1. Neuroscience
Download icon

Dreaming with hippocampal damage

  1. Goffredina Spanò
  2. Gloria Pizzamiglio
  3. Cornelia McCormick
  4. Ian A Clark
  5. Sara De Felice
  6. Thomas D Miller
  7. Jamie O Edgin
  8. Clive R Rosenthal
  9. Eleanor A Maguire  Is a corresponding author
  1. Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, United Kingdom
  2. Department of Neurodegenerative Diseases and Geriatric Psychiatry, University Hospital Bonn, Germany
  3. Department of Neurology, Royal Free Hospital, United Kingdom
  4. Department of Psychology, University of Arizona, United States
  5. Nuffield Department of Clinical Neurosciences, University of Oxford, United Kingdom
Short Report
  • Cited 2
  • Views 2,257
  • Annotations
Cite this article as: eLife 2020;9:e56211 doi: 10.7554/eLife.56211

Abstract

The hippocampus is linked with both sleep and memory, but there is debate about whether a salient aspect of sleep – dreaming – requires its input. To address this question, we investigated if human patients with focal bilateral hippocampal damage and amnesia engaged in dreaming. We employed a provoked awakening protocol where participants were woken up at various points throughout the night, including during non-rapid eye movement and rapid eye movement sleep, to report their thoughts in that moment. Despite being roused a similar number of times, dream frequency was reduced in the patients compared to control participants, and the few dreams they reported were less episodic-like in nature and lacked content. These results suggest that hippocampal integrity may be necessary for typical dreaming to occur, and aligns dreaming with other hippocampal-dependent processes such as episodic memory that are central to supporting our mental life.

eLife digest

Dreaming has intrigued humans for thousands of years, but why we dream still remains somewhat of a mystery. Although dreams are not a precise replay of our memories, one idea is that dreaming helps people process past experiences as they sleep. If this is true, then part of the brain called the hippocampus that is important for memory should also be necessary for dreaming.

Damage to the hippocampus can cause a condition called amnesia that prevents people from forming new memories and remembering past experiences. However, studies examining dreaming in people with amnesia have produced mixed results: some found that damage to the hippocampus had no effect on dreams, while others found it caused people to have repetitive dreams that lacked detail. One reason for these inconsistencies is that some studies asked participants about their dreams the next morning by which time most people, particularly those with amnesia, have forgotten if they dreamed.

To overcome this limitation, Spanò et al. asked participants about their dreams immediately after being woken up at various points during the night. The experiment was carried out with four people who had damage to both the left and right hippocampus and ten healthy volunteers. Spanò et al. found that the people with hippocampal damage reported fewer dreams and the dreams they had were much less detailed.

These findings suggest that a healthy hippocampus is necessary for both memory and dreaming, reinforcing the link between the two. Hippocampal damage is associated with a number of diseases, including dementia. If these diseases cause patients to dream less, this may worsen the memory difficulties associated with these conditions.

Introduction

Dreaming has intrigued humans for thousands of years, being variously interpreted as having premonitory, religious, psychoanalytic, or mnemonic significance. Defined as an internally-generated subjective mental experience during the sleep state (Cipolli et al., 2017), dreaming can be present at initial sleep onset (hypnagogic sleep; Horikawa et al., 2013; Stickgold et al., 2000), during non-rapid eye movement (NREM) sleep (Antrobus et al., 1995; Foulkes, 1962; Nielsen, 2000; Siclari et al., 2017; Wamsley, 2013; Wamsley et al., 2007), and rapid eye movement (REM) sleep (Hobson et al., 2000; Maquet et al., 2000; Maquet et al., 2005).

Although dreams are not a precise replay of our memories (Fosse et al., 2003; Stickgold et al., 2001a), it has been proposed that dreaming may be associated with memory consolidation processes (Payne, 2010; Wamsley, 2014; Wamsley et al., 2010; Wamsley and Stickgold, 2019). Indeed, in rodents and humans, patterns of brain activity exhibited during a learning experience were found to be subsequently expressed during sleep (Peigneux et al., 2004; Wilson and McNaughton, 1994).

Bilateral damage to a brain structure called the hippocampus is known to adversely affect memory processing (Miller et al., 2017; Miller et al., 2020; Scoville and Milner, 1957; Spiers et al., 2001) and daydreaming (McCormick et al., 2018). While sleep dreaming has been examined previously in patients with hippocampal lesions, results are mixed, with some studies reporting repetitious and stereotyped dreams (Torda, 1969a; Torda, 1969b; Stickgold et al., 2000), whereas others claim that hippocampal damage has no effect on dreaming (Solms, 1997; Solms, 2013). Possible reasons for these disparate findings include differences in the sleep stages sampled, the inclusion in several studies of patients with psychiatric conditions and below-average IQ and, in other instances, patients were interviewed at a point temporally remote from any dreaming that might have occurred, presenting a challenge for these patients who usually suffer from amnesia.

It remains uncertain, therefore, whether hippocampal integrity is necessary for dreaming to occur. If, as we predicted, bilateral hippocampal damage degrades dreaming, this would reinforce the link between dreaming and hippocampal-dependent processes such as memory, potentially moving us closer to an understanding of why we dream. By contrast, if hippocampal-damaged patients who have a reduced ability to experience complex, imagery-rich, spatio-temporal mental events during wake are nevertheless able to have such experiences during sleep, this would need to be explicitly accounted for within theories of hippocampal function.

Here, we sought to mitigate the issues affecting previous studies while investigating the frequency, features and content of dreaming in rare patients with selective bilateral hippocampal damage and matched control participants. To do this we employed a provoked awakening protocol (Nguyen et al., 2013; Siclari et al., 2013; Wamsley et al., 2016). This widely-used approach involved waking participants up at various times during their night’s sleep to report their thoughts in that moment. In this way we examined sleep mentation in a direct and immediate manner.

Results

We assessed four patients (all right-handed males; mean age 58.25 years, SD ±20.82) with selective bilateral hippocampal lesions and a specific episodic memory deficit (Table 1, Material and methods, Supplementary file 1, Spanò et al., 2020). Patients were matched to ten healthy control participants (all right-handed males; mean age 59.2 years, SD ±15.89) based on a number of demographic factors including age, gender, body mass index, non-verbal IQ, and a range of sleep quality measures (Table 1; Material and methods, Supplementary file 1, Supplementary file 2). We conducted in-home sleep recording using portable polysomnography (PSG) on two consecutive nights – one habituation night to allow for familiarization with the PSG equipment and one experimental night for the collection of dream reports. The purpose of the PSG recording was to ensure that we awakened participants during both NREM and REM sleep, and in a similar manner for the patient and control groups. During the dream sampling night, participants were woken up after a period of 3 min from the onset of either NREM or REM sleep at various times throughout the night (Materials and methods, Figure 1A).

Experimental set-up and key findings.

(A) Two researchers were located in Room one which was adjacent to Room two where the participant slept. The participant was woken up at various times during their night’s sleep to report their thoughts in that moment. PSG recordings informed the decisions about when to awaken the participant to ensure sampling during non-rapid eye movement (NREM) and rapid eye movement (REM) sleep. We used a Bluetooth intercom system equipped with a camera for continuous visual monitoring and communication with the participant. (B) The number of total awakenings was not different between the patients (HPC) and control (CTL) participants. (C) There were also no significant group differences in the proportion of awakenings from NREM and REM sleep. (D) In contrast, the patients reported significantly fewer dreams than the control participants, expressed here as the total number of dreams divided by the total number of awakenings (+ / - 1 SEM; p=0.028). (E) The few dreams the patients had were significantly less rich in content compared to those of the control participants (n = 3 patients, as one patient had no dreams at all and was not included in this analysis; + / - 1 SEM; p=0.018). For other measures see Table 2.

