Best practice for motor imagery: a systematic literature review on motor imagery training elements in five different disciplines

None of the published reviews have analysed the design of the MI training session (MITS) to determine successful MI intervention techniques, such as the position of the person during MI, the number of MI trials, and the instruction mode and type. However, the MITS design is of vital importance for researchers and clinicians planning to implement MI interventions adapted to participant health status, age and gender. In this systematic literature review, we extracted and analysed 17 MITS elements based on the PETTLEP (physical, environment, timing, task, learning, emotion, perspective) approach. Furthermore, we analysed seven temporal parameters, including duration times and number of repetitions. In total, we analysed five disciplines in which MI represents an important training strategy.

Hall described the cognitive processes and neural basis of MI in a review on educational literature, based on a MEDLINE search [7]., and proposed a six-stage procedure for explicit learning of surgical skills: task definition, prior learning, mental rehearsal, reflection, problem solving and reality check.

In psychology, various tasks, participant groups and reporting statistics have been considered for MI. Driskell et al. summarised the effects of MP and determined under which conditions MI was most effective [13]. They defined five conditions of interest: 1) type of task, 2) retention interval, 3) experience level of trainees, 4) length of practice and 5) type of control group. The results of their meta-analysis showed a positive effect of MI when the following criteria were met: examination mainly of the cognitive aspects of the task performance, short retention interval, participants being novices to the task, and the MI session being about 20 minutes or shorter. They reported a non-significant trend for larger effects of MI compared with a non-treatment group and with an equivalent control treatment group.

In the sports psychology literature, six imagery models and frameworks were reviewed by Guillot and Collet [26]. The models included a four-component scheme originally designed by Martin et al., who described how MI influences cognitive, affective and behavioural outcomes [20]. The six-stage model from Munroe et al. was also evaluated, including the well known 'W' questions (where, when and why do athletes use MI, and what do they imagine?) [28]. This qualitative method includes the type (visual, kinaesthetic) and perspective (internal, external) of MI. MacIntyre and Moran extended the framework of Munroe et al. by adding the question: 'How should MI be executed and used by athletes?' [29], and they described a multimodal model that includes definition, outcome and importance of MI. Holmes and Collins introduced the PETTLEP framework, building on findings in functional neuroscientific research literature and experience in sport psychology [30]. PETTLEP aims to facilitate designing MI interventions for athletes, and comprises seven components (physical, environment, task, timing, learning, emotion and perspective). These components describe the physical position of the individual, the environment that has to be imagined, the task involved, the timing or duration of the imagery, the learning or changes involved during imagery, the emotions that are associated with the task, and imagery perspective. By contrast, the three-step model described by Watt et al. focused on MI ability and two image-generation approaches: 1) vividness, control, duration, ease, and speed; and 2) visual sensory methods [31]. The recent framework proposed by Guillot and Collet aimed to combine key components from previously described models. Their Motor Imagery Integrative Model in Sport (MIIMS) includes four MI outcomes: 1) motor learning and performance; 2) motivation, self-confidence and anxiety; 3) strategies and problem-solving; and 4) injury rehabilitation. The scheme aimed to combine different imagery types (visual, kinaesthetic, olfactory, tactile and auditory) to create a complete mental version of the movement [26].

MI research from sports psychology has been used in medicine, particularly in neurological rehabilitation [11‐13]. Literature reviews have evaluated the overall beneficial effect of MI [5, 9, 11], but none has described the MITS elements or temporal parameters. In this review, we analysed the MITS elements and temporal parameters that have been successfully used in different disciplines: Education, Medicine, Music, Psychology and Sports (in this review, we use the term 'Sports' for all studies that include athletes as participants and the term 'Psychology' for all studies including healthy participants who are not athletes).

Search terms were identified by a previous search of databases (including PubMed) and internet search engines (including Google and Google Scholar). Additionally, each searched database was checked for pre-defined MeSH terms, and where available, these terms were integrated into the search strategy. The following terms were used: 'mental imagery', 'mental practice', 'mental rehearsal', 'mental movements', 'eidetic imagery', 'visual imagery', 'guided imagery', 'motor imagery' and 'mental training'. The Appendix provides the complete search strategy for Scopus. The Scopus search strategy was adapted to individual databases and trial registers to account for specific vocabulary and syntax rules. No restrictions were made regarding year of publication, study design or age of the study population.

