Despite the clinical importance of diagnosing and treating nightmares (Gieselmann et al., 2019), the brain correlates of this enigmatic disorder are still largely unknown. Most studies of nightmare pathophysiology have focused on polysomnographically derived measures of sleep architecture (Germain and Nielsen, 2003; Nielsen et al., 2010b; Simor et al., 2012; Kis et al., 2014; Paul et al., 2015; Marquis et al., 2017; Blaskovich et al., 2019b), periodic leg movements (Germain and Nielsen, 2003), cardiac variability (Nielsen et al., 2010a; Simor et al., 2014; Perogamvros et al., 2019) or EEG metrics such as spectral power (Simor et al., 2013, 2014; Marquis et al., 2017; Blaskovich et al., 2019a), heartbeat-evoked potential (Perogamvros et al., 2019) and sleep spindle frequency and density (Nielsen et al., 2017, 2019; Picard-Deland et al., 2018a, b).

While such methods have provided useful clues to the neural mechanisms involved in nightmares, more direct brain imaging methods are beginning to identify implicated brain regions. Based on the findings of one study of a large cohort of patients with various types and localizations of brain damage, Solms (1997) proposed that some recurring nightmares are due to epileptiform activity in the temporal lobe. In contrast, a second study of 23 patients with bilateral calcification of the basolateral amygdala due to Urbach-Wiethe Disease (Blake et al., 2019) found that, compared to controls, patients with this lesion type have less frequent nightmares. Lesion studies therefore point to alterations in temporal regions and in the basolateral amygdala as contributors to nightmares.

There are still only a few research investigations of non-brain-lesioned patients with nightmares that employ brain imaging. All of these were realized during wakeful, resting-state conditions. Three of these were presented at conferences and/or published as abstracts by the same research group. One Suh et al. (2017) investigated resting-state seed-based functional connectivity in participants with Nightmare Disorder (n = 9) and healthy controls (n = 5) with a focus on posterior cingulate cortex, an important node of the Default Mode Network (Raichle et al., 2001). Compared to controls, nightmare participants had decreased functional connectivity in the right middle cingulate cortex and the medial frontal gyrus, and increased connectivity in the inferior frontal gyrus and parts of the cerebellum. A second Suh et al. (2018) investigated functional connectivity in 12 female Nightmare Disorder participants before and after 5 weekly sessions of Imagery Rehearsal Therapy, with the anterior cingulate cortex used as seed. Post-therapy, participants exhibited decreased connectivity between the anterior cingulate cortex and medial/orbitofrontal cortex and some parietal areas. Last, Park et al. (2019) compared the functional connectivity of Nightmare Disorder participants (n = 12) with that of an age- and sex-matched control group. Compared to controls, nightmare participants exhibited decreased connectivity between the posterior cingulate cortex and the left superior frontal gyrus. The results from these three investigations are quite inconsistent, possibly due to small sample sizes and the variable presence of trauma histories in the participants.

Three brain imaging investigations of nightmares have been published as journal articles. A recent single photon emission tomography (SPECT) study from our laboratory (Marquis et al., 2019a) showed that retrospectively measured dysphoric dream frequency and nightmare distress are negatively correlated with regional cerebral blood flow in several brain regions, but most consistently with blood flow in anterior cingulate cortex and medial prefrontal cortex. Additional findings from this study presented in conferences are that (1) these results are independent of psychopathology (Marquis et al., 2019b) and habitual dream recall frequency (Marquis et al., 2019c) and (2) multiple regression analyses suggest a lateralization pattern by which dysphoric dream frequency correlates primarily with left-hemisphere regions, whereas nightmare distress correlates primarily with right-hemisphere regions (Marquis et al., 2019d). Another recent study using near-infrared spectroscopy and a similar picture-viewing task during wakefulness, partially replicated past SPECT results for frontal regions (Carr et al., 2020b).

The third published study focused on Regional Homogeneity (ReHo) of functional magnetic resonance imaging (fMRI) blood oxygen level-dependent (BOLD) signals (Shen et al., 2016). ReHo is a measure of local neural synchronization obtained by computing Kendall’s concordance coefficient for each voxel and its (usually 26) nearest contiguous voxels (Zang et al., 2004). Shen et al. (2016) found that nightmare participants (N = 15) had, compared to the control group (N = 15), increased ReHo in the left anterior cingulate cortex and right inferior parietal lobule, and decreased ReHo in the left superior and inferior frontal gyri and bilateral middle occipital gyri. Additionally, the Physical Effect subscale of the Nightmare Experience Questionnaire (NEQ; Chen et al. (2014)), which measures “adverse effects of physical health, appetite, and other daily activities after nightmares,” was found to be positively correlated with ReHo values in the anterior cingulate cortex and inferior parietal lobule in the nightmare group (Shen et al., 2016). In the control group, however, this subscale was positively associated with ReHo values in the inferior frontal gyrus.

