Human Movement Science 35 (2014) 50–65
Contents lists available at ScienceDirect
Human Movement Science journal homepage: www.elsevier.com/locate/humov
Assessing motor imagery using the hand rotation task: Does performance change across childhood? Michael L. Butson a, Christian Hyde b, Bert Steenbergen c,d, Jacqueline Williams a,⇑ a
Institute of Sport, Exercise and Active Living and College of Sport and Exercise Science, Victoria University, Melbourne, Australia Cognitive Neuroscience Unit, School of Psychology, Deakin University, Melbourne, Australia c Radboud University Nijmegen, Behavioural Science Institute, Nijmegen, The Netherlands d Australian Catholic University, School of Psychology, Melbourne, Australia b
a r t i c l e
i n f o
Article history:
PsycINFO classification: 2330 2820 Keywords: Motor imagery Hand rotation Motor skill development Child development
a b s t r a c t This study examined at what age children can engage in the hand rotation task (as a measure of motor imagery); whether engagement changes across development and; the influence of age and motor skill on performance. Children were aged 5–12 years (N = 101; 52 girls), with no IQ or motor skill impairment. Less than 40% of 5–6 year olds completed the hand rotation with sufficient accuracy for further analysis, compared with 80% of 7–8 year olds, and 90% aged 9 and above. From age 7, either or both response time (RT) and accuracy conformed to the biomechanical constraints of corresponding physical movements. Although RT did not improve with age, accuracy did, with 11 year olds significantly more accurate than 7 and 8 year olds. Importantly, efficiency (RT/accuracy) improved with age and both age, in months, and motor skill level were significant predictors of efficiency, accounting for 35% and 8% of variability, respectively. Improvements in motor imagery ability during childhood are likely the result of increased neural efficiency, developing as the result of complex interactions between endogenous and exogenous factors. This highlights the need for a multidisciplinary approach to further our understanding of the emergence of motor imagery ability. Ó 2014 Elsevier B.V. All rights reserved.
⇑ Corresponding author. Address: College of Sport and Exercise Science, Victoria University, Footscray Park Campus, PO Box 14428, Melbourne, VIC 8001, Australia. Tel.: +61 3 9919 4025. E-mail address: [email protected] (J. Williams). http://dx.doi.org/10.1016/j.humov.2014.03.013 0167-9457/Ó 2014 Elsevier B.V. All rights reserved.
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
51
1. Introduction Motor imagery paradigms, which require individuals to produce a dynamic simulation of movement (without any overt accompanying movement), are increasingly being used to investigate the ability of an individual to mentally represent movement. Internal representations of movement are thought to play a critical role in a number of motor control processes, including mental rehearsal and observational learning (Jeannerod, 2001), motor planning (Wolpert & Ghahramani, 2000), and online movement control (Desmurget & Grafton, 2000; Izawa & Shadmehr, 2011). These representations are dynamic, in that they are constantly updated as a result of an individual’s movement interactions with their environment and changes in body kinematics that occur during development (Choudhury, Charman, Bird, & Blakemore, 2007; Miall & Wolpert, 1996; Wolpert, Ghahramani, & Jordan, 1995). It is currently unclear at what age these representations form or become consciously accessible, but it has been suggested that children must first gain some level of implicit knowledge of the relationship between the motor commands they generate, the environment and the effects on their moving body before they can accurately generate an internal representation of movement (Caeyenberghs, Wilson, Van Roon, Swinnen, & Smits-Engelsman, 2009). Understanding the emergence and development of these internal representations is crucial to our understanding of motor development and to enable us to better understand the atypical motor imagery performance of children with motor skill impairment (see below). To examine these representations, we need to be sure that the motor imagery tasks being utilized are age-appropriate. Currently, the majority of studies examining motor imagery in children have utilized tasks that have been borrowed from adult studies and though for the most part these tasks are supported by neuroimaging data that indicates they can effectively engage participants in motor imagery (e.g., de Lange, Helmich, & Toni, 2006; Kosslyn, Digirolamo, Thompson, & Alpert, 1998; Parsons, 1987; Parsons & Fox, 1998), their valid use in children is less well established. Therefore, the first aim of this experimental study was to explore whether one such task, the hand rotation task, engages children aged 5–12 years in motor imagery. To establish this, response times and accuracy to posturally congruent and posturally incongruent stimuli were compared in order to confirm the presence of biomechanically constrained movement simulations (indicating the use of motor, rather than visual, imagery: see below). Following this, if analyses were to indicate that children were effectively engaged in motor imagery while performing the hand rotation task, the study aimed to determine how performance changes across age and what influence motor skill level has on performance.
