Malouin, Francine PT, PhD; Richards, Carol L. PT, PhD; Jackson, Philip L. PhD; Lafleur, Martin F. PhD; Durand, Anne PhD; Doyon, Julien PhD
A major difficulty when using mental practice through motor imagery is to determine to what extent a person is able to generate a mental representation of movements. The question is even more problematic after a stroke because results from chronometric studies indicate that, contrary to patients with a lesion in the motor cortex,1 patients with lesions in the superior regions of the parietal cortex may have impaired motor imagery ability.2 In addition, recent findings indicating a slowing of the imagery process after stroke suggest that temporal characteristics of motor imagery could be modified after stroke.3,4 Together these results suggest that some patients may not be able to engage in motor imagery and hence may not benefit from mental practice.5 Therefore, it is important to assess motor imagery ability.6
To this end, movement imagery questionnaires have been developed, two of the most widely used are the Movement Imagery Questionnaire (MIQ),7 followed by a revised and shorter version, the MIQ-R8, and the Vividness of Motor Imagery Questionnaire (VMIQ).9 These, however, are self-report questionnaires that have been used largely in young healthy adults.
The MIQ was designed to evaluate the vividness of movement imagery in the visual and the kinesthetic dimensions. The initial version of the MIQ includes 18-items (nine for the visual subscale, nine for the kinesthetic subscale). Each item involves arm, leg, or whole body movements (eg: arm abduction, jumping, front roll). The MIQ does not measure imagery vividness directly, but rather the ease/difficulty with which the subject imagines the movement on a seven-point scale (1 = very easy to see/feel; 7 = very hard to see/feel). The administration involves four steps: (1) the subject assumes a specific starting position; (2) the movement is described and the subject is asked to perform it; (3) the subject reassumes the starting position and is required to imagine producing the movement (no actual movement is made); (4) the subject is required to rate the ease/difficulty with which they imagine the movement on the seven-point scale. The reliability of the MIQ, studied in 50 young individuals (26 women and 24 men; mean age 21 years) has been judged acceptable, with reported Pearson correlation coefficients of 0.83 for both subscales, for test-retest at a 1-week interval, and internal consistency coefficients of 0.87 and 0.91 (Cronbach’s α) for the visual and kinesthetic dimensions, respectively.10 Lorant and Gaillot11 reported Pearson correlation coefficients of 0.88 (visual scale) and 0.87 (kinesthetic scale), for test-retest at a 3-week interval in a group of 32 students in physical education (mean age 20 years); likewise Cronbach α values of 0.87 and 0.91 were computed for each scale, respectively. Similar levels of internal consistency were found with Cronbach α values of 0.89 and 0.88 for the visual and kinesthetic scales, respectively.12 In addition, the bifactorial structure of the MIQ has been confirmed by a common factor analysis,11,12 indicating that the MIQ measures both the visual and kinesthetic dimensions of movement imagery.
The MIQ was revised in 1997 by Hall and Martin8 resulting in a shorter version: the MIQ-R. The MIQ-R includes only eight items (four items for each subscale) the rating scale was reversed (7 = very easy to see/feel; 1 = very hard to see/feel), and some of the items were reworded to enhance the clarity of the descriptions. Several items were removed from the original version to reduce the length, eliminate redundant items, and, more importantly, get rid of more physically demanding items (such as front roll and 360-degree turn) that some subjects refused to perform physically (step two). Administration of the questionnaire involves the same four steps as described above. Concurrent validity of the MIQ-R has been studied in a group of 50 subjects (24 men and 26 women (mean age 21 years; range: 18–41 years) by Hall and Martin8 who correlated MIQ-R scores to MIQ scores. They reported coefficients of correlation of −0.77 (negative values due to the reversal of the MIQ-R rating scale) for the two subscales suggesting that the MIQ-R is an acceptable revision of the MIQ. More recently, using the scores from 134 healthy subjects (mean age of 24 years and range 17–60 years) who had completed the French version of the MIQ-R, Lorant and Nicolas13 were able to confirm the internal consistency of the MIQ-R with a Cronbach α of 0.82 for both subscales as well as the bifactorial structure of the revised version. In the same study, the authors examined the test-retest reliability in a subgroup of 46 subjects who were tested at a 3-week interval. They reported Pearson correlation coefficients of 0.86 and 0.90 for the visual and kinesthetic subscales, respectively,13 indicating reliability levels slightly higher than those reported for the former MIQ4 (r = 0.83 for both subscales).