In sum, the neural correlates of nightmares remain largely unknown although lesion studies point to a role for temporal lobe regions and the basolateral amygdala in nightmare production. Existing brain imaging studies focusing on connectivity have (1) only been published as abstracts; (2) employed small sized samples; and (3) obtained inconsistent results but converge partially with two published brain imaging studies (Shen et al., 2016; Marquis et al., 2019a).

In light of the paucity of brain imaging research on nightmares, we aimed to document the ReHo correlates of nightmare-prone individuals using multiple nightmare severity measures. We compared ReHo of the BOLD signal for groups of frequent nightmare recallers and healthy controls. We also assessed whether there are graded relationships between scores on the Nightmare Distress Questionnaire and ReHo values in both groups. Finally, we assessed relationships between retrospective and prospective disturbed dreaming frequencies and ReHo values.

Our analyses yielded several significant findings. For simplicity, in the following section, we label the analyses according to their corresponding table number. Therefore, #2 corresponds to group comparisons in Table 2, #3 corresponds to NDQ-ReHo correlations for the NM group in Table 3, #4 the same correlations as in #3 but for the CTL group (Table 4), #5 corresponds to correlations between retrospective disturbed dreaming frequency estimates in Table 5 and #6 corresponds to correlations between prospective disturbed dreaming frequency estimates in Table 6. The following summarizes central results that (a) are featured in multiple analyses, (b) have a special theoretical relevance, or (c) are relevant in the context of previous empirical research on the correlates of NMs.

Notable results are found in frontal regions, namely the medial prefrontal gyrus (increased ReHo for analyses #2 and #3, decreased ReHo for analysis #5) and the inferior frontal gyrus (increased ReHo for analyses #2–#3, decreased ReHo for analysis #4). Surprisingly, results involving the anterior cingulate gyrus were only found in analysis #5 (decreased ReHo/negative correlation). The middle temporal gyrus was featured in analyses #3 and #4 (increased ReHo) and #5 (decreased ReHo); the results involving other temporal areas were inconsistent. Among occipital regions, the fusiform gyrus was consistently featured (increased ReHo) in analyses #2, #4 and #5 while the precuneus was featured in analyses #2 and #4 (increased ReHo) and #3 (decreased ReHo). For parietal regions, the supramarginal gyrus was featured in analyses #2, #4 and #6 (increased ReHo) and the inferior parietal lobule was featured in analyses #3 and #5 (decreased ReHo) and #4 and #6 (increased ReHo). Among subcortical regions, the thalamus was featured in analyses #2 and #4 (decreased ReHo) and #5 (increased ReHo). The hippocampus was featured in analysis #5 (increased ReHo). Finally, the cerebellum was featured in analyses #2, #5 and #6 (decreased ReHo) and #4 (increased ReHo). In sum, we found nightmare severity to be associated with altered ReHo in various frontal, temporal, occipital and parietal regions.

As is evident from this summary, there are inconsistent results between analyses. However, some of these inconsistencies may have theoretical relevance. For example, inconsistencies between analyses #3 and #4 may indicate brain correlates of nightmare distress that are unique to frequent NM sufferers. Such results could challenge the hypothesis of shared brain structures involved in all disturbed dreaming/nightmares (Levin and Nielsen, 2007), but more work will be necessary to test this hypothesis. Inconsistencies between analyses #3 and #5 could be the result of different neural correlates underlying NM frequency and NM distress. Limited evidence points to only partially shared brain correlates for the two variables (Marquis et al., 2019a).

One of our findings is consistent with those of published results that also used a ReHo approach (Shen et al., 2016). We found a negative correlation between ReHo values in the inferior parietal lobule and NDQ scores in the NM group: Shen et al. obtained a similar result using a Nightmare Experience Questionnaire subscale.

However, apart from this similarity our findings differed from published results (Shen et al., 2016). We did not find group differences for ReHo values in the anterior cingulate cortex nor in the inferior parietal lobule. We found that frequent nightmare recallers have increased ReHo values in the inferior frontal gyrus, while Shen et al. (2016) found the opposite result. Shen et al. observed a positive correlation between the NEQ Physical Effect subscale and the ReHo values in the inferior frontal gyrus (CTL group only), while in our CTL group we observed a negative correlation between NDQ scores and ReHo values in the same region.