Table 1 provides an overview of all databases, trial and dissertation registers, and conference proceedings searched. Database retrievals were imported into a reference management software package (EndNote; Thomson Reuters, Carlsbad, CA, USA). In total, 21,739 references were retrieved in February 2007. The literature search was repeated in June 2010 for references published between spring 2007 and 2010 in the largest databases of each discipline: the Education Resources Information Center (ERIC), Scopus, Répertoire International de Littérature Musicale (RILM), PsycINFO and SPORTDiscus. This search resulted in 5,741 additional references.

The references were selected for review inclusion based on the following criteria:

The following exclusion criteria were used:

During the manual selection process (Figure 1), articles were evaluated based on title, abstract or keywords. Two of the authors (CS and RH) reviewed the articles; CS evaluated all references, and RH evaluated a randomly selected proportion (20%) of the initial number of references selected from each discipline. Full texts were ordered if no decision could be made based on the available information. If one reviewer could not reach a decision for a reference, this reference was discussed by both reviewers, and if both reviewers had not been able to agree on a decision (which was not the case in this investigation) a third reviewer (JB) would have been consulted.

Owing to the large number of references, EndNote search options were used to eliminate studies based on the exclusion criteria.

To confirm the selection congruency between both independent reviewers, the inter-rater congruency was calculated. Reviewer agreement ranged between 78% and 100% (average 96%) for the five disciplines. Because some studies reported more than one MI intervention, the total exceeded the number of included studies. Each MI intervention was subsequently analysed as an independent investigation.

Information on study methods, MITS elements and temporal parameters were extracted by three researches (RB, AS, CS) and checked for accuracy (CS). Table 2 summarises all extracted information. Figure 2 illustrates the temporal parameters and the MITS terminology.

Two rating lists were used because studies with different quantitative designs were included. The Physiotherapy Evidence Database (PEDro) list was used to evaluate randomised controlled trials (RCTs) (maximum of ten points) and non-RCTs (maximum of eight points) [32]. For case series or single cases experimental designs, the 11-point Single Case Experimental Design (SCED) scale was used [33]. All studies were rated by CS based on detailed rating guidelines. Studies received one point for each fulfilled methodological criterion on the respective rating list. The higher the achieved score, the better the study quality.

Raw information was extracted into Excel (Microsoft Corp., Redmond, WA, USA). After coding and classification, MI intervention data was imported into statistical analysis software packages (SPSS versions 16 and 17 (SPSS Inc., Chicago, IL, USA), MATLAB version 2009b (The MathWorks Inc., Natick, MA, USA)) for frequency analyses, frequency and mean comparison tests and visualisation. MI intervention data was not pooled or analysed for significant differences because of the variability in experimental settings and missing information in MI intervention descriptions. The heterogeneity between MI interventions was also present in standard deviations of temporal parameters. All MI interventions were classified into two categories: positive change (129 MI interventions, 91.5%), and no or negative change from pre- to post-test (12 MI interventions, 8.5%). MITS elements and temporal parameters of studies with positive change were summarised under the term 'average positive MI intervention' and used for comparison in three analyses.

First, trend analyses were performed to identify MITS elements for MI interventions with positive results versus no changes or negative results. Further analyses aimed to identify changing trends in MITS element frequencies in MI interventions with positive results for five different disciplines, integration approaches, MI training focus, session type, age and gender groups and MI intervention modifications. Secondly, the χ² was used to test for significant differences between actual and expected observation frequencies for each MITS element. The tests were performed if 20% of the expected frequency showed an amount of 5 or more [34]. Thirdly, for temporal parameters, normal distribution was tested using the Kolmogorov-Smirnov test, and variance homogeneity was confirmed by the Levene test. Depending on the test results, group means were compared using Student t-test or Mann-Whitney U-test. The tests were used if at least five observations were available to estimate the statistic. For all temporal parameters, group means were compared against the average positive MI intervention.

For all statistical tests p ≤ 0.05 was considered significant.

In total, 133 studies were included in the analysis, reporting 141 MI interventions in five disciplines: Education (9 Interventions), Medicine (37), Music (5), Psychology (79) and Sports (11). For the studies published between 1941 and 2010, there were peaks in 1989/1990 (8 publications), in 2004 (18) and 2007 to 2009 (38). In Medicine, MI publications first appeared in 2000, with a steady increase until 2010. These studies originated from Europe, Australia/New Zealand, the Americas, Asia, and the Middle East.

