1 Introduction

Stroke and traumatic brain injury combined constitute the second-largest health-related expense in the US [1], with around 1.5 million new victims each year. This is largely due to the chronic disability and loss of vocational abilities experienced by many survivors.

The strongest predictors of return to work and quality of life following an acquired brain injury are found in the cognitive domain, while physiological characteristics such as etiology, location, and lesions size are weaker predictors [2,3,4,5]. Important cognitive sequelae include impairments in the executive domain [6], which are correlated with increased fatigue [7, 8], and have a relatively strong prognostic value for recovery in general [9].

1.1 Expectations of brain injury sequelae

Cognitive sequelae can, in part, be attributed to expectations of injury sequelae formed pre-injury. In other words, the expected sequelae become a self-fulfilling prophecy [10, 11], likely through a mechanism with high similarity—if not identity—to the nocebo effect [12]. Therefore, injury expectation may be an important target in rehabilitation, counselling, and psychotherapy to improve the patients’ functioning, activities, and social participation [13,14,15]. This article reports the potential effects of using a hypnotic procedure to change such negative expectations in the chronic phase following acquired brain injury. Changing expectations of psychological and behavioral outcomes may be integral to how hypnosis works [16, 17]. Since the expectation is the main driver of placebo effects (at least in the mental domain), it is no surprise that hypnosis and placebos are highly similar. Hypnosis may be regarded as systematically exploiting the placebo effect [18]. We have a theoretical framework for explaining these effects that reconciles patient’s reported experiences [19]. In, short, if cognitive sequelae are caused or sustained by negative expectations and nocebo effects, hypnosis could be well suited to counter such effects.

Most speech-based interventions have at least a component of expectation modification through suggestions. We opted to use mindfulness meditation as an active control group because it is one of the most closely related intervention modalities [20,21,22].

1.2 Hypnosis vs. mindfulness meditation

Hypnosis is a broad term referring to widely varying practices. Therefore, it is superficial to analyze identities and discrepancies between hypnosis and mindfulness meditation at a general level [23, 24]. American Psychological Association defines hypnosis as “A state of consciousness involving focused attention and reduced peripheral awareness characterized by an enhanced capacity for response to suggestion”. The suggestion is “suggested alterations in physiology, sensations, emotions, thoughts, or behavior during hypnosis” [25].

As with hypnosis, mindfulness meditation is a broad term that may refer to widely varying practices [26]. However, one standardized treatment program, Mindfulness-Based Stress Reduction (MBSR), defines mindfulness as a psychological state characterized by Focused Attention and Open Monitoring [26, 27]. Attention is focused when it can stay with a particular object or thought for an extended time, with little or no distraction from other objects or thoughts. Open Monitoring is observing the stream of thoughts without judging or acting on them, however disturbing or salient they may be. For brain-injured patients, such thoughts may include unfavorable comparisons between the current state and the pre-morbid state and the concurrent desire to return to pre-morbid functioning.

While both hypnosis and mindfulness typically include a state of focused attention, the difference is in the suggestions provided to the patient: mindfulness meditation typically seeks to align expectations with the actual mental, bodily, and physical circumstances. This is called “insight” or “awareness”. In contrast hypnosis typically seeks to align expectations to the reality proposed by the hypnotist, thereby achieving congruent mental, bodily, and physical effects as a self-fulfilling prophecy. This might be called “strategic self-deception”, without the negative connotations of the term [28]. To illustrate, a typical mindfulness approach to smoking cessation would focus on noticing and understanding the craving for nicotine without being absorbed. In contrast, a typical hypnotic approach would be the suggestion that there is no craving in the first place. In the present experiment, we suggest to brain-injured participants that their brain plastically re-organizes to be similar to the pre-injury brain and that the patient’s abilities and experiences will return to the pre-injury level as well [29].

Returning to the importance of expectations of brain injury sequelae, we speculate that mindfulness typically helps the patient by downplaying the importance of existing expectations, while hypnosis directly changes expectations to the benefit of the patient. In the process, hypnosis may even enhance the role of these “new” expectations. In this way, the expectation is an important mediator for mindfulness and hypnosis alike [23], at a general level, but they typically seek different kinds of expectation modifications.