Table 1
Demographic characteristics.
GroupAge
(years)
Chronicity
(years)
LHPC volume (mm3)RHPC volume (mm3)LHPC % volume loss relative to CTLaRHPC % volume loss relative to CTLaWASI
CTL59.20 (15.89)n.a.3173.18a (338.89)3285.91a (300.81)n.a.n.a.14.50 (2.37)
HPC161625062803−21.03%−14.70%12
HPC272817361698−45.29%−48.32%10
HPC3721126072755−17.84%−16.16%12
HPC4281128192804−11.16%−14.67%14
  1. All patients (HPC1-4) and control participants (CTL) were right-handed males. Mean and standard deviation in parentheses are shown for control participants and individual data for the four patients. aThe control group consisted of eleven participants (mean age 55.64 years ± 16.47). LHPC = left hippocampus; RHPC = right hippocampus; n.a. = not applicable; WASI = Wechsler Abbreviated Scale of Intelligence (Wechsler, 1999) Matrix Reasoning subtest scaled score. See Supplementary file 1 and Supplementary file 2 for additional neuropsychological and sleep quality data of the participants.

Table 2 shows the group summary data and the results of the between-group statistical analyses for each of the measures that are described below. While, for the sake of economy, we present the findings in terms of these group comparisons, given the small sample of these rare patients, caution should be exercised in interpreting the results. We, therefore, include the individual patient data in Tables 1 and 2 permitting the patients to be considered also as a series of case studies.

Table 2
Dream characteristics.
HPCCTLHPC1HPC2HPC3HPC4
M (SD)M (SD)UESP-Value
General analysesa
Number of awakenings6.75 (2.75)6.90 (1.97)18.50.110.82910.004.005.008.00
Proportion of awakenings during NREM0.73 (0.09)0.67 (0.13)14.00.470.3830.800.750.600.75
Proportion of awakenings during REM0.23 (0.17)0.33 (0.13)13.00.550.3090.00c0.250.400.25
Number of probes per awakening3.82 (1.75)4.22 (1.13)17.00.230.6715.803.501.604.38
Dream frequency0.23 (0.17)0.67 (0.28)4.51.450.0280.400.250.000.25
Proportion of dreams during NREM0.38 (0.48)0.52 (0.20)14.50.430.4151.000.000.000.50
Proportion of dreams during REM0.38 (0.48)0.48 (0.20)16.00.310.5550.001.000.000.50
Proportion of no dreams0.65 (0.31)0.21 (0.24)4.01.520.0220.600.751.000.25
Proportion of blank dreams0.13 (0.25)0.12 (0.15)15.50.350.4960.000.000.000.50
Overall qualitative attributesb
Number of informative words43.17 (16.06)95.55 (55.20)7.00.810.17655.5025.00.49.00
Complexity2.67 (0.58)3.32 (0.70)5.51.000.0983.002.00.3.00
Vividness3.10 (0.79)3.88 (1.08)7.50.750.2032.804.00.2.50
Bizarreness1.58 (1.01)2.27 (0.95)8.00.700.2322.751.00.1.00
Emotional valence2.75 (0.25)2.81 (0.22)13.00.190.7202.753.00.2.50
Proportion of self-references0.84 (0.29)0.90 (0.19)15.00.001.0001.001.00.0.50
Content characterizationb
Internal (episodic) details4.08 (1.47)9.13 (3.56)2.01.540.0285.753.00.3.50
External (semantic/other) details0.17 (0.29)1.07 (1.60)8.50.640.2580.000.00.0.50
Content score5.75 (0.90)13.10 (6.49)1.01.740.0186.755.00.5.50
  1. M = mean; SD = standard deviation; ES = effect size; HPC = hippocampal-damaged patients; CTL = control participants; NREM = non-rapid eye movement sleep; REM = rapid eye movement sleep; HPC1−4 = each individual hippocampal-damaged patient. P-values relate to between-group non-parametric Mann-Whitney U tests with significant differences depicted in bold. aAll patients included; means are per awakening. bHCP3, who had no dream reports at all, was excluded; means are per dream report. cFor HCP1, during 20% of his awakenings towards the end of the night, the EEG cap stopped functioning and so designation to NREM or REM sleep was not possible. Hence, it could be that this zero score for REM awakenings is an underestimate, given that REM is more common in the latter part of the night. Note that his dream reports from these awakenings were still included in the dream qualitative attributes and content analyses. See Table 2—source data 1 file for the data underpinning this table.

The number of awakenings was not different between patients and control participants (Figure 1B). Furthermore, there were no significant group differences in the proportion of awakenings from NREM and REM sleep (Figure 1C). After an awakening, participants were instructed via a two-way intercom to describe everything that was going through their mind before they were woken up in that moment. They were occasionally probed (e.g. Can you tell me more about that?) to obtain further information (Materials and methods). The amount of probing did not differ between groups. Dream reports were subsequently transcribed for further analyses (Materials and methods; Figure 2).

Example dream reports.

Experimenter probing is shown in italics. HPC1−4 = the four hippocampal-damaged patients; CTL = an example control participant.

Although the two groups were woken up a similar number of times, they differed in terms of dream frequency, with patients reporting significantly fewer dreams compared to control participants (Figure 1D). There were no group differences in terms of the proportion of dream recall during NREM and REM sleep. Of note, one patient did not report any dreams at all, but instead typically stated ‘I can’t picture it’ (Figure 2).

Perhaps the patients reported fewer dreams than control participants simply because they forgot any dreams they may have had. Three different types of awakening were evident: when participants reported a dream, when they stated they did not dream at all (no dream), and when they dreamt but could not recall the content (this is known as a blank dream). Although there was a significant difference between patients and controls for dreams (see above) and for no dreams, they did not differ in terms of the proportion of blank dreams, which was low for both groups. This suggests that patients could distinguish between situations when they did not dream and those when they dreamt but could not remember. However, no validated objective measure of dreaming exists, and this should be borne in mind when interpreting participants' subjective reports. It is notable that in previous studies, these particular patients could retain information over several minutes, including reporting on their daydreaming (e.g. McCormick et al., 2016; McCormick et al., 2018), which speaks against a rapid decay of sleep mentation as an explanation for their reduced dream frequency.

The patients had so few dream reports that comparisons with the control participants should be treated with caution. Nevertheless, we wondered whether any differences in features and content were evident between the groups for the few dreams the patients reported. Because one patient had no dreams whatsoever, he was not included in these analyses.

We first examined the number of informative words (see Materials and methods; Stickgold et al., 2001b) used in the dream narratives and, while the patients used fewer such words, overall there was no significant difference between the groups. Nevertheless, adjudicating between the possibility of patients having a generic problem with expressing themselves verbally versus merely having little to describe because the few dreams they had were so impoverished, is challenging. The same issue pertains for assessments during wake. A number of studies addressed this concern during tasks involving the imagination of scenes or future scenarios, counterfactual thinking, describing the present and pictures of scenes. Bilateral hippocampal damage does not affect narrative construction or verbal descriptive ability (e.g. Race et al., 2011; Race et al., 2013; Mullally et al., 2012; Mullally and Maguire, 2014; Miller et al., 2020). Considering specifically the patients in the current study, they too had no difficulty performing verbal fluency tests or other tasks where total word count was measured (Supplementary file 1). Given these general and specific findings, it is unlikely that the patients’ performance was driven by an underlying expressive verbal problem.