The study designs comprised 91 randomised controlled trials (RCTs), 22 controlled clinical trials (CCTs), 15 case series (CSs) and 13 single-case research designs (SCRDs). Study quality was rated on a 10-point scale for RCTs (4 to 9), an 8-point scale for CCTs (3 to 6), an 11-point scale for CSs (4 to 11), and on an 11-point scale for SCRDs (7 to 10).

On average, RCTs and CCTs scored 6 on the 10-point PEDro scale, whereas CSs and SCRDs scored 6 and 8, respectively, on the 11-point SCED scale (on both, higher scores indicate better quality). Examples of MI instructions were available for 29 MI interventions, and changes in MI content during the MI intervention period were reported in 31 MI interventions. An overview of essential study characteristics is provided for each discipline separately (Table 3, Table 4, Table 5, Table 6 and Table 7).

The MITS elements for all MI interventions were compared (Figure 3A). Frequency analyses of MI interventions with positive results revealed a number of key MITS elements present in a successful intervention design: MI was performed in individual sessions and added after PP; MI sessions were supervised and not directed; locations of MITS and the position of the participants during MI were both task-specific; participants received acoustic and detailed MI instructions, which were mainly standardised and live; during MI practice, participants kept their eyes closed.; the perspective used during MI practice was chosen from an internal view combined with a kinaesthetic MI mode; and MI interventions were mainly investigated with motor-focused tasks.

Only 17 reports mentioned an MITS for familiarisation before the MI intervention began. The reporting rate of all MITS elements ranged between 26% for the description of closed or open eyes to 95% for MI instruction individualisation. The most frequently reported MITS elements in successful MI interventions are listed in Table 2.

MI interventions with no change or with negative results predominantly used directed MITS. If MI integration was reported, MITS were embedded between or performed simultaneously with physical trials. Owing to the lack of reporting, the ordering of MI and PP could not be identified in 90% of all MI interventions. Only two MI interventions mentioned an MITS for familiarisation before MI intervention began. For the subsequent analyses only successful MI interventions with positive results were considered.

MI interventions with positive results had almost twice the duration of MI interventions with no change or negative results: study duration (34 days), MI intervention duration (21 days), total MITS count (13), (the number 13 stands for the number of MITS in MI interventions with positive results)MITS duration (17 minutes) and total MI time (178 minutes). By contrast, MI interventions with no change or negative results had a larger number of MITS per week (3) and a larger number of MI trials per MITS (34).

In the described analyses, only positive MI interventions were considered. The two disciplines with the youngest participants were Psychology (aged up to 9 years) and Sports (10 to 19 years). Most MI interventions were carried out with students (20 to 29 years) in Education, Psychology, Music, and Sports. Participants aged 50 and older were included only for Medicine and Psychology. Gender-specific investigations were carried out in Medicine, Psychology and Sports.

Detailed discipline-specific frequency analyses of MITS elements revealed the following differences (italics) from the average positive MITS.

For Education, participants (Figure 4A-F) performed MI predominantly before physical practice (PP) during directed MITS. Three MITS elements showed both categories: position of participants during MITS (task-specific and not task-specific), instruction mode (life and pre-recorded), and perspective (internal and external). MI content focused on cognitive task-related activities. MI mode was not reported. Regarding the temporal parameters in Education, the study and MI intervention duration and the total MI time were less than half of those in the average positive MI intervention, but the MITS duration was twice as long as in the average positive MITS. The number of MI trials per MITS was not reported.

MI interventions in Medicine (Figure 5A-F), the MI interventions predominantly used directed MITS with pre-recorded MI instructions. All temporal parameters had longer durations and total counts compared with the average positive MITS, especially for study and MITS duration, total MITS count and total MI time.

MI interventions in Music (Figure 6A-F) tended to be embedded into PP. MI instructions in Music were mainly written. Instruction mode and type allowed for more than one categorisation. Location of the MITS, and the MI perspective and mode used were not described. Almost all temporal parameters had lower durations and numbers than in the average positive MITS, particularly the number of MI trials per MITS, but the study duration was higher for MI interventions in Music.