The VMIQ, developed by Isaac et al.,9 is a 48-item questionnaire in which subjects are required to rate the vividness of their imagery on a five-point scale (1 = as clear and vivid as normal vision; clear; 5 = no image at all) for a series of 24 physical activities under two conditions: (1) while imagining someone else executing the movement (visual subscale) and (2) while imagining themselves performing the movements (kinesthetic subscale). Contrary to the MIQ, the subjects are not required to perform physically the movement before imagining it. The test-retest reliability of the VMIQ examined in a group of 220 students (high school and university levels), tested twice at a 3-week interval, yielded a Pearson correlation coefficient of 0.76. Although the VMIQ was developed to assess visual and kinesthetic imagery, a structural factor analysis did not confirm the bifactorial structure of the VMIQ.14 In addition, when scores from the VMIQ were correlated with scores from a visual imagery test, the VVIQ (the Vividness of Visual Imagery Questionnaire),15 and the visual subscale of the MIQ, it yielded correlation coefficients of 0.78 and 0.65, respectively. In contrast, the correlation between the VMIQ scores and the kinesthetic subscale of the MIQ was only 0.49.8 Altogether, these observations suggest that the VMIQ does not tap into the kinesthetic component of imagery as does the MIQ.
Based on the above findings, the MIQ appears to be the measure of choice to assess motor imagery. The MIQ, however, has been developed and used to assess motor imagery ability in healthy adults and athletes,10–13 and its use in rehabilitation raises many difficulties. Mainly because of the self-reporting nature of the MIQ and the high physical demands of several items, the MIQ and the MIQ-R are unsuitable for persons with physical disabilities. For this reason, we have developed the Kinesthetic and Visual Imagery Questionnaire (KVIQ-20). The KVIQ-20 is a motor imagery questionnaire adapted for persons who, for different reasons, have to be guided in the rating of their imagery and who are not able to stand or to perform complex movements. Like the MIQ, the KVIQ-20 was developed to assess both the visual and kinesthetic dimensions of motor imagery and the testing also involves four steps. Importantly, however, the KVIQ is not self-administered. In addition, unlike the MIQ, the rating scale of the KVIQ-20 assesses the vividness of each dimension of motor imagery (clarity of the image/intensity of sensation) on a five-point ordinal scale. Because the administration of the KVIQ-20 may take up to 45 minutes with persons with disability or those who have more difficulty concentrating, a short version has been created (the KVIQ-10) to make it more suitable for clinical use.
Therefore, the purposes of this study were to: (1) examine the test-retest reliability of the KVIQ-20 and its short version (the KVIQ-10) in healthy subjects and subjects with stroke; (2) investigate the internal consistency of both KVIQ versions; and (3) explore the factorial structure of the two KVIQ versions.
To study the test-retest reliability of the two versions of the KVIQ, 19 persons who sustained a cerebral vascular accident (CVA), a group of 46 healthy individuals (CTL) for comparative purposes, and an age-matched CTL group of 19 healthy subjects (aCTL) was also extracted from the total CTL group (Table 1). The internal consistency and factorial structure were examined in 131 persons, including subjects from the reliability study (Tables 1 and 2). All subjects gave informed, written consent for their participation in the study. The protocol was approved by the Ethics Committee of the Institut de Réadaptation en Déficience Physique de Québec, where the study was conducted.
Reliability Study: Test-Retest
To be included in the CVA group, subjects had to present a hemiparesis subsequent to a stroke. The exclusion criteria were: (1) lesion in the cerebellum or midbrain; (2) severe aphasia; (3) severe perceptual (eg, apraxia, hemineglect) and severe cognitive impairments; (4) other neurological conditions (eg, Parkinson’s disease, dementia). To be included in the CTL group, subjects had to be between 18 and 80 years old and without physical or intellectual impairments.
Internal Consistency and Factor Analysis
For these analyses, patients were recruited from different ongoing studies in which the KVIQ-20 was used as an evaluation tool. Thus, in addition to subjects from the CVA and CTL groups (see above), three groups of persons with either a visual (acquired blindness) or musculoskeletal impairment (lower limb amputation and immobilization) were included. The subjects with acquired blindness had a total loss of vision for at least two years and had received orientation training. The subjects in the amputation group had a traumatic amputation of one lower limb (transfemoral or transtibial), were ambulatory, and had completed their prosthetic training. In the immobilization group, subjects had been in a cast for at least two weeks due to a fracture of the ankle and had not resumed weight-bearing on the affected limb. Subjects with sensation loss due to neuropathy or severe phantom pain were excluded.