1.3 Cognitive rehabilitation using hypnosis and mindfulness

The idea that hypnotic suggestion could be used in neurorehabilitation has been suggested multiple times in the academic literature [30,31,32]. Hypnosis has been extensively studied for its effectiveness in pain relief [33,34,35,36]. Hypnosis has also been successfully applied in motor rehabilitation [37,38,39,40], aphasia [41], vertigo, and a range of related sequelae [42, 43]. Indeed, patients with stroke or concussion have been shown to be as hypnotizable as a non-injured reference sample [31, 44].

However, cognitive rehabilitation using hypnosis has received limited attention. Our research adds to this field by investigating the impacts of suggestions and mindfulness meditation providing perspective on their effectiveness, as potential cognitive rehabilitation techniques.

A recent single-blinded randomized controlled trial with 68 acquired brain injury patients found that hypnotic suggestion improved working memory performance with large effect sizes of Cohen’s d = 1.55 and d = 2.03 [29] on two multi-test indices relative to passive controls. Furthermore, the effect did not wane during the follow-up period of 7 weeks. Using an active control group (mindfulness) in addition to passive controls, the study identified the contents of the suggestions as a selective driver of the effect over and above non-specific effects such as expectancy (with medium sized improvements relative to active controls). Importantly, hypnotic suggestion improved working memory to just above the population average, indicating a complete recovery in this respect.

To our knowledge, only a handful of other studies have used parallel-group designs to investigate the effect of hypnotic suggestion on cognitive impairment following acquired brain injury. All included suggestions that a particular harmful consequence of acquired brain injury no longer applies, likely to induce this expectation in the patient.

Milos [45] used hypnotic age regression to alleviate amnesia for the events surrounding automotive traumatic brain injury and succeeded in seven out of 20 patients with severe injuries. However, the reports were not verified on objective outcomes, so it is unclear whether these recollections reflect a genuine alleviation of amnesia or a more liberal response criterion [46, 47].

Sullivan et al. [48] recruited 24 brain-injured patients (of unspecified origin) in the chronic phase with an IQ between 50 and 75 at baseline and stratified them into three groups. The treatment group received a very short 7-sentence anxiety-reducing hypnotic suggestion, including “From now on, you will not feel nervous or afraid while doing things. You will feel very relaxed. You will be able to do jobs better than before” ([48], p. 97). Even from this very brief intervention, Sullivan et al. observed an improvement (Cohen’s d ~ 0.2) on one out of two neuropsychological tests compared to an active and a passive control group.

In a later study, Cui-Ping [49]Footnote 1 conducted individualized hypnotic treatment with 71 stroke patients in the sub-acute phase between 1 and 12 weeks following incidence. The outcome was compared to 49 patients in a passive control group. Cui-Ping observed large positive effects (d ~ 1) on overall functioning, anxiety, and depression compared to the control group.

Although Sullivan et al. [48] and Cui-Ping [49] do lend some support to the findings of Lindeløv et al. [29], they are severely underreported, leaving out crucial information about participant randomization, blinding of testers, statistical procedures, etc. In particular, it is not clear whether Cui-Ping [49] was peer-reviewed, there is no information about the hypnotic suggestions, and it does not report the interaction effect between gains in the treatment group and control group [50, 51]. Hence, more evidence is clearly needed to assess the potential effect (and magnitude of effect) of hypnotic suggestion on patients with acquired brain injury.

Several studies have aimed at (and succeeded in) decreasing fatigue and increasing motivation for physical rehabilitation following acquired brain injury using hypnotic suggestion [52, 53]. As such, improved cognitive functioning need not be an outcome per se, as it can also be used to facilitate other aspects of rehabilitation.

There is convincing evidence from two randomized controlled trials that MBSR can reduce average fatigue following acquired brain injury [54, 55]. Three randomized controlled trials indicate that mindfulness suggestion only has a small or no average effect on cognitive test performance [56,57,58], which has also been found in the general population [26]. As a potential exception, Johansson et al. [54] observed large average positive within-group effects in the treatment group. However, the authors did not conduct the critical interaction test to determine whether these effects exceeded that of the control group [50].

In the following, we build on this literature by reporting new empirical findings on using hypnosis for chronic cognitive sequelae of acquired brain injury. The aims of the present article were to examine effects of targeted hypnotic suggestion on self-defined treatment goals and sleep-related outcomes compared to an active control (non-targeted, mindfulness) in 49 patients with chronic cognitive sequelae following acquired brain injury, and to examine the potential overlap between subjective and objective measures of improvement.