Next we assessed overall qualitative features of dream reports using experimenter ratings (Materials and methods; De Gennaro et al., 2011; Oudiette et al., 2012). There were no significant group differences in terms of general complexity, vividness, bizarreness, emotional valence (there were no nightmares), and proportion of self-references. In order to probe the dream reports in more depth, we used two scoring methods that are often employed for examining complex mental events (Materials and methods). The first involved a scoring regime typically used for autobiographical memories, the Autobiographical Interview (Levine et al., 2002). This allowed us to measure the amount of episodic (internal) and non-episodic (external) details. The patients included significantly fewer internal details in their dream narratives relative to controls, whereas there was no significant group difference in terms of external details. A second method, usually employed for scoring the content of imagined scenes (Hassabis et al., 2007), showed that the dream reports of patients were significantly less rich in content compared to those of the control participants (Figure 1E).

Discussion

By studying rare patients with selective bilateral hippocampal damage we found that dream frequency was reduced compared to control participants, and the few dreams they had were less episodic-like in nature and lacked content. This accords with previous studies that reported stereotyped dreaming in patients with brain damage that extended beyond the hippocampi (Torda, 1969a; Torda, 1969b; Stickgold et al., 2000), and echoes the effects of hippocampal damage on imagination (Hassabis et al., 2007), episodic memory (Miller et al., 2017; Miller et al., 2020; Spiers et al., 2001) and daydreaming (McCormick et al., 2018). Given our patients’ circumscribed lesions, these results suggest that hippocampal integrity may be necessary for typical dreaming to occur. There are several possible explanations for degraded dreaming in the patients, which we consider in turn.

Despite the tendency to confine dreaming to the sleep state, studies have shown that there is often continuity between waking life experiences and dreaming (Andrillon et al., 2015; Fosse et al., 2003; Horikawa et al., 2013; Schredl and Hofmann, 2003). Given that the patients had difficulty retaining information over longer time scales while awake, then perhaps during subsequent sleep there was little material to process. Hence it could be that the capacity to dream was intact, but underused, in the patients.

Alternatively, the core capacity for dreaming might have been compromised, with this in turn affecting memory processing. Dreaming has been linked with the consolidation of information into long-term memory (Payne, 2010; Wamsley, 2013; Wamsley, 2014; Wamsley et al., 2010; Wamsley and Stickgold, 2019), but it is unknown whether dreaming plays a functional role in this process (Wamsley, 2014). If it does, then patients with hippocampal damage may lack this dream-related mechanism for facilitating memory processing, and this could contribute to their amnesia.

Another aspect of sleep that has been associated with episodic memory consolidation is slow-wave sleep (SWS), a stage within NREM sleep (Rasch and Born, 2013). We have previously shown that the hippocampal-damaged patients tested here had significantly reduced SWS and slow wave activity, whereas the time spent in other sleep stages was comparable to that of control participants (Spanò et al., 2020). It is unlikely that the patients’ degraded dreaming was caused solely by decreased SWS. This is because we sampled dreaming across both NREM and REM sleep, and of the dreams sampled during NREM sleep in the control participants, just 5% (SD 11.3) were during SWS. Overall, however, reduced slow wave activity along with sub-optimal dreaming – if dreaming plays a functional role – may constitute a twofold blow that adversely affected the proper functioning of episodic memory in these patients.

A different perspective on the current results involves looking beyond memory. The hippocampus has been implicated in a range of other cognitive functions including thinking about the future (Addis et al., 2007; Hassabis et al., 2007; Kurczek et al., 2015), spatial navigation (Maguire et al., 2006; O'Keefe and Nadel, 1978), daydreaming (Karapanagiotidis et al., 2017; McCormick et al., 2018), aspects of decision making (Mullally and Maguire, 2014; McCormick et al., 2016), and visual perception (Lee et al., 2005; McCormick et al., 2017; Mullally et al., 2012). Patients with bilateral hippocampal damage are also impaired at mentally constructing scene imagery (Hassabis et al., 2007), and scene imagery features prominently across cognition (Clark et al., 2019; Clark et al., 2020). Consequently, it has been proposed that one role of the hippocampus could be to facilitate the scene construction process (Hassabis and Maguire, 2007; Maguire and Mullally, 2013). This may explain why dreaming is degraded in the context of hippocampal pathology, as it too typically involves (and possibly requires) scene imagery.

In summary, while the functional role of dreaming is as yet unknown, we conclude from our results that hippocampal integrity may be a prerequisite for typical dreaming to occur, and that dreaming seems to align with an array of other cognitive functions that are hippocampal-dependent and which play crucial roles in supporting our everyday mental life.

Materials and methods

Participants

For all patients, hippocampal lesions resulted from leucine-rich glycine-inactivate-1 antibody-complex limbic encephalitis (LGI1-antibody-complex LE; Miller et al., 2017; Miller et al., 2020). This study was conducted a median of 9.5 years after hippocampal damage occurred (mean 9 years ± SD 2.45). Patients (HPC) and the dream control participants (CTL) were closely matched on a number of demographic factors: gender (all males), age (Table 1; MWU = 19.00, p=0.89, Cohen’s d = 0.08), body mass index (HPC mean 27.68 ± 2.51; CTL 25.79 ± 2.41; MWU = 14.00, p=0.40, Cohen’s d = 0.47) and general cognitive ability assessed with the Matrix Reasoning subtest of the Wechsler Abbreviated Scale of Intelligence (WASI; Wechsler, 1999; Table 1; MWU = 7.00, p=0.06, Cohen’s d = 1.13). Participants had no history of psychiatric disorders (e.g., depression, anxiety). Each participant gave written informed consent for participation in the study, for data analysis and for publication of the study results. ‘Materials and methods’ were approved by the University College London Research Ethics Committee.

The patients entered the sleep study having already been characterized, relative to matched healthy control participants, in terms of their lesion selectivity and neuropsychological profile as part of previous research studies. Full details of that characterization process are available in McCormick et al., 2016, McCormick et al., 2017, McCormick et al., 2018 and Spanò et al., 2020. In summary, manual (blinded) segmentation of the hippocampi from T2-weighted high resolution structural MRI scans (0.5 × 0.5 × 0.5 mm voxels) showed that our patients had substantial volume loss relative to controls in the left (MWU = 2.00, p=0.009, Cohen’s d = 1.83) and right (MWU = 3.00, p=0.013, Cohen’s d = 1.67) hippocampus (Table 1). Expert neuroradiological examination confirmed there was no damage outside of the hippocampi. In addition, automated whole brain voxel-based morphometry showed there were no volume differences between patients and controls anywhere else in the brain. Supplementary file 1 provides the neuropsychological profile (summary data and statistical analyses) of the patients across a range of cognitive tests, and indicates the selective nature of their memory loss.

Since all patients included in the current study had suffered from LGI1-antibody-complex LE, our findings might potentially not generalize to other forms of hippocampal amnesia. However, it is important to note that other aetiologies that lead to hippocampal-mediated amnesia such as viral encephalitis, hypoxic brain injury secondary to drug overdose, or toxic shock syndrome are associated with circumscribed hippocampal lesions, but frequently also involve anatomical damage elsewhere (Heinz and Rollnik, 2015; Raschilas et al., 2002). In addition, these aetiologies lead to co-morbidities and broader cognitive impairment (Heinz and Rollnik, 2015; Hokkanen and Launes, 2007; McGrath et al., 1997; Peskine et al., 2010; Thakur et al., 2013; Rosene et al., 1982), which were absent from the clinical and neuropsychological profile of the patients reported here. Therefore, the selection of such a rare group of patients with circumscribed hippocampal lesions allowed us to pinpoint the direct role of the hippocampus in dreaming without the interference of potential confounds associated with heterogeneity in aetiology.