1.1. The hand rotation task A widely used task to assess motor imagery capacity is the hand rotation task (see, for example, Caeyenberghs, Tsoupas, Wilson, & Smits-Engelsman, 2009; de Lange et al., 2006; Deconinck, Spitaels, Fias, & Lenoir, 2009; Funk, Brugger, & Wilkening, 2005; Parsons, 1987, 2003; Ter Horst, Van Lier, & Steenbergen, 2010; Williams, Anderson, et al., 2011; Williams, Thomas, Maruff, Butson, & Wilson, 2006). In this task participants are required to make a decision on the laterality of a stimulus hand that can be presented at varying degrees of angular rotation, as well as in different views (back versus palm) and sometimes in different postures (Ionta & Blanke, 2009; Ter Horst et al., 2010). Neuroimaging and self-report data indicate that the task can elicit the use of motor imagery when participants imagine moving their own hand into the position of the stimulus hand to aid in determining laterality (de Lange et al., 2006; Kosslyn et al., 1998; Parsons, 1987; Parsons & Fox, 1998). For many years, researchers interpreted linear increases in response time that occurred in line with rotation of the stimulus from 0° (fingers up) to 180° (fingers down) as evidence that motor imagery was occurring (Caeyenberghs, Tsoupas, et al., 2009; Lust, Geuze, Wijers, & Wilson, 2006; Mutsaarts, Steenbergen, & Bekkering, 2007; Williams et al., 2006, 2004). However, visual imagery, a functionally and neurophysiologically distinct form of imagery involving the rotation of objects, follows the same pattern of response (i.e., response increases with angular rotation of stimulus). For studies without corroborating neuroimaging data, the challenge was to demonstrate that the task was indeed
52
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
engaging participants in motor, rather than visual, imagery. To do this, the biomechanical constraints of real movements, which are evident in motor but not visual imagery, are used as evidence to support the use of motor imagery. For example, if the participant has their hands palm down on a computer keyboard, responses to stimuli in the back view will be faster than to those in the palm view due to postural congruence (Ter Horst et al., 2010). Further, as hands are physically easier to rotate in the medial direction versus the lateral direction, responses are usually faster and more accurate for hands presented in a medial rotation than those rotated laterally when motor imagery is used (Williams, Omizzolo, Galea, & Vance, 2013). As these task design and analysis techniques are a more recent addition to the hand rotation task, few studies with children have employed them. Hence, it is difficult to confirm the use of a specific motor imagery strategy during performance of the hand rotation task with many of these older studies. Accordingly, we are limited in our interpretation of these studies and, more specifically, what we can infer about the emergence and development of motor imagery. 1.2. At what age can children engage in the hand rotation task using motor imagery? To date, the limited evidence regarding this question has been equivocal. Importantly, a fundamental consideration when interpreting hand rotation task performance in children is that both cognitive development and motor imagery capacity need to be considered. Importantly, children need to have the cognitive capacity to determine left hands from right. This places a lower limit on the age at which the hand task is suitable for use and is a factor that has not been addressed in the studies described below. A recent review by Gabbard (2009) indicated that children aged 5–6 years may be capable of accurately imagining movement, though performance is very variable. By age 7 years, performance appeared to conform to similar constraints and patterns as adults, though quality at this age is reduced and improves through childhood. Generally, when considering the hand rotation task in typically developing groups, it appears that by 7 years of age, responses are generally above chance (Caeyenberghs, Tsoupas, et al., 2009; Williams, Thomas, Maruff, & Wilson, 2008; Williams et al., 2006), but only three studies have included children younger than 7 years when utilizing the hand rotation task. In the first, Funk et al. (2005) found that 12 out of 20 children, aged 5–7 years, who attempted a hand rotation task were able to perform at a level exceeding chance. Analysis of response time and accuracy demonstrated not only that the children were performing in a way that matched that expected on mental rotation tasks, but their responses conformed to the biomechanical constraints of real movement. That is, as rotation from the participants resting hand position to the stimulus hand position became more awkward, response time increased. Interestingly, a later study by Krüger and Krist (2009) found a rotation effect (i.e., response time increased with angular disparity), but only limited motor effects in a sample of children aged 5– 6 years. In contrast, motor effects were clear in a group of 7 year old children. It is important to note however, that the task used by the authors involved presentation of a stimulus image, along with two smaller images, one left hand and one right hand, at the bottom of the screen for comparison purposes. Such a presentation, while easier for children, can present an opportunity for participants to use a visual imagery matching strategy, rather than motor imagery per se. Finally, a recent study using the hand rotation task included children aged from 5 to 17 years of age (Dey et al., 2012). Analysis of age effects was limited to correlations between age and both response time and accuracy, with only the latter being significantly related to age. However, given the lower end of the age range, the failure of the authors to analyze age effects beyond correlational analysis makes further interpretation of this data difficult. Importantly, assessment of developmental shifts in task performance profile for the hand rotation task within this large and neuro-cognitively diverse age-span is not possible. Nonetheless, taken together it appears that children as young as five may be able to engage in motor imagery to solve the hand rotation task if they have the required cognitive capacity. However, evidence is mixed and a well-designed study exploring when the capacity for imagery emerges is currently lacking.