The KVIQ-20 and KVIQ-10
The structure of the KVIQ-20 is similar to that of the MIQ7; it includes visual (V) and kinesthetic (K) subscales, and the assessment process also comprises four steps. The KVIQ-20 includes 20 items (10 movements in each of the V and K subscales) representing gestures from different body parts (Table 3). The gestures are simple movements of the head, shoulders, trunk, upper limbs, and lower limbs. These movements were selected for their use in persons with physical disabilities, and for this reason, all 10 movements are performed from a sitting position. In addition, instead of a seven-point scale to rate the difficulty to see or feel the movement, the KVIQ-20 uses a five-point scale to rate the clarity of the image (V subscale) and the intensity of the sensations (K subscale); a score of 5 corresponds to the highest level of imagery and a score of 1 to the lowest (Table 4). Unlike the MIQ, which is a self-reported questionnaire, the KVIQ-20 is administered by an examiner who reads the instructions and records the score. The subjects are required to rate their imagery using the operational definition of each category (eg, 5 = image as clear as seeing) and the numbered scale is used only for computation of the data.
This version includes 10 items (five movements in each of the V and K subscales) instead of 20 (Table 3). The items retained correspond to a representation of body movements comparable to that of the KVIQ-20: one trunk movement, one proximal movement, and one distal movement of the upper and lower limbs.
The testing procedures for the administration of the questionnaire were similar for all subjects. The visual imagery was assessed first, followed by the kinesthetic imagery. The subject was sitting in a chair with a backrest, and the examiner sat in front of the subject. The following steps were used for each item: (1) the subject was asked to assume the start position demonstrated by the examiner; (2) the examiner demonstrated the movement and then the subject was required to execute the movement physically only once (patients unable to execute the movement physically with the affected limb were requested to use the unaffected limb); (3) the subject was then asked to return to the starting position and to imagine performing the same movement that he or she had just executed; (4) the examiner asked the subject to rate on the five-point scale the clarity of the visual image or the intensity of the sensations associated with the imagined movement. For the present study, we have tested items of the upper and lower limbs bilaterally to determine whether the reliability was similar when imagining moving the affected or the unaffected limb (or dominant and nondominant). For the test-retest study, the same procedure was repeated about seven to 14 days later by the same examiner.
The KVIQ was administered by three examiners: two physiotherapists, a senior therapist who had worked for more than 20 years and a junior therapist with two years of experience, both in orthopedic and neurology areas. The third examiner was a third year student in occupational therapy. The senior physiotherapist who had been trained in the administration of the KVIQ by the developers of the questionnaire was responsible for the standardization (written documentation) and training of the other examiners. Live and videotaped demonstrations followed by discussions about scoring were used for training new examiners. Assessments were also videotaped for review of the scoring when necessary.
For each group and each session, the scores from the visual and kinesthetic subscales (maximal values: KVIQ-20 = 50 and KVIQ-10 = 25) and from the total KVIQ (maximal values: KVIQ-20 = 100 and KVIQ-10 = 50) were computed. For both versions, the reproducibility of the measurements of each subscale and total scale was estimated for each group using intraclass correlation coefficients (ICCs) (one-way random effect model, SPSS) and the 95% lower confidence limit.16 The standard error of measurement (SEM) was also calculated for each outcome measure using the following formula: SEM = √ςe2, where ςe2 is the error variance and equals the mean square error term from an analysis of variance.17 In addition, comparisons were made between the visual and kinesthetic subscores using the paired Student t test.
The homogeneity of items composing each subscale was evaluated with Cronbach’s α. Alpha values were tested against a minimal value of 0.70 and a 95% confidence interval was estimated with the ICC function of the reliability procedure from the SPSS software (two-way random model).
The latent structure of the two KVIQ versions was assessed with the principal factor extraction technique and oblique rotation since the visual and kinesthetic factors were expected to correlate. All the analyses were performed with SPSS (v.11.0.1 and 13, Chicago, IL).
Test-Retest Reliability of the KVIQ-20
The mean time interval between test and retest sessions was 13.4 days (median 7; min-max seven to 50), 11.1 days (median 8; min-max: seven to 32) and 10.1 days (median 8; min-max 6 to 32) in the CVA, aCTL, and CTL groups, respectively. There was no significant difference between groups for the delay between test-retest (p > 0.05). For both versions and groups, comparisons between KVIQ scores at test-retest sessions yielded no difference (p > 0.05). The mean (1 standard deviation) scores recorded at test and retest, together with respective ICCs for each group and the two KVIQ versions are reported in Tables 5, 6, and 7. Within each group, comparable reliability levels were found for both versions. The ICCs ranged from 0.81 to 0.90 in the CVA group, from 0.73 to 0.86 in the aCTL group, and from 0.72 to 0.81 in the CTL group, indicating good reproducibility levels (Tables 5, 6, and 7). Table 8 reports the estimated components of variance related to subjects, time, and random error from the ANOVA analyses. As can be seen, the variance due to subjects for visual and kinesthetic imagery was low in the CTL and aCTL groups compared to that of the CVA; in addition, for kinesthetic imagery the variance due to time in the aCTL group was high compared to the other groups.