We propose that three main factors contribute to the discrepancies between our results and those from this past study (Shen et al., 2016). First, for the correlational analyses, the two studies used a different questionnaire measure (the NDQ and NEQ). The two questionnaires measure similar but not identical constructs, and there is no study comparing these measures. Second, the subject pools of NM sufferers are different and, arguably, not comparable. While our NM participants experienced more frequent NMs than did those of Shen et al., the level of nightmare distress of their group is unclear. They diagnosed their participants with DSM-5 criteria (American Psychiatric Association, 2013), which is very different from measuring levels of nightmare distress. Clinically, nightmare distress is considered more important than nightmare frequency. Third, their control group was recruited with very stringent criterion (0–1 nightmares per lifetime). This method of selecting participants may have increased the contrast between groups. As suggested by a reviewer, one way to reconcile results from the two studies is to propose different neural correlates for different levels of nightmare severity. This possibility should be addressed in future work because, as indicated earlier, the study cohorts are arguably not comparable.

Our results are more in line with a recent SPECT study from our laboratory (Marquis et al., 2019a). While not present in all our analyses, both studies highlight the possible role of the anterior cingulate cortex and medial prefrontal cortex. Both studies also have found nightmares to be associated with cortical regions involved in sensory processing (parietal, occipital, and temporal cortices). Finally, both studies produced stronger results for retrospective than for prospective disturbed dreaming estimates.

Our neurocognitive model of nightmares (Levin and Nielsen, 2007; Nielsen and Levin, 2007) proposes that nightmares are produced mainly by dysfunction in a limbic-prefrontal circuit comprising the amygdala, hippocampus, anterior cingulate cortex and medial prefrontal cortex. As stated earlier, the limited brain imaging literature on nightmares mainly supports the possible involvement of the anterior cingulate and medial prefrontal cortex. In a previous article by our group, these results were interpreted primarily as evidence of a cross-state alteration of emotional processing. While our neurocognitive model emphasizes a role for these structures in emotion regulation, other research also suggests roles in emotional appraisal (Dixon et al., 2017) and emotional generation/expression (Etkin et al., 2011). Recent studies suggest some form of cognitive appraisal involved in the experience of nightmare distress (Gieselmann et al., 2020), but it is unknown which structures contribute to this evaluative process. We hypothesize that the medial prefrontal and anterior cingulate cortices are likely candidates, among other possible structures.

Our results show that retrospective disturbed dreaming frequency was positively associated with ReHo values in the hippocampus. Albeit not a result found in all of our analyses, it is the first result supporting a role for the hippocampus in nightmare production. From a theoretical standpoint, the possible role of the hippocampus — as for the other regions featured in our neurocognitive model of nightmares—is supported by: (1) brain imaging findings that the hippocampus remains active during REM, (2) brain imaging findings suggesting hippocampal alterations in post-traumatic stress disorder, a disorder featuring repetitive nightmares, and (3) a role for the hippocampus in emotion regulation, including in fear conditioning [reviewed in Levin and Nielsen (2007)]. Accordingly, another theoretical model of REM sleep’s role in emotional memory and emotional adaptation also emphasizes hippocampal involvement (Walker and Van Der Helm, 2009). More study is needed to clarify how the hippocampus (and the previously mentioned frontal regions) may contribute to nightmare production. For example, we provide evidence that REM theta activity, a presumed correlate of hippocampal activity, is a marker of frequent nightmares (Marquis et al., 2017).

The findings of associations between nightmares and activity in parietal, occipital and temporal regions would challenge a view that nightmares are only produced by a limbic-prefrontal circuit. While acknowledging that the implication of other regions is likely, the neurocognitive model does not propose a role for these other regions. It is possible that these results are better understood in reference to a ‘differential susceptibility’ model, in which nightmare sufferers are hypothesized to have heightened sensory processing for both positive and negative stimuli (Carr and Nielsen, 2017; Carr et al., 2020a, b). For the moment, no specific brain correlates are posited for this heightened sensory processing that is specific to nightmares, but our results raise the possibility that nightmares are associated with alterations in sensory processing in addition to alterations in emotion processing. Nonetheless, this possibility is speculative, especially given the paucity of literature on the brain correlates of nightmares, as mentioned in the introduction.