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
53
1.3. Motor imagery in older children Motor imagery ability has been assessed using the hand rotation task more readily in older children, generally aged between 7 and 12 years (Caeyenberghs, Tsoupas, et al., 2009; Deconinck et al., 2009; Lust et al., 2006; Williams, Anderson, et al., 2011; Williams, Anderson, Reid, & Reddihough, 2012; Williams et al., 2006, 2008, 2004). Unfortunately, due to a variety of methodological and design issues, the studies provide neither a complete picture of performance nor imagery development. For example, improvements with age in children aged 7–12 years were demonstrated by Caeyenberghs, Tsoupas, et al. (2009), but the authors did not examine biomechanical effects in their data. As noted, analysis of biomechanical effects provides clear evidence as to whether motor imagery has been utilized to complete the hand rotation task. Other studies have examined biomechanical effects on both response time and accuracy (Deconinck et al., 2009; Williams, Anderson, et al., 2011; Williams, Reid, Reddihough, & Anderson, 2011), but these studies grouped children across the 7–12 years age span without examining changes with age. Notwithstanding these limitations, taken together, evidence from these studies demonstrates that children beyond 7 years of age are capable of engaging in motor imagery when completing the hand rotation task and that performance continues to improve throughout late childhood. However, without confirmation of these findings through analysis of both biomechanical and age effects on performance in a single sample, we must be circumspect when drawing conclusions regarding motor imagery performance in children based on the existing evidence. 1.4. What influence do motor skills have on imagined hand rotation performance in childhood? Remembering that theory suggests internal movement representations are updated in line with movement experience (Choudhury et al., 2007; Miall & Wolpert, 1996; Wolpert et al., 1995), it is not surprising that many studies have compared clinically motor impaired groups with non-impaired peers, with the expectation that poor motor skills result in a reduced ability to accurately update internal representations of movement. Studies have included children with Developmental Coordination Disorder (Deconinck et al., 2009; Maruff, Wilson, Trebilcock, & Currie, 1999; Williams et al., 2006; Williams et al., 2008; Wilson, Maruff, Ives, & Currie, 2001; Wilson et al., 2004), Attention Deficit Hyperactivity Disorder (Lewis, Vance, Maruff, Wilson, & Cairney, 2008; Williams et al., 2013) and congenital hemiplegia (Mutsaarts et al., 2007; Williams, Anderson, et al., 2011; Williams, Reid, et al., 2011). From these studies, we know that children between the ages of 7 and 11 years without motor impairment often perform significantly better than children with motor impairment. This suggests a relationship between motor imagery ability and motor skill level, which marries well with the expectation that internal representations of movement will be more accurate in individuals with increased exposure to a wider range of motor experiences. Still, only two studies have examined the relationship between motor imagery and motor skill level in children without motor skill impairment, with both supporting the link between motor skill level and motor imagery ability (Caeyenberghs, Tsoupas, et al., 2009; Gabbard, Cacola, & Bobbio, 2011a). Only Caeyenberghs, Tsoupas, et al. (2009) used the hand rotation task, identifying a significant positive correlation between performance accuracy and motor skills. However, the outcome of this study needs to be interpreted with some caution as the authors used unstandardized raw scores as their measure of motor ability. This makes it difficult to determine what proportion of the observed relationship resulted from improved motor performance compared with that resulting from broader developmental changes occurring with age. 1.5. Summary Our review of the literature indicates that the following points remain outstanding: (1) the age at which children can effectively engage in the hand rotation task; (2) whether the hand rotation task is an appropriate measure of motor imagery throughout childhood- i.e., whether performance on the hand rotation task throughout childhood is constrained by the biomechanical and postural constraints of real movement; (3) How motor imagery ability, as measured by the hand rotation task, changes during development and; (4) the influence motor skill level on performance on the hand rotation task.