Other features associated with LGI1-antibody-complex LE in its initial presentation – such as focal seizures and hyponatraemia related to hypothalamic damage – are also unlikely to explain the effects we observed. Our patients were seizure-free when they were discharged after initial admission, they were not prescribed antiepileptic medication, and none of the patients had seizure recurrence following initial treatment. Thus, unlike in temporal lobe epilepsy, which is associated with ongoing seizures and hippocampal sclerosis (Kapur and Prevett, 2003), our patients enabled us to study effects on dreaming that were not coincidental with, and sequelae of, seizure activity. Moreover, patients were not undergoing treatment for hyponatraemia, which is consistent with published evidence that persistent hyponatraemia is not a characteristic feature of LGI1-antibody-complex LE (Bastiaansen et al., 2017). Therefore, the findings in the current study are unlikely to stem from the above-mentioned potential issues.

As dreaming might be influenced by sleep quality (Schredl, 2009), we confirmed that patients did not differ significantly from control participants on subjective measures of the general quality and pattern of sleep, as well as on objective measures of sleep-related breathing disorders, and sleep-wake patterns across one week (see Supplementary file 2). Specifically, participants completed standardized questionnaires assessing habitual sleep habits over the last month (The Pittsburgh Sleep Quality Index; Buysse et al., 1989), level of daytime sleepiness (The Epworth Sleepiness Scale; Johns, 1991), and chronotype – whether someone is a ‘morning’ or an ‘evening’ type of person (The Morningness-Eveningness Questionnaire; Horne and Ostberg, 1976). Moreover, we assessed the severity of obstructive sleep apnoea with the WatchPAT-200 (Itamar Medical Ltd., Caesarea, Israel), a wrist-worn device that measures the Peripheral Arterial Tone (PAT) signal by means of a plethysmographic based finger-mounted probe. Signals were automatically analysed with the zzzPAT software (version 4.4.64 .p, Itamar Medical Ltd., Caesarea, Israel) to identify respiratory events and sleep states. The outcome measure employed in this study was the PAT apnoea-hypopnea index (AHI), which provides the number of apnoea and hypopnea events per hour during the night.

In order to assess sleep-wake patterns, participants wore an Actiwatch 2 (Phillips Respironics Mini-Mitter) for seven consecutive days and nights on their non-dominant wrist. Light and activity data were collected in 30 s epochs and analyzed using the Philips Actiware 6.0.2 software package (Respironics Actiware 6.0.2). Data were scored based on available guidelines (Chow et al., 2016), with a medium sensitivity (40 activity cpm), with sleep onset occurring after an immobility period of 10 min, and rise time following an increase in activity level and in light level above 1.0 μW/cm2. Variables of interest were sleep efficiency (in percent), total sleep time (in minutes), sleep fragmentation index (an index of restlessness), night-to-night variability for sleep duration (Lemola et al., 2013), average bedtime and mean sleep midpoint (clock time halfway between bedtime and rise time).

Procedure

Request a detailed protocol

Participants slept in the same room on both nights. Two sleep researchers were located in a separate, nearby, room one of whom performed sleep staging in real time during online visualization of noise-reduced EEG recordings. An independent, registered sleep technologist, blind to participant group membership and the study aims, later off-line scored the PSG recordings to verify the sleep staging, in line with the revised American Academy of Sleep Medicine manual (Berry et al., 2015).

For all participants, awakenings occurred after sleep onset, throughout the night. We aimed to assess dreams from both NREM and REM sleep, and we therefore staggered awakenings at intervals that allowed for entry into these stages, or approximately between 30–90 min intervals. For example, a participant may have been woken at 30 min, returned to sleep and was woken again at 90 min. A maximum of 10 awakenings were scheduled per night. Awakenings were not collected at pre-ordained points during the night (for example, after the third REM period for all participants), rather they were based on participants’ specific sleep architecture in order to maximize the number of reports collected. Every participant had a different sleep onset and duration, and so the awakenings were not scheduled at precisely the same clock time across participants. However, as shown in Supplementary file 2, the sleep quality of the patients and the controls was well-matched, and this included total sleep time, bedtime and midpoint of the night.

Once a decision to awaken a participant was made, after a 3 min period without stage shift of either NREM or REM sleep, the other sleep researcher played a non-stressful 500 Hz neutral tone via a two-way, Bluetooth intercom system equipped with a camera for continuous visual monitoring. After the tone was played, if the participant did not wake up, his name was spoken. This two-step awakening procedure was repeated up to five times, if required (Dumel et al., 2015). After awakening, participants were instructed by intercom to tell the experimenter everything that was going through their mind before they were woken up. What they said was recorded and subsequently transcribed.

Participants were occasionally probed (e.g. Can you tell me more about that?) to obtain further information. Probing followed a structured protocol. This involved first asking participants to freely describe what was in their minds immediately after awakening. Whenever a participant’s response was not clear or only covered parts of the dream, the experimenter asked general follow-up questions, which could echo information already provided (e.g. ‘…It was a conversation happening in the locker room’. Can you give us any more specifics about the conversations or anything else that you recall?). Crucially, this probing never involved leading the participant, as can be observed in the examples provided in Figure 2. This approach is very similar to that of well-established tasks that assess autobiographical memory recall (Levine et al., 2002) and scene imagination ability (Hassabis et al., 2007) during wake, where probing in this manner is widely accepted.

Dream analyses

Request a detailed protocol

Transcriptions of dream reports were analysed by a researcher who was blind to participant group membership. Double scoring was performed on 20% of the data by a second researcher. We assessed across-experimenter agreement with inter-class correlation coefficients, with a two-way random effects model looking for absolute agreement, which indicated excellent agreement between the experimenters' scoring (range: 0.9–1.0).

A dream was defined as any report that included at least one person, one place or one event (Foulkes and Rechtschaffen, 1964). Dream frequency was calculated as the total number of dreams divided by the total number of awakenings. Word count included words that provided information about the dream (i.e., informative words), and excluded repetitions, hesitations and fillers, secondary elaborations and metacognitive statements (Stickgold et al., 2001b). Dream complexity was experimenter scored on a 6-point scale using an adaptation of the Orlinsky score (Oudiette et al., 2012); excluding the no dream and blank dream options, which we recorded separately). This ranged from a participant remembering a specific topic, but in isolation, for example a fragmentary action, scene, object, word, or idea unrelated to anything else, to a participant remembering an extremely long and detailed dream sequence of five or more stages. Vividness referred to the clarity and detail of a dream, and was experimenter scored using the 6-point scale of De Gennaro et al., 2011, ranging from no image at all, to perfectly clear and as vivid as normal vision. Bizarreness/implausibility of dreams was experimenter scored using the De Gennaro et al., 2011 6-point scale. This was based on the presence of bizarre elements (impossible characters or actions) and/or improbable plot (discontinuity or unusual settings). The emotional valence of a dream was scored using a 5-point Likert scale ranging from very negative to very positive, with three indicating a neutral tone. For self-references, one point was awarded per dream report if a participant reported he was the agent of an action, thought or feeling (e.g. ‘I was driving my car down a nearby road with a friend...”).

We used two other scoring methods that are often employed for examining complex mental events in order to probe the dream reports further. The first was the Autobiographical Interview which identifies internal and external details, a distinction that can be conceptualized as the difference between episodic and non-episodic/semantic information, respectively (Levine et al., 2002). Internal details refer to the main event described by the participant and comprises the subcomponents: Event (happenings, individuals present, weather conditions, physical/emotional actions, reactions in others); Time (time of the day, year, season, month, day of the week, hours); Place (location of the event); Perceptual (auditory, olfactory, tactile, visual and visual details, body position, duration of time); Thought/emotion (emotional states, thoughts, implications). The internal details score is the sum of these subcomponents. Anything tangential to the main dream event was scored as external details, including Event (details from other events outside of the dream); Semantic (general, ongoing, extended knowledge/event/state of being); Repetition; Other (metacognitive statements, editorializing). The external details score is the sum of these subcomponents.