To illustrate individual variations in imagery, visual, and kinesthetic KVIQ-20 scores from each subject recorded at test and retest are given for each group in Figure 1. Comparisons between visual and kinesthetic subscores within each group indicated higher visual subscores in the CVA group (KVIQ-20: p = 0.02 at test and retest, KVIQ-10: p = 0.058 at test; retest: p = 0.01), the aCTL group (KVIQ-20 test: p = 0.01, retest: p = 0.05, KVIQ-10 test: p = 0.019, retest: p = 0.013) and the CTL groups (KVIQ-20: p < 0.001 at test and retest; KVIQ-10: p = 0.04 at test and retest). Figure 2 illustrates individual visual subscores relative to corresponding kinesthetic subscores for each group. As can be seen, the visual scores were generally larger (squares above the line) than kinesthetic subscores.
The ICCs and SEM values for items of the upper and lower limbs on each side in the CVA and aCTL groups are reported in Table 9. In the CVA group, the ICCs when imagining movements on the affected side (upper and lower limbs combined: seven items) ranged from 0.71 to 0.87 and from 0.86 to 0.94 on the unaffected side. When upper limb (three items) and lower limb items (four items) were analyzed separately, ICC values when imagining movements on the affected and unaffected side ranged, respectively, from 0.69 to 0.93 and from 0.78 to 0.90. In the aCTL group, for the upper and lower limbs combined (seven items), ICCs ranged from 0.75 to 0.89 when imagining movements on the dominant side and from 0.81 to 0.92 on the nondominant side. When upper limb (three items) and lower limb items were analyzed separately, ICC values in the aCTL group ranged, respectively, from 0.62 to 0.79 and from 0.60 to 0.85. Thus, the reproducibility is better in both groups when items of the upper and lower limbs are combined. Moreover, in both groups, the less reproducible conditions were associated with less intersubject variance or a greater random error component of variance.
Internal Consistency and Factor Analysis
The Cronbach α values for the visual and kinesthetic subscales of the two KVIQ versions given in Table 10 ranged from 0.87 to 0.94. Items-to-corrected items correlations ranged from 0.63 to 0.83 (KVIQ-20) and from 0.62 to 0.77 (KVIQ-10), indicating that no item was problematic.
Two factors were extracted and the promax rotation was necessary to obtain an adequate solution. For both versions, all the visual items were explained by the first factor, and all the remaining kinesthetic items were explained by the second factor (Tables 11 and 12). All the items were well defined by the factor solution, and communalities ranged from 0.44 to 0.73. For the KVIQ-20, the two factors explained 63.4% of the total variance (visual: 47.6%, kinesthetic: 17.8%) and 67.7% for the KVIQ-10 (visual: 45.6%, kinesthetic: 20.1%). The correlation between the two factors was 0.46 for both versions.
The results from the test-retest analysis indicate that both versions of the KVIQ can be used with reliability in healthy subjects and in patients post-stroke. The ICC values were all >0.81 in the CVA group, indicating very good stability of the measures. Although the level of reproducibility was good in the CTL groups with ICCs ranging from 0.73 to 0.86 (aCTL) and from 0.72 to 0.81 (CTL), it was not as good as in the CVA group. Interestingly, the level of reproducibility in the large CTL group (age range 20–78 years) and the aCTL group (age range 45–78 years) was comparable, suggesting that age did not markedly affect the reliability of the KVIQ scores. The examination of the variance output (Table 8) revealed that items with lower reproducibility levels had smaller intersubject variations, which make them more sensitive to time variation and can explain the lower ICCs. The other source of variation for visual imagery in the CTL group and kinesthetic imagery in the aCTL group was the variance due to time (Table 8), which indicates that these conditions were more variable over time in these groups. Thus, present findings that motor imagery measures are more stable in the CVA group leads one to question whether the perception of motor imagery becomes more stereotyped after cerebral lesions.