54
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
Related to the first point, this study aimed to determine whether children below the age of 7 could effectively engage in the hand rotation task. Initially, this included assessing their performance on a baseline task which required them to identify the laterality of hands presented in the most basic view – back of hand at 0°. This allowed us to exclude children who were not yet able to accurately identify left from right hands. Based on the findings of Funk et al. (2005), we expected that approximately 50% of children in this age group who attempted the hand rotation task (after successfully completing the baseline task) would be able to complete the task with adequate levels of accuracy for analysis (operationalized as >50% correct to the back view stimulus at 0° of rotation) and their responses would conform to the biomechanical limitations of real movement. With respect to the second point, we explored whether the hand rotation task was an appropriate tool for investigating motor imagery in children aged 7–12 years by examining whether the postural constraints of real movements influenced hand rotation performance across the age groups. Specifically, we expected to see that responses would be faster and more accurate to hands rotated in the medial direction (compared to those rotated laterally) and to hands presented in the back view (posturally congruent) compared to those in the palm view (posturally incongruent). We expected these patterns to be relatively stable across age, which would be taken as an indication that the children were indeed engaging in motor imagery. After confirming the use of motor imagery, the third point was addressed by examining age-related changes in performance throughout childhood. Though categorizing age was not considered ideal (see Section 6), to enable comparison with previous research, participants were first grouped according to chronological age to examine task performance. In line with Caeyenberghs, Tsoupas, et al. (2009), we expected response times to decrease and accuracy to increase with age. To better understand development of performance on the task, we also created an ‘efficiency’ variable, by dividing response time by accuracy, to account for any speed-accuracy trade-off in the data. This variable has not been utilized previously, but we expect it to provide a more thorough description of the data than analysis of response time or accuracy alone. We also expected efficiency to improve with age. Finally, we examined the influence of age (in months, as a continuous variable) and motor skill level on performance efficiency. It was expected that age would be the biggest predictor of performance, but we expected motor skill level to further contribute to the variability of performance. 2. Method 2.1. Participants Initially, 101 children aged 5–12 years (52 girls) were recruited from a primary school in Melbourne, Australia, and by advertising throughout the university, including at university-run school holiday Sports Camps. Participants were required to have no physical or neurological condition causing motor skill impairment and no intellectual, visual, or hearing impairments. One 11 year old boy was excluded due to a diagnosis of high functioning autism and an 8 year old girl was excluded due to non-compliance throughout the assessment. Further descriptions by age group are included in the Section 7. 2.2. Measures 2.2.1. Motor skills The McCarron Assessment of Neuromuscular Development (MAND; McCarron, 1997) is a test-battery that assesses fine and gross motor skills in children from 3.5 years of age, with normative data provided through to adulthood. Age-scaled scores are used to calculate a Neuromuscular Development Index (NDI) score, with a mean of 100 and standard deviation of 15. 2.2.2. The Wechsler abbreviated scale of intelligence (WASI) The WASI (Wechsler, 1999) is an abbreviated scale of intelligence that provides an estimate of IQ. The two subtest version was used in this study, which includes the vocabulary and matrix reasoning
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
55
subtests. Unfortunately, WASI norms commence at age 6, leaving us without IQ data for the 5 year old children. 2.2.3. Motor imagery task – hand rotation Single hand stimuli were presented on a laptop computer using E-Prime software (Psychology Software tools, Pittsburgh, PA, USA). The participants were required to decide whether each stimulus, presented in the center of the screen and measuring 9 cm by 6 cm, was a left or right hand, as quickly and as accurately as possible. The stimulus remained on screen until the participant pressed one of two keys on the computer keyboard (d for left, k for right), or 10 s had elapsed. There was a random delay of 2–3 s between trials. E-Prime was used to record response time (RT) to the nearest 1 ms and response accuracy. Younger children (5–6 year olds) first completed a baseline version of the task. In this version, participants were presented with 10 stimulus hands at 0° of rotation – meaning stimulus fingers were pointing up. There were an equal number of left and right hands, which were presented in a random order. After each response, visual feedback was provided on the screen (either CORRECT in green font or INCORRECT in red font). This was provided to enable the researcher to mark down the total number correct in real time to determine whether the experimental task should be administered. Participants received two practice trials. Only participants scoring at least 6/10 for the baseline task were administered the experimental task. The experimental task was completed by all children aged 7 years and above, and those 5 and 6 year olds reaching the criterion level in the baseline task. Five practice trials preceded the presentation of 80 stimulus hands. An equal number of left and right hands were presented randomly in 45° increments between 0° and 315°, providing ten trials at each angle of rotation. These hands could be presented in either a palm or back of the hand view (five trials of each at each angle), with stimulus view (palm v back – five trials each at each angle), laterality (medial v lateral – 30 trials of each within the task) and angle were completely randomized. 