2 Methods

2.1 Participants

We recruited 52 participants who had sustained acquired brain injury at least 1 year prior. Of these, 49 successfully completed the study. Our cohort was diverse, covering a broad spectrum of lesioned brain areas, severities, and time elapsed since the injury. Participants were randomly allocated to the non-targeted (mindfulness) and targeted hypnosis through a coin flip. (See the supplementary material for excerpts from the manuscripts, as well as further details on hypnosis induction).

The method and the secondary and primary outcomes of this randomized actively controlled trial have been reported in Lindeløv et al. [29]. In this paper, we report all quantitative secondary outcomes.

2.2 Interventions

The hypnosis was written in a manuscript as dictated by the hypnotist to reduce experimenter effects. Thus, the treatment was not individualized. The induction and termination were identical for all participants. We administered two hypnosis protocols. Each protocol was four one-hour treatment sessions with new manuscripts, including the induction and the termination.

The “targeted” hypnotherapy focused on suggestions about regaining pre-morbid cognitive abilities. We used various techniques towards this end, including regression to the pre-morbid state, suggestions about extensive brain plasticity during hypnosis, and suggestions about an ongoing experience of ease and automaticity of thought. We call these “targeted” suggestions because they directly target the core problem of improving working memory-related functioning.

The “non-targeted” hypnotherapy was an active control of the “targeted” therapy. Following the standardized hypnotic induction, the “non-targeted” hypnosis borrowed suggestions from Mindfulness meditation, asking participants to focus on staying directed at the present moment and accepting any pleasant and unpleasant thoughts or sensations as they are, without judging them or acting on them. There was no direct mention of the brain injury or working memory-related functioning. However, the termination of the hypnosis did include the post-hypnotic suggestion that patients would awake “refreshed and feel better”. As mentioned in the introduction, Mindfulness meditation has shown small- or no improvement on cognitive tests; thus, the non-targeted group served as an active control to the targeted group. There was a passive control group, which is not reported here because they did not participate in the interviews where self-report measures were collected.

2.3 Procedure

This study is a randomized controlled trial with a parallel-groups phase, including a follow-up phase. This is followed by the second phase of the intervention, where both groups receive the targeted intervention. Participants were assigned to the two groups using a coin toss.

The “targeted-first” group received targeted suggestions in both phase 1 and phase 2. The “targeted-last” group received a non-targeted suggestion in phase 1 and crossed over to receive a targeted suggestion in phase 2. Thus, the targeted-last group serves as an active control to isolate the effect of “targetedness” in phase 1 and after the break. Both groups received four (1 h) sessions of targeted suggestion in phase 2. This phase mainly serves to see dose–response effects in the targeted-first group and address the possibility that unresponsive patients were, by chance, randomized to the targeted-last group. Participants were tested and interviewed before and after each phase.

2.4 Outcomes

Semi-structured interviews were conducted before and after each treatment phase. Here, we report on the results from the closed questions.

Patient Reported Outcomes (PROs): at baseline, patients were asked to self-define a set of PROs; the outcomes were thus self-defined treatment goals by the patients. They were followed up in the subsequent interviews, asking for progress on each PRO using the ordinal scale “much worse”, “worse”, “same” (no progress), “better,” “much better,” and “not a problem anymore” (maximum progress, hereafter called “achieved” for brevity). Rehabilitation science, in general, has recently moved toward a larger focus on PROs. Rather than having the experimenter decide on the important outcomes, PROs let the patients define the outcomes and report their progress towards achieving these outcomes [59].

Sleep: we asked participants to report their daily duration of nightly sleep, daily sleep, and daily rest at each of the four testing sessions. Furthermore, patients filled out the European Brain Injury Questionnaire (EBIQ) [60], and a close relative filled out the relative-rated EBIQ [61].

WAIS-III Working Memory Index [62] is an index score that is computed from the performance on three tasks: digit span (forwards and backward), letter-number sequencing, and mental arithmetic.