A second method we employed, the Scene Construction Test (Hassabis et al., 2007), is usually used for scoring the content of imagined scenes. Here we focused on the content score, which comprises four subcomponents: Entities Present (objects or people); Spatial References (places or spatial relationships between entities); Sensory Descriptions (details that describe an entity); Thought/emotion/action (thoughts, emotional states, action descriptions). The content score is the sum of these subcomponents.

Sleep physiology

Request a detailed protocol

All participants underwent PSG in their homes using a Brain Products system (GmbH, Gilching, Germany). The purpose of the PSG recording was to ensure that we awakened participants during both NREM and REM sleep, and in a similar manner for the patient and control groups. Two trained sleep researchers arrived at a participant’s home approximately three hours before the usual bedtime to set up for the PSG. Equipment was then removed by a researcher the following morning upon awakening. PSG was recorded using a 24-electrode cap (EasyCap; based on the international 10–20 system) including the following EEG channels: Fp1, Fp2, F3, F4, C3, C4, P3, P4, O1, O2, F7, F8, T7, T8, P7, P8, Fz, Cz, Pz, Oz, FT9, FT10 referenced to average mastoids (M1 and M2) (sampling rate = 500 Hz). This montage also included two bipolar electrooculogram channels (EOG), two electromyogram channels (EMG) and two electrocardiogram channels (ECG). Sleep staging was performed based on EOG, EMG and the following derivations: F3/M2, F4/M1, C3/M2, C4/M1, O1/M2, O2/M1.

Statistical analyses

Request a detailed protocol

All statistical analyses were performed with SPSS 25.0 (IBM Corporation). Given that the data did not meet the assumptions of normality and homogeneity necessary for parametric statistics, between-group analyses were performed using non-parametric Mann-Whitney U tests. We also calculated the effect sizes using non-parametric Cohen’s d for all tests performed. In all analyses, the significance level was set at 0.05.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
    The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications, Version 2.2
    1. RB Berry
    2. R Brooks
    3. CE Gamaldo
    4. SM Harding
    5. RM Lloyd
    6. CL Marcus
    7. BV Vaughn
    (2015)
    Darien, Illinois: Academy, American of Sleep Medicine.
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
    Dream reports from different stages of sleep
    1. WD Foulkes
    (1962)
    The Journal of Abnormal and Social Psychology 65:14–25.
    https://doi.org/10.1037/h0040431
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
    A self-assessment questionnaire to determine morningness-eveningness in human circadian rhythms
    1. JA Horne
    2. O Ostberg
    (1976)
    International Journal of Chronobiology 4:97–110.
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
    The Hippocampus: a manifesto for change
    1. EA Maguire
    2. SL Mullally
    (2013)
    Journal of Experimental Psychology: General 142:1180–1189.
    https://doi.org/10.1037/a0033650
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
    The Neuropsychology of Dreams
    1. M Solms
    (1997)
    New York: Psychology Press.
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
    Wechsler Abbreviated Scale of Intelligence
    1. D Wechsler
    (1999)
    New York, NY: The Psychological Corporation, Harcourt Brace.
  76. 76

Decision letter

  1. Laura L Colgin
    Senior Editor; University of Texas at Austin, United States
  2. Muireann Irish
    Reviewing Editor; University of Sydney, Australia
  3. Jess Payne
    Reviewer

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This is a timely and elegant study exploring dreaming in patients with bilateral hippocampal damage due to LGI1-antibody-complex LE. Studies like this have a strong potential to inform us about the neural basis of dreaming, which remains largely unknown.

Decision letter after peer review:

Thank you for submitting your article "Dreaming with hippocampal damage" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Laura Colgin as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Jess Payne (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission. In recognition of the fact that revisions may take longer than the two months we typically allow, until the research enterprise restarts in full, we will give authors as much time as they need to submit revised manuscripts.

Summary:

This is a timely and interesting study from the Maguire lab exploring the role of the hippocampus in dreaming. Four patients with bilateral hippocampal damage, due to LGI1-antibody-complex LE were tested at home using a tightly controlled forced awakening protocol informed by PSG. These patients are extremely rare, presenting with selective memory disturbances due to circumscribed hippocampal insult, which adds to the novelty of the study. The authors used an elegant experimental design including a habituation protocol and the use of PSG to guide awakenings in participants. Interestingly, the current study findings are in line with older papers by Torda from the 1960s emphasising dream abnormalities in amnesia, as opposed to more recent investigations which characterise dreaming as largely intact in such patients (Solms, Stickgold). While the reviewers were all enthusiastic about this study, a number of methodological concerns were nevertheless raised.

Essential revisions:

1) It was suggested that the novelty of the study was somewhat oversold given a number of studies that have explored dreaming in patients with extensive hippocampal damage. It is argued that these studies are flawed for several reasons, including because damage actually extended beyond the hippocampus and patients had broader cognitive impairments. This may be the case, but the clear presence of dreaming in these patients (e.g. Torda, 1969 studies; Stickgold et al., 2000) still shows that dreaming can persist in the face of bilateral hippocampal lesions. The current study is therefore not the first to test whether dreaming requires the hippocampus and such statements should be tempered accordingly.

2) Information on the timing of dream awakenings should be expanded in the Materials and methods. It is stated that participants were awoken after a solid 3min period of REM or NREM sleep, but how, specifically, was the timing and number of awakenings determined? How often were they awoken, and at what time of night? Were awakenings stopped when some maximum number was reached? Were REM and NREM reports collected at equivalent times of night? Was the clock time of awakenings equivalent between patients and controls? Further details would be helpful to address these comments.

3) One methodological weakness was the verbal interaction between the researchers and patients during dream reporting. Subtle differences in how questions are posed have a large effect on the quantity and quality of dream reports. Thus, it is ideal to use pre-recorded verbal prompts for all participants, avoiding systematic bias in the way that dream reports are elicited from patients vs from controls. In this regard, it could be a problem that participants "were occasionally prompted (e.g. Can you tell me more about that?) to obtain further information". While the number of prompts did not differ significantly between control and patient groups, the study is underpowered to detect all but the largest differences here. Numerically, control participants were prompted for more information more often than the patients. There also could have been subtle differences in the quality of interaction between researchers and participants that affected reporting.

4) It is reasoned that "If patients reported fewer dreams due to an inability to recollect what had just been in their mind, the two groups should differ in terms of the proportion of blank dreams". There exists no validated measure of dreaming other than participants' subjective report, and there is no evidence that participants can meaningfully distinguish between having had a dream that they cannot recall and having not had a dream at all. It is therefore of questionable validity to presume that "blank dreams" indicate that dreaming occurred and was forgotten, whereas "no dream" reports indicate lack of dreaming.

5) The very small sample of patients leads to a high risk of both Type II and Type I error. While a large sample is not feasible with this rare and difficult-to-test population, strong conclusions based on non-significant differences between groups should be avoided. For example, it is unclear whether patients are not just describing their dream experiences using fewer words than controls. Importantly, some of the methods used to score richness of detail (for example that used in Hassabis et al., 2007) are based on the number of words used – So differences in so-called "episodic richness" could be in part attributed to any differences in the mere ability of patients to describe their experience, rather than the production of dreaming itself.

6) A related point is whether it would be more informative to present the data as a case series rather than aggregating data from 4 extremely rare (and in some cases only 3) patients as a group. The figures currently present the group means, but individual data points should be included to enable the reader to see the spread of scores across patient and Control groups. A case series would enable the authors to further compare and contrast the case who had no dream reports and to potentially understand the heterogeneity across these patients. It may be that aggregating the dreaming performance across these patients masks important individual differences, as for some measures Controls appear to outperform patients by almost double (e.g., informative content) yet this is not statistically significant at the group level.