2.3. Procedure The study protocol was approved by the Human Research Ethics Committee at Victoria University, Melbourne, Australia. Informed consent was obtained from participants and their parent/guardian, with each seen by a researcher one-on-one. Task order was randomized across participants. For the baseline and experimental tasks, participants were seated in front of a computer screen with their left hand on the d key and the right hand on the k key. They were instructed to imagine their own hand in the position of the stimulus that appeared and to use this to guide their decision as to whether the stimulus on the screen was a left or right hand. The participants were encouraged to respond as quickly and as accurately as possible. 2.4. Analysis Data were analyzed using SPSS 19.0 statistical software. Group means for motor and IQ scores were submitted to univariate ANOVA. For the experimental hand rotation task, each participant’s mean RT and accuracy (proportion correct) was calculated for each angle, as well as for palm and back view and medial and lateral rotations. Only children scoring above 50% (or 0.5) for accuracy when responding to hands presented in the back view at 0° (deemed the easiest stimulus presentation) were included in further analyses. This cut-off was chosen rather than a value that was significantly above chance level due to the variability in performance of some children – for example, some children performed very well when responding to hands in the back view versus hands in the palm view. This may have left them with overall values below that designated as significantly above chance, but was valuable data that we deemed important to include. After excluding children with poor accuracy, only limited data was available for the 5 and 6 year old children, thereby precluding statistical analysis. Instead, descriptive information was calculated for RT and accuracy for back (medial and lateral rotations) and palm view (medial and lateral rotations) to determine whether responses were constrained by biomechanical factors constraining real movement.
56
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
For older children, two repeated measures ANOVAs were used to examine the effect of biomechanical constraints on RT and accuracy individually (i.e., Direction (2: medial v lateral) View (2: palm v back) Age). Significant findings were followed up with pairwise comparisons with Sidak corrections for multiple comparisons. In line with other research (see, for example, Harris et al., 2000; Roelofs, van Galen, Keijsers, & Hoogduin, 2002; Williams, Reid, et al., 2011), general hand rotation performance was analyzed by combining palm and back views and collapsing medial and lateral rotations to provide mean values for responses from 0° to 180° (45° increments; 16 trials per angle). Both RT and accuracy were submitted to repeated measures ANOVAs with age group as a between-subjects measure. Finally, an efficiency measure was calculated by dividing individual RT values by accuracy and then calculating individual means at each angle from 0° to 180°. The resulting efficiency values were also submitted to repeated measures ANOVA with age group as a between-subjects measure. Of note, this type of efficiency measure, commonly referred to as the Inverse Efficiency Score (IES), has been criticised when accuracy levels are below 90%, as is the case in the current data set (Bruyer & Brysbaert, 2011). Bruyer and Brysbaert suggest careful inspection of the data when using IES to ensure that it accurately portrays the relationship between accuracy and response time. Inspection of Fig. 3 in the results indicates that the efficiency measure here does accurately reflect the relationship between accuracy and response time. In addition, we calculated efficiency using two other methods – one which transformed the variables and a second which standardized them, prior to performing an efficiency calculation. In both instances, the resulting efficiency values produced results that were very similar, though slightly stronger, to the IES. Thus, the more conservative IES measure was chosen to describe efficiency, as this is more commonly used to describe speed-accuracy trade-offs and will likely be more comparable in the future. While multiple comparisons increase the risk of Type I error, correcting the family-wise error rate unduly increases the likelihood of committing a Type II error (see O’Keefe, 2003; Saville, 1990). Accordingly, adjustments to the standard alpha level of p < .05 were not made during analysis. To determine whether age and motor skill level would predict efficiency on the hand rotation task, a two-step hierarchical regression analysis was employed. Age was entered as a continuous variable, to avoid the reduced power and residual confounding that occurs when artificially dichotomizing continuous variables (Royston, Altman, & Sauerbrei, 2006). In the first step, age in months, expected to be the strongest predictor of efficiency, was entered into the regression. In the second step, MAND NDI was added, allowing us to determine if it also made a unique contribution to efficiency after accounting for the effect of age.