Trail Making Test log(B − A) index: patients draw lines between consecutive numbers to test psychomotor speed (form A) and alternate between consecutive numbers and letters to test the additional cost of task switching (form B). The difference in completion times (B − A) operationalizes task-switching ability [63], and the logarithm renders our data normally distributed.

The subjective experience of progress: participants were asked to evaluate their performance in the Working Memory and Trail Making tests (see above), which preceded the interview. They reported their performance relative to the baseline test performance using the ordinal scale “much worse,” “worse,” “same,” “better” or “much better.” This constituted their perceived improvement, while the actual change in test performance constituted the objective improvement. The degree to which the two correspond (their correlation) was taken as an index of metacognitive accuracy with Kendall’s τ = 1 being fully accurate and τ = 0 being no meta-cognitive information. A one-sided Kendall correlation (τ) with a stretched-beta prior width = 0.5 was used for computing p and BF [64], the prior expressing a weak skepticism of strong correlation coefficients given prior literature on this correspondence [65, 66].

Participants were unaware that there were two treatment groups and were thus blinded to this contrast. The neuropsychological tester was blinded to the allocation of each participant. The hypnotist, who was not blinded, collected the sleep and PRO data. We discuss blinding in Sect. 4.1.

3 Results

3.1 Participants

Patient characteristics are presented in Table 1. The patient group was heterogeneous because we had no a priori knowledge or hypothesis that patient characteristics were important for the effect. It has previously been shown that age, hypnotizability, etiology, and duration since injury did not modulate the effect [29].

Table 1 Participant characteristics at baseline
Full size table

Participants were similar in terms of age, gender, years since injury and hypnotizability was measured with the SHSS:C (Stanford Hypnotic Susceptibility Scale: Form C) [29, 67]. Despite having suffered a brain injury, the participant’s hypnotizability of 7.2 was comparable to the general population as has previously been found [44]. As expected, they had reduced working memory functioning with a mean WAIS-III index score of 82.6 compared to the population average of 100.

3.2 Patient reported outcomes

Frequent PROs included regaining the ability to (1) take part in a conversation with multiple people or with one person amidst background chatter, (2) read without quickly tiring, (3) remember new names, phone numbers, or characters in fiction, and (4) do grocery shopping.

The distribution of progress reports can be seen in Fig. 1. The categories “much worse” and “worse” were not used by patients and are therefore omitted from further analysis. There was only anecdotal evidence of a difference in the distributions of progress reports between the targeted-first group and the targeted-last group after phase 1 (X2 = 10.82, df = 3, p = 0.013, BF = 2.8) but the evidence was strong for a difference at follow-up (X2 = 18.25, df = 3, p = 0.00039, BF = 333.6). At the final test, when all patients had received the targeted intervention, the evidence is ambiguous as to whether the distribution of improvements is identical or different between the two groups (X2 = 6.28, df = 3, p = 0.099, BF = 0.58), but a visual inspection shows that any differences are small because they concern adjacent categories rather than a general skew.

Fig. 1
figure 1

Reports of improvements on Patient Reported Outcomes (PROs) relative to baseline. Each panel sums to 100% of the responses given for that group-phase combination. The numbers on each bar are the percentages and counts. There are evident improvements in both groups. Notably, only a few of the PROs maintain a “same” status, and the achievement of the goals occurs only following treatment with the initially targeted suggestion

Full size image

Only after targeted suggestions did the participants begin to report that they had achieved their PROs (after phase 1 in the targeted-first group and after phase 2 in the targeted-last group). The magnitude of the effect following the full 8-session program is substantial, with the achievement of 25.4% of the PROs initially set by the participants, and just 13.6% of the PROs lack progress.

These results indicate that the effect of targeted suggestions differs from the effect of the active control: non-targeted (mindfulness) suggestions. While both improved on treatment goals, the targeted suggestion has better long-term effects and exclusively leads to the achievement of the PROs.

3.3 Sleep and rest

The interviews elicited complete reports from 16 participants in the targeted-first group and 8 participants in the targeted-last group.

Figure 2 shows the individual and summarized sleep- and rest trends. We used the total daily duration of sleep and rest per individual as a crude index. We conducted a Bayesian Wilcoxon Signed Rank test on the change from baseline to the follow-up session. Hypnotic suggestion reduced daily sleep and rest by around an hour in the targeted-first group (Mdn = − 57.5 min) and in the targeted-last group (Mdn = − 52.5 min), p = 0.0008, BF = 30.9 (see priors and posteriors in the Appendix). There was no (strong) evidence that these effects differed (W = 73, p = 0.60, BF = 0.5) as assessed using a Bayesian two-tailed Mann–Whitney U test on change scores [64].