7) The authors argue that "if dreaming is unaffected by HC lesions, this would suggest that complex, associative, imagery-rich, spatio-temporal mental events can occur without hippocampal input, requiring a fundamental re-consideration of prominent theories of hippocampal function". It was suggested that the authors should be a bit more cautious in their language given that most HC output is believed to be blocked during REM sleep (which is where most imagery rich dreaming occurs). In fact, given that REM sleep dreaming may be mostly cortical (temporarily devoid of HC input), these results become even more interesting, because the REM sleep dreams obtained here were still degraded compared to controls. These issues may lie beyond the scope of the present study but the authors might find these points relevant for future work.

8) While the proportion of awakenings between NREM and REM were not significantly different between patients and controls, the reviewers were concerned regarding the number of dream reports solicited from REM sleep in general, and especially in the patients. Proportion of awakenings in REM in patients looks to be only about 20%. This is concerning given that REM sleep is where most genuine "dreams" occur (according to standard definitions of dreams as opposed to the mental content that can be obtained from NREM sleep). The authors should comment on this.

https://doi.org/10.7554/eLife.56211.sa1

Author response

Essential revisions:

1) It was suggested that the novelty of the study was somewhat oversold given a number of studies that have explored dreaming in patients with extensive hippocampal damage. It is argued that these studies are flawed for several reasons, including because damage actually extended beyond the hippocampus and patients had broader cognitive impairments. This may be the case, but the clear presence of dreaming in these patients (e.g. Torda, 1969 studies; Stickgold et al., 2000) still shows that dreaming can persist in the face of bilateral hippocampal lesions. The current study is therefore not the first to test whether dreaming requires the hippocampus and such statements should be tempered accordingly.

Apologies, we did not intend to imply that our study was the first to examine dreaming in hippocampal-damaged patients. Rather, we sought to note possible reasons for the disparity among extant findings, and to outline how our study helps to mitigate issues that may have influenced previous results. We have now gone through the entire manuscript and revised it to ensure that our intent is clearer and that the language is appropriately tempered. Examples of how we have done this include:

Abstract: “The hippocampus is linked with both sleep and memory, but there is debate about whether a salient aspect of sleep – dreaming – requires its input.”

Introduction: “While sleep dreaming has been examined previously in patients with hippocampal lesions, results are mixed, with some studies reporting repetitious and stereotyped dreams (Torda, 1969a,b; Stickgold et al., 2000), whereas others claim that hippocampal damage has no effect on dreaming (Solms, 1997; 2013). Possible reasons for these disparate findings include differences in the sleep stages sampled, the inclusion in several studies of patients with psychiatric conditions and below-average IQ and, in other instances, patients were interviewed at a point temporally remote from any dreaming that might have occurred, presenting a challenge for these patients who usually suffer from amnesia.”

“Here, we sought to mitigate the issues affecting previous studies while investigating the frequency, features and content of dreaming in rare patients with selective bilateral hippocampal damage and matched control participants.”

Discussion: “By studying rare patients with selective bilateral hippocampal damage we found that dream frequency was reduced compared to control participants, and the few dreams they had were much less episodic-like in nature and lacked content. This accords with previous studies that reported stereotyped dreaming in patients with brain damage that extended beyond the hippocampi (Torda, 1969a,b; Stickgold et al., 2000)…”

2) Information on the timing of dream awakenings should be expanded in the Materials and methods. It is stated that participants were awoken after a solid 3min period of REM or NREM sleep, but how, specifically, was the timing and number of awakenings determined? How often were they awoken, and at what time of night? Were awakenings stopped when some maximum number was reached? Were REM and NREM reports collected at equivalent times of night? Was the clock time of awakenings equivalent between patients and controls? Further details would be helpful to address these comments.

For all participants, awakenings occurred after sleep onset, throughout the night. We aimed to assess dreams from both NREM and REM sleep, and we therefore staggered awakenings at intervals that allowed for entry into these stages, or approximately between 30-90 minute intervals. For example, a participant may have been woken at 30 minutes, returned to sleep and was woken again at 90 minutes. A maximum of 10 awakenings were scheduled per night. Awakenings were not collected at pre-ordained points during the night (for example, after the third REM period for all participants), rather they were based on participants’ specific sleep architecture in order to maximize the number of reports collected. Every participant had a different sleep onset and duration, and so the awakenings were not scheduled at precisely the same clock time across participants. However, as shown in Supplementary file 2, the sleep quality of the patients and the controls was well-matched, and this included total sleep time, bedtime and midpoint of the night. We now include this additional information in the Materials and methods:

“For all participants, awakenings occurred after sleep onset, throughout the night. We aimed to assess dreams from both NREM and REM sleep, and we therefore staggered awakenings at intervals that allowed for entry into these stages, or approximately between 30-90 minute intervals. For example, a participant may have been woken at 30 minutes, returned to sleep and was woken again at 90 minutes. A maximum of 10 awakenings were scheduled per night. Awakenings were not collected at pre-ordained points during the night (for example, after the third REM period for all participants), rather they were based on participants’ specific sleep architecture in order to maximize the number of reports collected. Every participant had a different sleep onset and duration, and so the awakenings were not scheduled at precisely the same clock time across participants. However, as shown in Supplementary file 2, the sleep quality of the patients and the controls was well-matched, and this included total sleep time, bedtime and midpoint of the night.”

3) One methodological weakness was the verbal interaction between the researchers and patients during dream reporting. Subtle differences in how questions are posed have a large effect on the quantity and quality of dream reports. Thus, it is ideal to use pre-recorded verbal prompts for all participants, avoiding systematic bias in the way that dream reports are elicited from patients vs from controls. In this regard, it could be a problem that participants "were occasionally prompted (e.g. Can you tell me more about that?) to obtain further information". While the number of prompts did not differ significantly between control and patient groups, the study is underpowered to detect all but the largest differences here. Numerically, control participants were prompted for more information more often than the patients. There also could have been subtle differences in the quality of interaction between researchers and participants that affected reporting.

This is an issue into which we put a lot of thought prior to data collection, mindful of the point highlighted by the reviewers. We agree that uncontrolled probing could affect findings. However, we were concerned that pre-recorded probing could also influence results. For instance, the latter could come across as somewhat impersonal, be of limited utility if participants suspect it is automated, and could provoke confusion in people who are being woken from sleep in the middle of the night. To achieve controlled probing that was nevertheless sensitive and useful, the elicitation of dream reports in our study followed a structured protocol. This involved first asking participants to freely describe what was in their minds immediately after awakening. Whenever a participant’s response was not clear or only covered parts of the dream, the experimenter asked general follow-up questions, which could echo information already provided (e.g. “…It was a conversation happening in the locker room”. Can you give us any more specifics about the conversations or anything else that you recall?). Crucially, this probing never involved leading the participant, as can be observed in the examples now provided in new Figure 2.