3. Results 3.1. Sample characteristics Participant information is provided Table 1. Neither NDI nor IQ differed between ages (p = .79 and .66, respectively). Due to the small N for the 12 year olds, these children were excluded from further analysis.
Table 1 Sample characteristics by age. Age group
N % Male NDI M SD IQ M SD
5 years
6 years
7 years
8 years
9 years
10 years
11 years
12 years
7 71.4 102.86 16.01 – –
11 36.4 104.50 16.64 101.44 12.43
15 60 104.07 19.61 105.85 8.12
11 45.5 97.73 13.61 99.00 10.94
16 43.8 94.69 15.60 107.10 15.01
20 45 96.95 18.19 106.41 9.61
17 52.9 96.41 13.54 104.79 12.60
2 0 95.00 8.49 107.50 2.12
Note: NDI = Neurodevelopmental Index from MAND; IQ = estimate of intelligence from WASI (norms begin at age six).
57
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
3.2. Motor imagery performance – 5 and 6 year olds Five and six year old children initially completed the baseline hand task. Data was missing for one 5 and one 6 year old due to computer malfunction. Of the 6 remaining 5 year olds, three (50%) reached the criterion of an accuracy score of P6/10 correct. Of the 10 remaining 6 year olds, eight (80%) reached the same level. These 11 children then attempted the hand rotation task. One 5 year old and three 6 year olds commenced the task, but it was later aborted as it was clear that they were not responding to the stimuli. Of the seven children who completed the task, two 6 year olds failed to score above 50% for the trials where the hand was presented in back view at 0° (accuracy levels of 38% and 50%; see Fig. 1 for percentage of children completing the task in each age group). The accuracy of the remaining five children ranged from 75% to 100% (one 5 year old and one 6 year old achieved 100%), but only two (one of each age) scored above 50% on the task when hands from both views and all angles of rotation were combined. The children completing the hand rotation task had a similar MAND NDI (M = 102) to those who did not (M = 104). Table 2 presents descriptive data for these five children, comparing RT and accuracy of the different views and rotation directions. Response accuracy is very low except for when hands were presented in the simplest view (back view and medial rotations), but there are clear differences in participants responses that appear to reflect biomechanical constraints. Palm view responses, for example, are less accurate than back view (remembering that participant’s hands were in a posture compatible with back view) and medially rotated hands were more accurately recognized than laterally rotated hands. RT showed similar effects, with the exception of the lateral rotations in palm view – the reduced RT here might reflect the difficulty of this view and an increased likelihood of participants simply guessing.
Fig. 1. Percentage of children in each age group who reached the inclusion criterion of accuracy greater than 50% when responding to hands in the back view at 0° of rotation.
Table 2 Hand rotation task performance for five children aged 5 and 6 years who reached the designated level on the baseline hand task (N = 5). View/rotation direction
Response time (ms)
Accuracy (% correct)
Back view, medial rotation Back view, lateral rotation Palm view, medial rotation Palm view, lateral rotation
1904.25 2400.82 2241.82 1732.32
79.90 56.70 46.70 23.40
(825.50) (1115.89) (953.34) (552.32)
(13.79) (27.34) (21.75) (15.01)
58
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
Fig. 2. Response time (left) and accuracy (right) to medially versus laterally rotated stimuli (top) and palm versus back view stimuli (bottom). Error bars represent ±2 SD.