Fig. 2
figure 2

Self-reported hours of nightly sleep, daily sleep, and daily rest for 24 participants where full datasets were available. The rightmost panel is the medians and inter-quartile ranges for total sleep- and rest duration per day. In the other panels, each line represents an individual’s progression on nightly sleep, daily sleep, and daily rest, respectively. A slight vertical jitter of ± 2 min is added to visualize the density of overlapping lines—in particular at zero (no sleep or rest)

Full size image

The self-reported duration of sleep did not correlate with the self-reported “Problems with sleep” item from the European Brain Injury Questionnaire at any of the four testing sessions (Kendall’s τ = 0.09, − 0.14, − 0.12, and − 0.03 for session 1 through 4).

3.4 Perceived and objective test performance

We observed no correlation between perceived and objective changes in performance in either of the two groups at post-test and follow-up (see Table 2). In fact, there was no correlation between perceived and objective changes in performance when collapsing all data from all testing sessions and both groups on the Working Memory Index (τ = − 0.01, p = 0.57, BF = 0.17) nor the Trail Making Test (τ = 0.05, p = 0.28, BF = 0.36), as is also evident in Fig. 3. In other words, the participants had little or no insight into the magnitude of their improvements or deteriorations in their performance on the neuropsychological tests. This is surprising considering that most patients experienced large improvements of z > 1.5 in both tests (see Fig. 3; grid lines represent z scores of 1, 2, 3, …). This is particularly evident in the targeted-first group, where the majority of the participants perceived no change in their testing performance immediately following the targeted-suggestion phase, even though this was the most rapid increase in objective performance at any point in the experiment.

Table 2 Kendall correlations between perceived and objective improvements on two neuropsychological tests at post-test and follow-up relative to baseline performance
Full size table
Fig. 3
figure 3

Changes on perceived performance change (shape) and objective performance change since baseline (symbol location) on the two neuropsychological tests for each phase and group. The dots near the intersection between the solid lines in each panel correspond to no improvement on the objective outcomes while the dots in the upper right correspond to large improvement on both outcomes. The grid lines correspond to standardized mean differences of 0, 1, 2… where differences larger than 0.8 are conventionally labeled as “large” effects (Cohen, 1992). While changes on the two tests are correlated, reflecting a general improvement in working memory performance, categories of perceived improvement are completely overlapping, indicating that the participants had little or no insight into their objective improvement on these tests

Full size image

3.5 One latent improvement?

Above, we found a low validity of perceived improvement on the neuropsychological tests as an index of actual improvement. We have no such objective measures of other outcomes (sleep, PROs, reports by relatives, etc.), but an assessment of convergent validity between effects on all dependent variables can be used to infer the structure of the underlying improvement(s). At one extreme, there is a hypothesis of one latent improvement. An exploratory factor analysis should yield one factor with strong loadings from all dependent variables. At the other extreme, there is no systematicity resulting in weak joint loadings of dependent variables in factor analysis.

To test this, we used a Bayesian exploratory factor analysis [68] on the treatment effects reported in the previous sections: the Working Memory Index, the Trail Making Test, perceived neuropsychological improvement, total sleep and rest duration, and average progress on PROs. In addition to this, we added the self-rated European Brain Injury Questionnaire (EBIQ) core scale [60, 61] and the relative-rated EBIQ core scale. Two hundred out of 1029 data points were missing and imputed using the corresponding conditional (posterior predictive) distribution during MCMC sampling.

A one-factor solution was strongly preferred including just three outcome measures: the Working Memory Index (mean loading = 0.56, 95% CI = [0.29, 0.91]), the Trail Making Test (mean loading = 0.67, 95% CI = [0.35, 0.96]), and the relative-rated EBIQ with a more moderate factor loading (mean loading = 0.26, 95% CI = [0.00, 0.46]). Pairwise correlations between all outcomes are available in the Appendix.