This approach is very similar to that of well-established tasks that assess autobiographical memory recall (Levine et al., 2002) and scene imagination ability (Hassabis et al., 2007) during wake, where probing in this manner is widely accepted. We, therefore, believe that this parsimonious approach to probing did not have a substantive effect on our findings. We note that the numerical difference in the mean number of probes per awakening was small between groups [HPC 3.82 (1.75); CTL 4.22 (1.13)], as was the effect size (0.23; p=0.671) with no hint that a large difference was masked. We now describe our probing procedure in more detail in the Materials and methods:

“Participants were occasionally probed (e.g. Can you tell me more about that?) to obtain further information. Probing followed a structured protocol. This involved first asking participants to freely describe what was in their minds immediately after awakening. Whenever a participant’s response was not clear or only covered parts of the dream, the experimenter asked general follow-up questions, which could echo information already provided (e.g. “…It was a conversation happening in the locker room”. Can you give us any more specifics about the conversations or anything else that you recall?). Crucially, this probing never involved leading the participant, as can be observed in the examples provided in Figure 2. This approach is very similar to that of well-established tasks that assess autobiographical memory recall (Levine et al., 2002) and scene imagination ability (Hassabis et al., 2007) during wake, where probing in this manner is widely accepted.”

4) It is reasoned that "If patients reported fewer dreams due to an inability to recollect what had just been in their mind, the two groups should differ in terms of the proportion of blank dreams". There exists no validated measure of dreaming other than participants' subjective report, and there is no evidence that participants can meaningfully distinguish between having had a dream that they cannot recall and having not had a dream at all. It is therefore of questionable validity to presume that "blank dreams" indicate that dreaming occurred and was forgotten, whereas "no dream" reports indicate lack of dreaming.

We agree with this point, and in the revised manuscript have now significantly tempered the interpretation of this finding. We still allude to this result and retain the related data in new Table 2 because the distinction between no dreams and blank dreams is common in the literature and will be of interest to some readers.

“Perhaps the patients reported fewer dreams than control participants simply because they forgot any dreams they may have had. Three different types of awakening were evident: when participants reported a dream, when they stated they did not dream at all (no dream), and when they dreamt but could not recall the content (this is known as a blank dream). Although there was a significant difference between patients and controls for dreams (see above) and for no dreams, they did not differ in terms of the proportion of blank dreams, which was low for both groups. This suggests that patients could distinguish between situations when they did not dream and those when they dreamt but could not remember. However, no validated objective measure of dreaming exists, and this should be borne in mind when interpreting participants' subjective reports. It is notable that in previous studies, these particular patients could retain information over several minutes, including reporting on their daydreaming (e.g. McCormick et al., 2016, 2018), which speaks against a rapid decay of sleep mentation as an explanation for their reduced dream frequency.”

5) The very small sample of patients leads to a high risk of both Type II and Type I error. While a large sample is not feasible with this rare and difficult-to-test population, strong conclusions based on non-significant differences between groups should be avoided. For example, it is unclear whether patients are not just describing their dream experiences using fewer words than controls. Importantly, some of the methods used to score richness of detail (for example that used in Hassabis et al., 2007) are based on the number of words used – So differences in so-called "episodic richness" could be in part attributed to any differences in the mere ability of patients to describe their experience, rather than the production of dreaming itself.

We concur with the need for caution in interpreting null results in small samples, and that is why we are happy to implement the reviewers’ excellent suggestions of tempering the overall tone, as well as including individual patient data to facilitate their consideration as a case series (see more on this latter issue in the response to point 6).

The number of informative words per dream report did not differ significantly between the patient and control groups, although there was variability across the board. Nevertheless, adjudicating between the possibility of patients having a generic problem with expressing themselves verbally versus merely having little to describe because the few dreams they had were so impoverished, is challenging. The same issue pertains for assessments during wake. A number of studies addressed this concern during tasks involving the imagination of scenes or future scenarios, counterfactual thinking, describing the present and pictures of scenes. Bilateral hippocampal damage does not affect narrative construction or verbal descriptive ability (e.g. Race et al., 2011, 2013; Mullally et al., 2012; Mullally and Maguire, 2014; Miller et al., 2020). Considering specifically the patients in the current study, they too had no difficulty performing verbal fluency tests or other tasks where total word count was measured (Supplementary file 1). Given these general and specific findings, it is unlikely that the patients’ performance was driven by an underlying expressive verbal problem. We now consider this point in more detail in the revised manuscript:

“We first examined the number of informative words (see Materials and methods; Stickgold et al., 2001b) used in the dream narratives and, while the patients used fewer such words, overall there was no significant difference between the groups. Nevertheless, adjudicating between the possibility of patients having a generic problem with expressing themselves verbally versus merely having little to describe because the few dreams they had were so impoverished, is challenging. The same issue pertains for assessments during wake. A number of studies addressed this concern during tasks involving the imagination of scenes or future scenarios, counterfactual thinking, describing the present and pictures of scenes. Bilateral hippocampal damage does not affect narrative construction or verbal descriptive ability (e.g. Race et al., 2011, 2013; Mullally et al., 2012; Mullally and Maguire, 2014; Miller et al., 2020). Considering specifically the patients in the current study, they too had no difficulty performing verbal fluency tests or other tasks where total word count was measured (Supplementary file 1). Given these general and specific findings, it is unlikely that the patients’ performance was driven by an underlying expressive verbal problem.”

6) A related point is whether it would be more informative to present the data as a case series rather than aggregating data from 4 extremely rare (and in some cases only 3) patients as a group. The Figures currently present the group means, but individual data points should be included to enable the reader to see the spread of scores across patient and Control groups. A case series would enable the authors to further compare and contrast the case who had no dream reports and to potentially understand the heterogeneity across these patients. It may be that aggregating the dreaming performance across these patients masks important individual differences, as for some measures Controls appear to outperform patients by almost double (e.g., informative content) yet this is not statistically significant at the group level.

We have now re-worked the manuscript to accommodate this excellent suggestion as follows:

– Figure 1 has been expanded to constitute an overall summary of the experimental set-up, the number of awakenings, and the main group differences in terms of dream frequency and content. This permits an at-a-glance appreciation of the substantive aspects of the study, and may be useful for readers, particularly those who do not work in the field.

– We now include a new Table 1 which contains individual patient data for a range of variables including hippocampal volume and percent hippocampal volume loss relative to control participants.

– New Table 2 provides all of the dream-related variables as before, but has been expanded to include the individual data for each patient for every measure. We have now also uploaded the related Table 2-Source Data 1 file which contains the individual data for patients and controls. Along with new Table 1, new Table 2 permits interested readers to examine the patients as a series of case studies. We retain the formal between-group comparisons for those readers who will be interested in these analyses, while tempering them appropriately in the text.

– New Figure 2 provides an example dream report from each of the four patients and one from an example healthy control participant to further facilitate the case study approach. We selected those that precluded personal identification of participants, given data protection legislation.

The reviewers suggested we include the data for each patient in case they masked factors that might have influenced the results. Consideration of individual differences in such a small sample of rare patients is not possible. Nevertheless, these data de-emphasise several possible explanations for the results. For example, the findings are unlikely to have been driven by the extent of hippocampal volume loss – HCP3, who had no dreams at all, did not have the greatest hippocampal volume loss (this aligns with the wider literature where evidence of one-to-one mapping between hippocampal volume and function is mixed even among healthy controls). Sampling during specific sleep stages did not seem to have systematically affected performance. For instance, HCP3, with no dreams, had 40% of his awakenings during REM sleep, which should have maximised his opportunity for imagery-rich dreaming. By contrast, HCP1, despite having a low proportion of awakenings during REM had the highest dream frequency among the patients.

Overall, we believe these revisions render the manuscript more informative, whilst retaining clarity. We now preface the Results section as follows:

“Table 2 shows the group summary data and the results of the between-group statistical analyses for each of the measures that are described below. While, for the sake of economy, we present the findings in terms of these group comparisons, given the small sample of these rare patients, caution should be exercised in interpreting the results. We, therefore, also include the individual patient data in Tables 1 and 2 permitting the patients to be considered instead as a series of case studies.”