Alterations in regions involved in sensory processing may also have clinical relevance. They may help clarify mechanisms underlying the efficacy of Imagery Rehearsal Therapy, a short-term intervention efficient in diminishing nightmare frequency and alleviating nightmare distress (Krakow and Zadra, 2006; Morgenthaler et al., 2018). Despite some recent progress (Rousseau and Belleville, 2018; Gieselmann et al., 2019), the mechanisms underlying interventions specifically targeting nightmares remain unclear. But it has been proposed that linking the brain correlates of nightmares to those of lucid dreaming may give insight into mechanisms underlying reductions in nightmare frequency (Carr et al., 2020b).

At present, no theory of nightmares can account for all the current findings. Despite some limitations that will be described below, the present study builds upon other brain imaging studies of nightmares (Shen et al., 2016; Marquis et al., 2019a) and highlights the need for theoretical developments that would reconcile all the findings. Our study also provides insight into possible common neural substrates (Harvey et al., 2011; Feldker et al., 2017) for nightmares and other psychiatric disorders for which nightmare frequency is elevated (Swart et al., 2013). While beyond the scope of this article, it is worth pointing out that ReHo is altered in a number of such disorders, such as post-traumatic stress disorder (Ke et al., 2017), primary insomnia (Wang et al., 2016) and depression (Iwabuchi et al., 2015). There is also ongoing research on the genetic similarity between nightmares and other disorders (Ollila et al., 2019).

It is possible that our focus on disturbed dreaming frequency, rather than on bad dreams or nightmare frequency separately, influenced the results. Since nightmares are considered to be more emotionally and visually intense than bad dreams (Fireman et al., 2014; Robert and Zadra, 2014), they may well have stronger, more easily observable brain correlates. However, our participants did not show much variability in their nightmare frequencies, so combining bad dreams and nightmares optimized the power of our statistical tests.

There are also limitations in our MRI acquisition procedures. Although we matched the sequences from two scanners as closely as possible and made appropriate statistical corrections, we cannot rule out the possibility that our results were impacted. We also cannot rule out the possibility that left-handed participants in our two groups differed in the nature of their brain lateralization profiles—especially with such small samples. Nonetheless, the inclusion of left-handed individuals increases the representativity of our sample and the inclusion of similar proportions of left-handers in each group at least partially mitigates the possibility of significant neural differences. Some characteristics of our study sample may also have affected results. Past research has associated nightmares with neuroticism and/or trait anxiety (Blagrove et al., 2004; Miro and Martinez, 2005) and group comparisons between nightmare recallers and control participants often show differences on these measures (Simor et al., 2012; Carr et al., 2018). Our nightmare participants were not seeking treatment, did not display heightened trait anxiety and were likely less distressed by their nightmares than patients in other clinical studies; thus, our findings may not be generalizable to a broader population of nightmares sufferers.

This study aimed to document the neural correlates of nightmare-prone individuals using regional homogeneity of the BOLD signal. We compared ReHo values between frequent nightmare recallers and controls. We also correlated ReHo values with nightmare severity variables: NDQ scores, retrospective disturbed dreaming frequency and prospective disturbed dreaming frequency. Our results mainly implicate cortical (parietal, occipital, temporal) areas, as did a previous study from our laboratory (Marquis et al., 2019a). Other results implicating the medial prefrontal cortex, anterior cingulate cortex, and hippocampus (in one analysis), are consistent with a neurocognitive model (Levin and Nielsen, 2007; Nielsen and Levin, 2007). However, the present findings also highlight the limitations of existing pathophysiological models of nightmares as none of these models can easily reconcile all of our results. Thus, more research and theoretical development is needed. While studying nightmare recallers during wakefulness is convenient and useful in demonstrating cross-state alterations in emotional processing or brain structure/function, future research should include brain imaging during sleep along with dream collection. In time, insights gained into the mechanisms underlying nightmare production promise to translate into improvements in clinical practice.

The datasets presented in this article are not readily available because participants did not agree to this in their signed consent forms. Requests to access secondary results and statistical analyses can be obtained upon request. Requests to access the datasets should be directed to corresponding author.

This study was reviewed and approved by the CIUSSS-NIM Ethics Committee (Hôpital du Sacré-Coeur de Montréal; ID number 2014-1039) and the Ethics Committee Affiliated with the Functional Neuroimaging Unit (ID number CMER-RNQ-15-16-17). Participants provided written informed consent to participate in the study.

L-PM: conceptualization, data collection, data analysis, manuscript writing, and revision. S-HJ, CB-C, and MC: data collection and analysis. VD: conceptualization and data analysis. TP: conceptualization, data collection, and analysis. J-PS and JM: conceptualization and funding acquisition. TN: conceptualization, funding acquisition, supervision, manuscript writing, and revision. All authors read and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.