This analysis confirms that the improvements in the Working Memory Index and the Trail Making Test reflected a latent working memory improvement, which also moderately expresses itself in overt behavior as assessed by relatives on the EBIQ. Interestingly, all outcomes reported by the participants themselves failed to reach this convergent validity, including the self-rated EBIQ.

4 Discussion

The presented findings suggest that there is a positive effect of targeted hypnotic suggestion on self-reported real-life goals and fatigue following acquired brain injury. The effects of “targetedness” of the hypnotic suggestion have previously been shown to be instrumental for improvement on neuropsychological tests [29], and we find a similar effect on achieving specific behavioral goals. While we found evidence that hypnosis reduced the need for sleep by around one hour daily, the “targetedness” of the hypnotic suggestions did not enhance or diminish that positive effect notably as sleep improved equally in the “non-targeted” group. Given that most participants reported that these sequelae had been stable for years (an informal multiple-baseline), we consider it plausible that at least part of the improvements can be attributed to the hypnotic procedures, but future studies should determine the merits of this claim, e.g., by comparing with other types of hypnosis or standard insomnia treatment.

The use of hypnosis was motivated by the observation that expectations of acquired brain injury sequelae seem to be important predictors of manifest impairment [10, 11]. Since changing expectations of psychological and behavioral outcomes may be integral to how hypnosis works [16, 17], the use of hypnosis could also be effective in neurorehabilitation. In a recent theoretical framework, we deliberate on the mechanisms of change in hypnosis for cognitive rehabilitation arguing that hypnosis is a powerful way to attenuate negative self-expectancies induced by the brain injury [19]. Our framework aligns with recent appeals for the use of hypnosis in neurology [14, 69]. The current study is not designed to evaluate the credibility of this motivation, but one result does lend support to it: participants erred on the side of being too pessimistic about their objective improvements, which may be a consequence of a “negative surprise” if they expected to be able to perform even better. This potential mechanism remains to be tested directly. If true, this would indicate that the positive effects on sleep in both treatment groups could be due to common factors in the two treatments, e.g., relaxation and reduced worrying. We argue that there are several methodological issues with previous studies reporting positive effects of hypnotic suggestions for cognitive rehabilitation (e.g., [45, 48, 49]) More peripheral evidence from Sapp’s [70] successful anxiety-reduction in stroke- and TBI patients, as well as the major reduction of vertigo in concussed patients by Cedercreutz et al. [43], adds support to the hypothesis that hypnotic suggestion can be effective following acquired brain injury [43, 70]. Our study result lends further support to the idea that hypnosis, but also mindfulness-based interventions can affect sleep and the need for rest which we hypothesize to be mediated through its effect on emotional distress.

Our research outcomes provide additional evidence supporting the effectiveness of hypnosis in the neurorehabilitation of acquired brain injury. The positive findings on standardized neuropsychological measures of working memory reported in our companion paper [29] dovetail with the improvement in subjective measures of the PROs and their reported reductions in the need for sleep and rest reported in the present paper. Thus, our research adds to the growing body of evidence in favor of using hypnosis for cognitive sequelae following acquired brain injury. Hypnosis is being extensively researched for pain relief [33,34,35,36] and has recently seen a positive resurgent application within motor rehabilitation [37,38,39,40]. A similar resurgence has been seen for the application in the cognitive neurorehabilitation domain in recent years [29, 71]. However, the method has not been adopted into neurological rehabilitation practice [72], and has not been incorporated into meta-analytic comparisons [73, 74] despite acknowledging the promising findings of Lindeløv et al. [29]. The absence may be attributed to the scarcity of multiple high-quality RCTs performed by independent research groups.

In line with existing literature on the placebo effect and negative expectations following injury [10, 11] our approach focuses on the impact of expectations on outcomes and rehabilitation. By utilizing hypnosis to alter these expectations our results align with theories suggesting that hypnosis can be used to harness the placebo effect to the benefit of the patient [18]. We explore this hypothesis extensively in our framework for understanding these effects [29]. The neuropsychological test results from the present randomized actively controlled trial [29] found large effect sizes while addressing many of the methodological shortcomings in the previous literature. One of such shortcomings is the comparison to an adequate active-control condition. Here, we argue that an intervention that contains the same initial hypnotic induction but diverges during the main hypnosis by borrowing from a mindfulness tradition rather than reshaping specific expectations for enhancing working memory provides an adequate control condition.