3.3. Motor imagery performance – 7 to 11 year olds Of the 79 children aged 7–11 years, 8 were excluded after failing to score above 50% when responding to hands presented in the back view at 0° of rotation. This included three 7 year olds, two 8 year olds, one 9 year old and two 10 year olds (see Fig. 1). Biomechanical constraints are demonstrated in Fig. 2. A repeated measures ANOVA on RT identified a significant effect of view, Wilks’ K = .46, F (1,66) = 76.79, p < .001, gP2 = .54, with responses significantly faster to hands presented in the back, compared with palm, view (p < .001). There was also a significant interaction between direction and age, Wilks’ K = .82, F (4,66) = 3.54, p = .011, gP2 = .18. Pairwise comparisons demonstrated that this interaction resulted from the significant effect of direction, where responses to stimuli rotated in the medial direction were faster than to stimuli rotated laterally, in the 8, 9 and 11 year old groups (p = .006, .003 and .05). For accuracy, repeated measures ANOVA identified a significant effect of view, Wilks’ K = .55, F (1,66) = 53.66, p < .001, gP2 = .45, and direction, Wilks’ K = .74, F (1,66) = 23.30, p < .001, gP2 = .26, as well as for age, F (4,66) = 5.25, p = .001, gP2 = .24. There were no significant interactions among
59
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
Fig. 3. Response time (left), accuracy (middle) and efficiency (right) averaged across views and angles for the hand rotation task for children aged 7–11 years.
variables. Responses were more accurate to hands presented in the back, compared with palm, view (p < .001), and to those rotated in the medial, compared to lateral, direction (p < .001). Regarding age, the 11 year olds were significantly more accurate than the 7 and 8 year olds (p = .006 and .003, respectively). General hand rotation task performance (combining both back and palm view and collapsing medial and lateral stimuli) for the children aged 7–11 years can be seen in Fig. 3. Repeated measures ANOVA demonstrated that there was a significant effect of angle of rotation on response time, Wilks’ K = .34, F (4,63) = 30.60, p < .001, gP2 = .66, but there was no effect for age (p = .12) or interaction between the two (p = .76). For accuracy, a repeated measures ANOVA also demonstrated a significant effect for angle of rotation, Wilks’ K = .60, F (4,63) = 10.44, p < .001, gP2 = .40, and also for age, F (4,66) = 5.85, p < .001, gP2 = .26, but there was no interaction between the two (p = .94). The 11 year olds were significantly more accurate than the 7 and 8 year olds (both p = .002). Finally, a repeated measures ANOVA demonstrated that there was a significant effect for both angle, Wilks’ K = .44, F (4,63) = 20.14, p < .001, gP2 = .56, and age, F (4,66) = 4.23, p = .004, gP2 = .20, on efficiency, but there was no interaction between the two (p = .99). Across groups, efficiency scores were significantly lower (indicating greater efficiency) at 0° and 45° compared to other angles (all p < .05) and also at 90° compared to 180° (p < .001). Across angles, the 11 year olds were more efficient than the 8 year olds (p = .022) and there was a trend toward the same when the 11 year olds were compared to the 7 and 9 year olds (p = .050 and .051, respectively). A two-step hierarchical regression analysis was performed to determine whether age and motor skill level would predict efficiency on the hand rotation task. In the first step, age was demonstrated to be a significant predictor of performance, accounting for 29% of the variance in hand rotation efficiency, F (1,66) = 26.93, p < .001, R2 = .29. Next, MAND NDI was added as a predictor variable in a second step to determine whether, after accounting for age, motor skill level would influence hand rotation efficiency. This model accounted for 38% of the variability in hand rotation efficiency, F Table 3 Summary of hierarchical regression results with hand rotation efficiency as the outcome variable and Age (in months), and Neurodevelopmental Index (NDI) as the predictor variables (N = 71). Step
Predictor variable
1
DR2
ANOVA DF
Sign. DF
df
.29
26.93
.000
1,66
.089
9.29
.003
1,65
Age (in months) 2 Age (in months) NDI Note: S2 = Squared semi-partial correlations.