MI interventions in Psychology (Figure 7A-F) most closely resembled the average positive MITS with similar distributions of MITS elements in instruction type and in MI perspective. MI interventions had the same number of MITS per week to the average positive MITS, but shorter MI intervention and MITS duration. Total MI time was half of that of the average positive MI intervention.

MI interventions in Sports (Figure 8A-F) reported embedded and directed MITS, [and after or between PP. Instructions during MITS were tailored and in written form. Study and MI intervention duration were almost twice as long as those in the average positive MI intervention. MI trials in a Sports MITS were only half of the number of the average positive MITS.

This analysis was performed in successful MI interventions, which included PP in the study design. Of the 34 retrieved MI interventions, 20 described an added and 14 an embedded MI training method (Figure 9A-F). There was a preference for added MITS to be directed using pre-recorded instructions. Neither the locations of MITS nor the position of the participants during MI practice were task-specific. Added MITS used a kinaesthetic or mixed MI mode. MI training embedded into PP tended to be supervised, and was implemented between physical trials of the same task. Most MI interventions did not report details on location and position.

The averages of the temporal parameters of both categories (added and embedded) differed from those of the average positive MI intervention; however, there was wide variation between MI interventions. The duration of the study MI intervention and MITS duration were longer for both categories than in the average positive MITS, but the number of MI trials per MITS was lower. MITS duration was longer in the added than in the embedded training methods.

Based on the primary focus of activities that were imagined, positive MI interventions were categorised into motor-focused (94), cognitive-focused (29) and strength-focused (6) activities. Compared with motor and cognitive-focused MI interventions, which were mainly published between the 1970s and 2010, the majority of strength-focused MI interventions were published in 1991 and in the period 2004 to 2009. Motor- and strength-focused MI interventions were often designed according to the average positive MITS (Figure 10A-F). Strength-focused MI interventions were investigated in healthy participants aged 20 to 39 years only. Motor-focused MI interventions had the highest number of MI trials per MITS and the longest MITS duration and total MI time.

Cognitive-focused MI interventions differed from the average positive MI intervention: there was a preference for MITS to be embedded and directed. Cognitive-focused MI interventions had shorter durations and lower numbers in all temporal parameters compared with motor-focused MI interventions.

This analysis could be performed for 37 positive MI interventions that reported details. In total, 21 MI interventions described MITS in group sessions, and 71 in individual sessions (Figure 11A-F).

Group MITS tended to be directed and embedded into PP, and included MI practice before and after PP. Neither the locations of the MITS nor the position of the participants during MI practice were task-specific. Both the MI perspective used during MI practice and the MI mode changed. Total MI time and number of MI trials per session were only half those of the average positive MI intervention.

For the individual sessions, we investigated two options: directed and non-directed MITS. Compared with the average positive MI intervention, individual sessions had larger values for many of the temporal parameters, particularly total MI time.

Participant age in successful MI interventions was classified into seven categories: up to 9 years (2 interventions), 10 to 19 years (18), 20 to 29 years (63), 30 to 39 years (13), 40 to 49 years (2), 50 to 59 years (9), and 60 and older (20). Two MI interventions did not mention the age of the participant and were thus not considered in this analysis.

There were only two MI interventions with participants aged up to 9 years, published in 1973 [35] and 2004 [36]. Both were studies in Psychology, which considered healthy children of both genders with an average age of 6 and 9 years, respectively, and were supervised with acoustic instructions. Rapp and Schoder described the MI intervention as a non-directed group session with live and standardised instructions. Children closed their eyes during MI as they imagined a motor-focused task [35]. No further details were provided. Taktek et al. designed the MI intervention as a directed session with pre-recorded instructions [36]. Participants used a task-specific position during MI, closed their eyes and used a kinaesthetic MI mode when imagining a cognitive-focused task. MI trials were preformed before PP trials using standardised and detailed instructions. Temporal parameters differed between both investigations.

The MI interventions (n = 18) with teenagers (10 to 19 years) were in the fields of Psychology and Sports. Investigations were designed as directed individual sessions. Where reported, MI was practiced either before, or before and after PP. Participants received their MI instructions in written form. Only the number of MI trials per MITS was less than that in the average positive MI intervention (Figure 12A-F).

MI interventions (n = 13) with participants aged between 30 and 39 years were mainly designed as added MITS with MI practice before PP and coarse (broad) MI instructions. MI mode was reported in two MI interventions as kinaesthetic and changing mode, respectively. Four temporal parameters had twice the duration or frequency than in the average positive MI intervention: study duration, MITS duration, total MITS count and total MI time. The number of MI trials per session was lower than those of the average positive MI intervention.