7) The authors argue that "if dreaming is unaffected by HC lesions, this would suggest that complex, associative, imagery-rich, spatio-temporal mental events can occur without hippocampal input, requiring a fundamental re-consideration of prominent theories of hippocampal function". It was suggested that the authors should be a bit more cautious in their language given that most HC output is believed to be blocked during REM sleep (which is where most imagery rich dreaming occurs). In fact, given that REM sleep dreaming may be mostly cortical (temporarily devoid of HC input), these results become even more interesting, because the REM sleep dreams obtained here were still degraded compared to controls. These issues may lie beyond the scope of the present study but the authors might find these points relevant for future work.

Whilst appreciating the interesting comment, consideration of this issue is beyond the scope of the present work, and would be better pursued using a bespoke experimental design in a future study. The simple point we were trying to make in the cited text is, we hope, relatively uncontroversial. That is, if patients with significantly reduced ability to experience complex, imagery-rich, spatio-temporal mental events during wake are able to have such experiences during sleep, then this would need to be accommodated within extant hippocampal theories, with further implications for understanding the role of cortical regions and hippocampal-cortical interactions in such mental events. We have now re-worded the text to better reflect our point:

“It remains uncertain, therefore, whether hippocampal integrity is necessary for dreaming to occur. If, as we predicted, bilateral hippocampal damage degrades dreaming, this would reinforce the link between dreaming and hippocampal-dependent processes such as memory, potentially moving us closer to an understanding of why we dream. By contrast, if hippocampal-damaged patients who have a reduced ability to experience complex, imagery-rich, spatio-temporal mental events during wake are nevertheless able to have such experiences during sleep, this would need to be explicitly accounted for within theories of hippocampal function.”

8) While the proportion of awakenings between NREM and REM were not significantly different between patients and controls, the reviewers were concerned regarding the number of dream reports solicited from REM sleep in general, and especially in the patients. Proportion of awakenings in REM in patients looks to be only about 20%. This is concerning given that REM sleep is where most genuine "dreams" occur (according to standard definitions of dreams as opposed to the mental content that can be obtained from NREM sleep). The authors should comment on this.

Dreams (and sleep-related visualisations) have been recorded in all sleep stages, even stage 1. REM is the stage in which people report “typically” bizarre or emotional content, but dreams with spatial content are also reported from NREM awakenings. Hence, we felt it was necessary to assess NREM and REM sleep in order to capture dream content that might be affected by hippocampal damage. Also, as noted by the reviewers, there was no significant difference between the patients and control participants in terms of the number of awakenings from NREM and REM. Moreover, and as now detailed on new Table 2 including its legend, the proportion of awakenings during REM in patients may be an underestimate. For HCP1, during 20% of his awakenings towards the end of the night, the EEG cap stopped functioning and so designation to NREM or REM sleep was not possible. Hence, it could be that his zero score for REM awakenings is an underestimate, given that REM is more common in the latter part of the night. If we assume that these awakenings occurred during REM, the patient group mean would then be 28%, which is even more similar to the controls (33%).

As alluded to in the response to point 7 above, the current study was not designed to investigate differences between NREM and REM dreaming, and so our data cannot speak to ongoing debates in that realm. Our goal was to sample the night’s sleep, which typically comprises ~20% REM, in as similar a manner as possible across the patient and control groups.

https://doi.org/10.7554/eLife.56211.sa2

Article and author information

Author details

  1. Goffredina Spanò

    Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Writing - original draft, Data scoring
    Competing interests
    No competing interests declared
  2. Gloria Pizzamiglio

    Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom
    Contribution
    Investigation, Writing - review and editing, Data scoring
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3567-1344
  3. Cornelia McCormick

    Department of Neurodegenerative Diseases and Geriatric Psychiatry, University Hospital Bonn, Bonn, Germany
    Contribution
    Writing - review and editing, Patient characterization
    Competing interests
    No competing interests declared
  4. Ian A Clark

    Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom
    Contribution
    Writing - review and editing, Advice on data scoring
    Competing interests
    No competing interests declared
  5. Sara De Felice

    Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom
    Contribution
    Writing - review and editing, Data scoring
    Competing interests
    No competing interests declared
  6. Thomas D Miller

    Department of Neurology, Royal Free Hospital, London, United Kingdom
    Contribution
    Writing - review and editing, Patient characterization
    Competing interests
    No competing interests declared
  7. Jamie O Edgin

    Department of Psychology, University of Arizona, Tucson, United States
    Contribution
    Writing - review and editing, Advice on conceptualization
    Competing interests
    No competing interests declared
  8. Clive R Rosenthal

    Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom
    Contribution
    Writing - review and editing, Patient characterization
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5960-4648
  9. Eleanor A Maguire

    Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom
    Contribution
    Conceptualization, Formal analysis, Supervision, Funding acquisition, Methodology, Writing - original draft
    For correspondence
    e.maguire@ucl.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9470-6324

Funding

Wellcome (101759/Z/13/Z)

  • Eleanor A Maguire

Wellcome (203147/Z/16/Z)

  • Eleanor A Maguire

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank the participants and their families for welcoming us into their homes and giving of their time so generously. We are also grateful to David Bradbury and Kamlyn Ramkissoon for their assistance.

Ethics

Human subjects: Each participant gave written informed consent for participation in the study, for data analysis and for publication of the study results. 'Materials and methods' were approved by the University College London Research Ethics Committee (reference number: 6743/004).

Senior Editor

  1. Laura L Colgin, University of Texas at Austin, United States

Reviewing Editor

  1. Muireann Irish, University of Sydney, Australia

Reviewer

  1. Jess Payne

Publication history

  1. Received: February 20, 2020
  2. Accepted: May 5, 2020
  3. Version of Record published: June 8, 2020 (version 1)

Copyright

© 2020, Spanò et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,257
    Page views
  • 276
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

  1. Further reading

Further reading

    1. Neuroscience
    Henry WP Dalgleish et al.
    Research Article Updated

    Many theories of brain function propose that activity in sparse subsets of neurons underlies perception and action. To place a lower bound on the amount of neural activity that can be perceived, we used an all-optical approach to drive behaviour with targeted two-photon optogenetic activation of small ensembles of L2/3 pyramidal neurons in mouse barrel cortex while simultaneously recording local network activity with two-photon calcium imaging. By precisely titrating the number of neurons stimulated, we demonstrate that the lower bound for perception of cortical activity is ~14 pyramidal neurons. We find a steep sigmoidal relationship between the number of activated neurons and behaviour, saturating at only ~37 neurons, and show this relationship can shift with learning. Furthermore, activation of ensembles is balanced by inhibition of neighbouring neurons. This surprising perceptual sensitivity in the face of potent network suppression supports the sparse coding hypothesis, and suggests that cortical perception balances a trade-off between minimizing the impact of noise while efficiently detecting relevant signals.

    1. Neuroscience
    Shenghong He et al.
    Research Article Updated

    Previous studies have explored neurofeedback training for Parkinsonian patients to suppress beta oscillations in the subthalamic nucleus (STN). However, its impacts on movements and Parkinsonian tremor are unclear. We developed a neurofeedback paradigm targeting STN beta bursts and investigated whether neurofeedback training could improve motor initiation in Parkinson’s disease compared to passive observation. Our task additionally allowed us to test which endogenous changes in oscillatory STN activities are associated with trial-to-trial motor performance. Neurofeedback training reduced beta synchrony and increased gamma activity within the STN, and reduced beta band coupling between the STN and motor cortex. These changes were accompanied by reduced reaction times in subsequently cued movements. However, in Parkinsonian patients with pre-existing symptoms of tremor, successful volitional beta suppression was associated with an amplification of tremor which correlated with theta band activity in STN local field potentials, suggesting an additional cross-frequency interaction between STN beta and theta activities.