Treatment effects of the magnitudes observed in the present article and in prior research are occasionally observed in the literature on cognitive rehabilitation. The consistency observed across studies on hypnotic suggestion (for pain and motor rehabilitation) has far-reaching implications for treatments in the neurological domain and warrants further investigation. Taken at face value, the close-to-zero correlation between objective performance and perceived performance implies that the participant’s reports are questionable as a measure of manifest improvement. This could imply that hypnosis, in general, may induce a lack of self-insight. However, we have since learned that poor self-insight is a general feature of healthy [75, 76] as well as brain-injured patient populations [65, 66] with similar correlation coefficients. Thus, we argue that the poor correspondence should not be attributed to hypnosis. The weak correlation we observe could also imply that the results from the other subjective outcome measures are invalid. However, a meta-analysis by Zell et al. [76] showed that the correlation is stronger for performances closer to activities of daily living (r ~ 0.3) than for mental abilities (r ~ 0.15), so the reports on sleep and PROs may have higher validity than the reports on cognitive performance [76]. One interpretation of this could be that metacognitive evaluations of first-order cognitive abilities are not a direct “read-out” of actual capabilities, but rather a sum of the observed outcome of behavior over a longer period. Thus, a sudden increase in function may not be captured by a metacognitive system. Much evidence suggests that introspection can be trusted as an indication of the current content of conscious experience, yet not as an index of the current functional state of a cognitive system [77,78,79]. Future studies should incorporate a variety of different measures to understand the relationship between cognitive improvement and metacognitive appreciation—including verification using relatives’ (subjective) reports as well.

We found evidence that self-reported effects on sleep and rest, PROs, and symptoms are not related (see also Supplementary Figure S1 in the Appendix). That is, there is no tendency across participants in how each of these improvements relates to other improvements, with the exception that there is a correlation between “outside” observations from relatives and neuropsychological tests. This lack of a general latent improvement factor suggests that individuals with acquired brain injury have very different subjective experiences of the treatment effects. The practical implication of this, if true, is that we currently cannot predict how each patient will perceive the effects, other than that the experiences will generally be positive. The self-reported performance on the neuropsychological tests systematically erred on the side of being too pessimistic, so patients may tend to report smaller effects than objectively true.

We believe that these observations highlight a need for observable outcome measures when assessing the effect of hypnosis—and possibly of (mindfulness) meditation—following acquired brain injury.

4.1 Limitations

The evidence presented here is primarily limited by small sample size. Future studies could additionally improve on the present work by differentiating sleep quality and sleep duration and should include objective measures of functioning and participation in society, e.g., return-to-work, caregiver burden, and observations of functional independence.

Lindeløv et al. [29] tested a number of sources of individual differences, including age at injury, current age, time since injury, injury type, baseline cognitive performance, and hypnotizability, but neither of these predicted differences in improvement in the working memory outcomes [29]. Further studies should attempt to balance patients on these prognostic variables.

The demand characteristics in a clinical trial generally encourage participants in the treatment groups to report improvements, though a review found that participants are not biased by this demand [80]. If they were, however, this would induce positive correlations between outcomes, which would result in one latent factor spanning all interview data. This was not observed, so the effect of any demand characteristics is likely to be small in the present study. The factor analysis should however be replicated in a larger sample, and the results from the factor analysis can be considered preliminary. Moreover, given that these are self-chosen outcome measures, the effects are scattered among different self-defined issues rather than a singular outcome. Further studies should explore functioning in everyday life using objective measures and possible mechanisms of change [71].

The fact that the interviewer was not blinded may have inflated the effect sizes. The lack of blinding in randomized controlled trials generally causes a small-to-medium effect size inflation of around d = 0.3 in general patient populations [81] and for brain-injured patients [73, 82]. The observed effects of hypnotic suggestion go beyond the effect of blinding and were also observed by the completely blinded tester.

5 Conclusion

We believe the present study strengthens the case for using hypnotic suggestion as an experimental adjunct to existing treatment for cognitive sequelae following acquired brain injury. More evidence is needed before deciding whether hypnotic suggestions could be implemented as a standard treatment or replace existing treatments. In other words, one should be ready to discontinue the treatment if it fails to replicate in high-quality randomized controlled trials.