In general, the reproducibility was better when patients imagined movements with the unaffected limbs than the affected limbs, and the wide range of ICCs emphasizes large variations across conditions (Table 9). Again, the main factor that can explain the lower reproducibility was a small intersubject variation. Similar observations can be made in the aCTL group. Overall, the best reproducibility levels were obtained when combining scores from the upper and lower limbs, thereby increasing the variance due to subjects.
Except for the aCTL group, the ICC values of the visual imagery subscores were lower compared to ICCs for the kinesthetic imagery subscores. Such a pattern could have resulted from an order effect linked to our testing procedure. Indeed, to ease the testing procedure, visual imagery for all 10 items was assessed first, followed by the kinesthetic imagery, instead of alternating between the visual and kinesthetic imagery for each item, a procedure used with the MIQ7 and MIQ-R.8 This is unlikely, however, for two reasons. First, our findings concur with those of Lorant and Nicolas,13 who examined the test-retest reliability of the MIQ-R in 46 subjects tested at a 3-week interval. They reported Pearson correlation coefficients of 0.86 and 0.90, respectively, for the visual and kinesthetic subscales. Second, we have found (Malouin et al., unpublished observation) that the order of testing has no effect on the scores in a group of 26 subjects in which one half of the subjects were tested first with visual imagery and the other half with kinesthetic imagery.
In fact, people generally get higher visual than kinesthetic scores,10,12,13 suggesting that visual imagery is easier than kinesthetic imagery. Thus, beginning the assessment with the visual subscale has the advantage of progressively exposing subjects to the concept of motor imagery. Another advantage of this testing procedure is that subjects do not have to switch repetitively from the visual to the kinesthetic descriptors between each item, thereby creating interference between the two dimensions. For these reasons, assessing visual before kinesthetic imagery appears to be a good strategy, especially in older and disabled persons who are more or less familiar with motor imagery and have limited attention and concentration skills. Finally, the reproducibility of the KVIQ measures using our testing procedure supports its use for persons with stroke.
The high Cronbach α values (range 0.87–0.94) confirmed the internal consistency of both versions of the questionnaire. The KVIQ demonstrated very good item-to-goal score consistency. Such item homogeneity suggests that all items measured the same construct. None of the items, if removed, affected the level of consistency, thus indicating that they all were contributing. Although the Cronbach α values of the KVIQ-10 were slightly lower than those of the KVIQ-20, they were nevertheless high (0.87 and 0.89, respectively) and equivalent to those reported for the MIQ10,12 and the MIQ-R13. The latter finding indicates that even with the removal of five movements, the KVIQ-10 maintains its internal consistency. Given that the KVIQ-10 can be administered in half the time, it appears to be a good choice, especially for assessing persons with physical disabilities. The KVIQ-20, however, remains of interest for research purposes as it provides a means of exploring motor imagery ability over a wider variety of movements (eg, head, elbow, knee). A longer version, including bilateral testing of the seven items related to limb movements is also useful for examining the symmetry of motor imagery ability especially in persons with unilateral impairment such as amputation, immobilization, and stroke.
The factor analysis confirmed the bifactorial structure of both versions. This finding supports the notion that the KVIQ-20 and KVIQ-10 assess two distinct dimensions of motor imagery: visual and kinesthetic. For each scale, the two factors explained more than 63% and 67% of the total variance, respectively. These values compared to the 65% reported for the MIQ-R7 but are higher than the values (45% and 54%) reported for the MIQ.11,12
The factor analysis also indicated that the two factors were correlated (0.46). Correlations of 0.58 and 0.39 have been reported by others.10,11,13 Such correlation between factors indicates that although the visual and kinesthetic scales share common characteristics, they still represent two distinct dimensions of motor imagery. Further analyses are under way to identify the source and contribution of the common variance.
Present findings indicate that the KVIQ-20 and the KVIQ-10 are two motor imagery questionnaires that can be used with reliability in healthy subjects and in persons post-stroke. Both versions of the KVIQ have a high level of internal consistency, and confirmation of their bifactorial structure indicated that they assess two distinct dimensions of motor imagery: visual and kinesthetic. Given the good psychometric properties of both versions, the KVIQ-10 appears to be a good choice for assessing persons with physical disabilities because it can be administered in half the time.
The authors thank the subjects who participated in this study. They also extend their gratitude to Denis Côté for statistical analyses, Daniel Tardif for illustrations, and Cynthia Bergeron and Valérie Poulin for data collection. This work was supported by Quebec Provincial Rehabilitation Research Network (FRSQ) and the CIHR.
Contact Dr. F. Malouin at: Francine.Malouin@rea.ulaval.ca to receive a copy of the questionnaire.