Only two MI interventions could be classified in the age group 40 to 49 years [37, 38]. The MI intervention described by Vergeer and Roberts was performed with healthy participants of both genders [37]. MITS elements resembled those of the average positive MI intervention, with group sessions and a multimodal approach for MI instructions (written, visual and acoustic). The second article, published by Guillot et al., described an MI intervention with participants (male and female) who had burns [38]. The authors used added, individual, supervised and directed MITS. Participant received detailed, tailored and written instructions.

The MI interventions (n = 9) with participants aged between 50 to 59 years were directed MITS (Figure 13A-F). Neither the locations of the MITS nor the position of the participants during MI practice were task-specific. Internal and external MI perspective options were offered. A changing MI mode was preferred. Temporal parameters had almost twice the duration of MI study and MI intervention than in the average positive MI intervention.

Gender-specific analyses of positive MI interventions showed the following distribution: sixty-nine MI interventions with participants of both genders, eight MI interventions with female participants only and thirty-four MI interventions with male participants only. Gender-specific investigations were carried out in Medicine, Psychology and Sports.

The MI interventions with participants of both genders (n = 69) were designed according to the average positive MI intervention. They had the longest study duration and total MI time and the largest number of MI trials per session compared with gender-specific MI interventions (Figure 14A-F).

The MI interventions (n = 8) with female participants only were primarily designed as embedded MITS with MI trials between PP trials. The locations of the MITS were both task-specific and non-task-specific. MI instructions were live or pre-recorded. MI interventions with female participants were investigated with mainly cognitive-focused tasks. The MITS duration and number of MI trials were lower than those of MI interventions with both genders and with male participants only; however, the duration of the MI intervention was longer.

The MI interventions (n = 34) with male participants used task-specific or non-task-specific positions, and the MI mode selected was the changing mode. Temporal parameters closely resembled those of the average positive MI intervention.

Successful MI interventions were categorised with regard to change in MI intervention during the total MI intervention period into change (n = 31) and no change (n = 38). Change in MI intervention could include three domains: MI dosage, MI content and MI time. MI interventions were excluded from the analysis if a change was not clearly described (n = 55) or if the categorisation was not applicable (n = 5) due to the study design.

Only minor differences were found between categories (Figure 15A-F). MI interventions with a change during the MI intervention period included directed MITS. Duration of study and MI intervention, total MITS count, and total MI time were almost twice those of the average positive MI intervention. MI interventions without a change during the MI intervention period were designed as embedded MITS with shorter study duration, lower number of MI trials and lower total MI time than in the average positive MI intervention.

A question frequently raised by clinicians is 'How should motor imagery be done?'. Our literature review aimed to answer this question and to describe which elements characterise successful MITS. It was not our intention to evaluate the effectiveness of MI or to compare effect sizes, as this has already been addressed in other literature reviews [1, 6, 11, 13]. The results of the trend analyses revealed changes in the frequencies of the MITS elements, which represent important variations between MI interventions. In addition, the review identified differences between the studies with positive results and those with no changes or negative results. Thus, the trend analyses might help clinicians to implement MI interventions successfully. By contrast, the χ² test revealed general frequency distribution differences only, which were often caused by frequency variations and did not represent actual trend changes. Owing to limitations in the reporting rates, the χ² test for MITS elements and the group mean comparison tests for temporal parameters could not be applied in many cases. We expect that our trend analyses in combination with the statistical test would be able to serve as indicators for potential future research directions.

Our analyses considered the differences in specific disciplines (Education, Medicine, Music, Psychology, Sports), MI integration types (added/embedded), session types (individual/group), focus of the task (motor, cognitive, strength), age, gender groups (female, male, both ) and change in content, duration and dosage. From 141 MI interventions, data were extracted and analysed for 17 MITS elements based on the PETTLEP approach and 7 temporal parameters.

The only available MeSH term for searches was 'mental imagery', which must be considered as an umbrella term for various mental techniques. MI is one technique focusing on movements, which is important in rehabilitation medicine. Historically, other terms have been used for the same purpose in literature. Our literature search included various terms associated with imagery, yielding a large initial reference count.

