to our questionnaire (Greyson, 1983). Indeed, in line with our results, a recent study revealed a rather high proportion of negative near-death experiences (14%) (Cassol et al., 2019). Interestingly, and in accordance with our assumptions, that study also included near-death experiences of different etiologies and showed a higher proportion of suicidal attempts among people with negative near-death experiences (Cassol et al., 2019).

Again unlike previous studies (Thonnard et al., 2013; Charland-Verville et al., 2015; Martial et al., 2017; Martial et al., 2018; Cassol et al., 2018), we found that near-death experiences occurred equally likely in truly life-threatening situations and situations that only just felt so. This discrepancy probably results from the fact that we interviewed unprimed lay people from a large cross-sectional sample, whereas previous studies were retro- or prospectively performed in specific populations such as cardiac arrest survivors (Van Lommel et al., 2001; Cassol et al., 2018). This substantiates previous reasoning that near-death experiences are real experiences and not merely products of fantasy proneness (Martial et al., 2018): People with confirmed near-death experiences (i.e., GNDES score ≥7 points) did not perceive their situations as more dangerous than those whose experience did not qualify as a near-death experience (i.e., GNDES score ≤ 6 points), which argues against tendencies towards overdramatizing.

Near-death experiences and REM sleep intrusion

Our central finding is that we confirmed a possible association of near-death experiences with REM sleep intrusion. Following multivariate analysis, REM sleep intrusion was the only factor that remained significantly correlated with near-death experiences (and indeed very much so: p = 0.0001). This finding corroborates and extends data from the previous study by Nelson and co-workers, in which the life-time prevalence of REM sleep intrusion in 55 humans with near-death experiences was compared with that in age- and sex-matched controls (Nelson et al., 2006). Sleep-related visual and auditory hallucinations and/or sleep paralysis assessed by a questionnaire like the one used in our study were substantially more common in cases with near-death experiences. The authors suggested that under circumstances of peril, near-death experiences are more likely in people with a tendency towards REM sleep intrusion and that REM sleep intrusion might explain much, if not all, of the semiology of these experiences (Nelson et al., 2006). Indeed, as shown in Table 5, two participants from our study gave spontaneous reports of classic REM sleep disturbances (rather than reporting their near-death experience as requested) akin to those seen in people with narcolepsy (Kondziella & Arlien-Soborg, 2006; Scammell, 2015).

Nelson et al. (2006) based the hypothesis of REM intrusion being associated with near-death experiences on several “lines of evidence”: REM intrusion during wakefulness occurs frequently in healthy people (from 1.2–32% for cataplexy to 24–28% for hypnagogic hallucinations (Ohayon et al., 1999; Cheyne, Rueffer & Newby-Clark, 1999)); REM intrusion is the hallmark of narcolepsy that shares semiological features with near-death experiences; and cardiorespiratory afferents may evoke REM intrusion by heightened vagal afferent activity (Nelson et al., 2006). Further, complex dream-like hallucinations like those of REM sleep are well-described with lesions near the mesopontine paramedian reticular formation and the midbrain cerebral peduncles (i.e., peduncular hallucinations) (Galetta & Prasad, 2017), suggesting that dysfunction of the REM-inhibiting serotonergic dorsal raphe nuclei and the noradrenergic locus ceruleus may facilitate REM intrusion (Hobson, McCarley & Wyzinski, 1975; Manford & Andermann, 1998; Kayama & Koyama, 2003; De Lecea, Carter & Adamantidis, 2012; Hasegawa et al., 2017). Nelson and co-workers therefore hypothesized that people with near-death experiences may have an arousal system predisposing them to REM intrusion (Nelson et al., 2006).

The anonymous nature of our online survey enabled us to avoid the selection bias that was a major point of criticism (Long & Janice Miner, 2007) of the Nelson et al. study (Nelson et al., 2006). While the latter used a case-control approach, we evaluated an unselected sample of unprimed adult lay people. We carefully adjusted for confounding factors including age, gender, place of residence, employment status and perceived danger, and we found that the association of near-death experiences and evidence for REM intrusion remained statistically highly significant. Of note, online surveys limit the influence of psychological bias as compared to face-to-face interviews or telephone surveys because there is no incentive to please the investigator by inventing or exaggerating memories (Peer et al., 2017). (There was no monetary incentive to do so either, since we instructed participants that their reimbursement was the same irrespective of whether they would report a near-death experience or not). Also, we recruited a much larger sample size than what can be achieved with lab-based behavioral testing or case-control studies, including respondents from 35 countries, which strengthens the validity and generalizability of our results. It therefore appears that the association of near-death experiences and REM intrusion is real, although future work needs to address the relationship between the two states: Does a tendency to REM intrusion predispose to near-death experiences, as Nelson et al. (2006) argue, or could it be the other way round, i.e., do near-death experiences lead to a propensity for REM sleep intrusion (Long & Janice Miner, 2007)? Either way, we suggest that characterizing the precise biological mechanisms leading to REM intrusion into wakefulness might offer new insights into the physiology of near-death experiences.

Limitations

Despite the advantages outlined above, online studies have limitations that should be acknowledged (Woods et al., 2015; Peer et al., 2017). First, complex clinical and ethical notions are impossible to fully implement in a survey form. For instance, almost a third of our participants claimed to have had two or more near-death experiences or near-death-like experiences, but we do not know if these participants only referred to one experience when completing the GNDES or several, which might have skewed GNDES scores in some cases. Second, although we assessed various demographic factors, there are many with potential importance such as religiosity (Greyson, 2006; Chandradasa et al., 2018) that we did not record. (Although we did assess religiosity in a previous online survey using the same crowdsourcing platform and found that most participants have a secular background (Kondziella, Cheung & Dutta, 2019)). We did neither inquire about medical history (e.g., diagnoses of epilepsy, narcolepsy, sleep deprivation and migraine), medication or use of alcohol and illicit drugs. Third, for obvious (and ethical) reasons participants were informed about the topic of the survey, which could have led to a selection bias by evoking the interest of survey participants with near-death experiences (and hence overestimation of the true prevalence of such experiences). However, as stated above, the prevalence rate in our study is only slightly higher than that reported in earlier studies (Knoblauch, Schmied & Schnettler, 2001; Perera & Padmasekara, 2005; Facco & Agrillo, 2012), and the entire survey was finished during such a short time frame (i.e., the number of required participants was reached within 3.5 h) that word-of-mouth communication of the survey’s topic appears very unlikely. Also, REM sleep intrusion was assessed using a self-reported questionnaire, the sensitivity and specificity of which remains undefined, and objective sleep testing might have yielded different results. However, sleep studies including polysomnography to confirm reports of REM intrusion would have been impractical in such a large multinational sample, and our questionnaire was identical to that from Nelson et al. (2006), reflecting screening for REM sleep intrusion as performed by neurologists in clinical practice (Kondziella & Arlien-Soborg, 2006; DelRosso, Chesson & Hoque, 2013; Scammell, 2015; Schneider & Mignot, 2017). Lastly, it should be noted that we did not inquire about the frequency of REM intrusion or the temporality of near-death experiences and REM intrusion. Thus, we do not know which of them occurred first; and if there is a causal relationship, the direction remains unknown, i.e., REM sleep intrusion might predispose to or be induced by near-death experiences.