Studies were included regardless of their study design, country and year of publication. This method enabled us to obtain a global view of the MI literature in different disciplines and of the MI approaches that were evaluated in different study designs. We used and adapted two widely accepted scales to evaluate all studies for their methodological quality.

The analysed studies primarily investigated the short-term effect of MI with a simple pre-/post-test design. The longest time period evaluated was a 6-month follow-up in an RCT by Moseley et al., in which significant improvements were seen in the MI treatment group compared with a control group [47].

Overall, data reporting in the selected articles was low, and the implications of this are highlighted by one of the least reported elements: imagery perspective. Depending on the chosen perspective (first or third person), different brain areas will be activated [48]. Publications on successful and non-successful athletes reported contradictory results for the imagery perspectives used [46, 49, 50]. Furthermore, Kim et al. investigated the exercise-related imagery perspective in middle-aged adults and, reported an internal:external perspective ratio of 1.8 [51]. Mulder et al. found a slightly better MI vividness in adults over 64 years when using the external MI perspective. The authors also mentioned that MI from an internal perspective is more important than MI from an external perspective in learning a motor skill [52, 53]. Furthermore, they could detect a shift in perspective related to age, with younger people more likely to use the internal perspective and older people more likely to use the external perspective. Taking imagery perspective as an example, future research should detail MITS elements more carefully.

There were two important sources of possible information bias: firstly, 51 references were not obtainable, and secondly, our selected references included only 12 MI interventions with no change or negative results versus 129 MI interventions with positive results. We therefore hypothesised that MI interventions without positive results are rarely published. This hypothesis is further supported by our identification during the selection process of abstracts detailing preliminary results of MI interventions but no follow-up full article describing the whole MI intervention and its final results. Nevertheless, the aim of this review was to analyse MI interventions with positive results, and to identify discipline-specific MI interventions and fundamental intervention designs.

We found that the reporting standard of MI intervention had improved in recent years; however, investigations published before 2007 often lacked details on MITS elements, which resulted in missing data in the frequency analyses. Many investigations included more than one experimental or control group. In such cases, we focused our analyses focused on the experimental group with the largest change in measurement between pre- and post-intervention measurement.

The MI interventions were heterogeneous, which explained the large standard deviations in temporal parameters.

Task evaluation is complex and subjective, and to date, no standard classification exists. In our review we classified the investigated tasks based on their main focus: motor, cognitive or strength.

Before applying an MI intervention, it is essential to evaluate the MI ability of the participants to determine whether they are able to perform MI. Additionally, MI ability might change over an intervention period. In the current literature review, we found that assessments of MI ability had been used in thirty-six studies with positive results [36‐38, 54‐83] and in five studies with no change or with negative results [84‐88]. Heterogeneity between the MI ability assessments used, which were partially custom-designed for individual MI interventions, prevented direct comparison and relation to the study results in this review. We hope this will encourage researchers to use assessments of MI ability and to state participant scores in their research reports.

This review does not include MI interventions that were published after June 2010, because of the reference selection and analysis process. However, we briefly mention new articles currently under review in Medicine. Braun et al. embedded MI training into regular therapy in patients with stroke in nursing homes and in patients with Parkinson disease at different disease levels. In both investigations, embedded MI did not show a significant advantage compared with the control group receiving regular care [89, 90]. These interventions may add information for analysing positive results versus no change or negative results.

The current review focused on MITS elements to improve motor function or learn a motor skill. Further reviews should consider the influence of MI on psychological factors, such as goal-setting [91], self-efficacy, motivation and mood [92], and working memory.

Search strategy Scopus database from 22 February 2007: (((TITLE-ABS-KEY("mental imagery")) OR (TITLE-ABS-KEY-AUTH("mental practice")) OR (TITLE-ABS-KEY-AUTH("mental rehearsal")) OR (TITLE-ABS-KEY-AUTH("mental movements")) OR (TITLE-ABS-KEY-AUTH("eidetic imagery")) OR (TITLE-ABS-KEY-AUTH("visual imagery")) OR (TITLE-ABS-KEY-AUTH("guided imagery")) OR (TITLE-ABS-KEY-AUTH("motor imagery")) OR (TITLE-ABS-KEY-AUTH("mental training"))) AND NOT (TITLE-ABS-KEY-AUTH("mental health"))) AND NOT (TITLE-ABS-KEY-AUTH("body image")) 2,556 references