b
Regression SE b
t
b
p
S2
31.96
6.16
5.19
.54
Contents lists available at ScienceDirect
Human Movement Science journal homepage: www.elsevier.com/locate/humov
Assessing motor imagery using the hand rotation task: Does performance change across childhood? Michael L. Butson a, Christian Hyde b, Bert Steenbergen c,d, Jacqueline Williams a,⇑ a
Institute of Sport, Exercise and Active Living and College of Sport and Exercise Science, Victoria University, Melbourne, Australia Cognitive Neuroscience Unit, School of Psychology, Deakin University, Melbourne, Australia c Radboud University Nijmegen, Behavioural Science Institute, Nijmegen, The Netherlands d Australian Catholic University, School of Psychology, Melbourne, Australia b
a r t i c l e
i n f o
Article history:
PsycINFO classification: 2330 2820 Keywords: Motor imagery Hand rotation Motor skill development Child development
a b s t r a c t This study examined at what age children can engage in the hand rotation task (as a measure of motor imagery); whether engagement changes across development and; the influence of age and motor skill on performance. Children were aged 5–12 years (N = 101; 52 girls), with no IQ or motor skill impairment. Less than 40% of 5–6 year olds completed the hand rotation with sufficient accuracy for further analysis, compared with 80% of 7–8 year olds, and 90% aged 9 and above. From age 7, either or both response time (RT) and accuracy conformed to the biomechanical constraints of corresponding physical movements. Although RT did not improve with age, accuracy did, with 11 year olds significantly more accurate than 7 and 8 year olds. Importantly, efficiency (RT/accuracy) improved with age and both age, in months, and motor skill level were significant predictors of efficiency, accounting for 35% and 8% of variability, respectively. Improvements in motor imagery ability during childhood are likely the result of increased neural efficiency, developing as the result of complex interactions between endogenous and exogenous factors. This highlights the need for a multidisciplinary approach to further our understanding of the emergence of motor imagery ability. Ó 2014 Elsevier B.V. All rights reserved.
⇑ Corresponding author. Address: College of Sport and Exercise Science, Victoria University, Footscray Park Campus, PO Box 14428, Melbourne, VIC 8001, Australia. Tel.: +61 3 9919 4025. E-mail address: [email protected] (J. Williams). http://dx.doi.org/10.1016/j.humov.2014.03.013 0167-9457/Ó 2014 Elsevier B.V. All rights reserved.
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
51
1. Introduction Motor imagery paradigms, which require individuals to produce a dynamic simulation of movement (without any overt accompanying movement), are increasingly being used to investigate the ability of an individual to mentally represent movement. Internal representations of movement are thought to play a critical role in a number of motor control processes, including mental rehearsal and observational learning (Jeannerod, 2001), motor planning (Wolpert & Ghahramani, 2000), and online movement control (Desmurget & Grafton, 2000; Izawa & Shadmehr, 2011). These representations are dynamic, in that they are constantly updated as a result of an individual’s movement interactions with their environment and changes in body kinematics that occur during development (Choudhury, Charman, Bird, & Blakemore, 2007; Miall & Wolpert, 1996; Wolpert, Ghahramani, & Jordan, 1995). It is currently unclear at what age these representations form or become consciously accessible, but it has been suggested that children must first gain some level of implicit knowledge of the relationship between the motor commands they generate, the environment and the effects on their moving body before they can accurately generate an internal representation of movement (Caeyenberghs, Wilson, Van Roon, Swinnen, & Smits-Engelsman, 2009). Understanding the emergence and development of these internal representations is crucial to our understanding of motor development and to enable us to better understand the atypical motor imagery performance of children with motor skill impairment (see below). To examine these representations, we need to be sure that the motor imagery tasks being utilized are age-appropriate. Currently, the majority of studies examining motor imagery in children have utilized tasks that have been borrowed from adult studies and though for the most part these tasks are supported by neuroimaging data that indicates they can effectively engage participants in motor imagery (e.g., de Lange, Helmich, & Toni, 2006; Kosslyn, Digirolamo, Thompson, & Alpert, 1998; Parsons, 1987; Parsons & Fox, 1998), their valid use in children is less well established. Therefore, the first aim of this experimental study was to explore whether one such task, the hand rotation task, engages children aged 5–12 years in motor imagery. To establish this, response times and accuracy to posturally congruent and posturally incongruent stimuli were compared in order to confirm the presence of biomechanically constrained movement simulations (indicating the use of motor, rather than visual, imagery: see below). Following this, if analyses were to indicate that children were effectively engaged in motor imagery while performing the hand rotation task, the study aimed to determine how performance changes across age and what influence motor skill level has on performance.
1.1. The hand rotation task A widely used task to assess motor imagery capacity is the hand rotation task (see, for example, Caeyenberghs, Tsoupas, Wilson, & Smits-Engelsman, 2009; de Lange et al., 2006; Deconinck, Spitaels, Fias, & Lenoir, 2009; Funk, Brugger, & Wilkening, 2005; Parsons, 1987, 2003; Ter Horst, Van Lier, & Steenbergen, 2010; Williams, Anderson, et al., 2011; Williams, Thomas, Maruff, Butson, & Wilson, 2006). In this task participants are required to make a decision on the laterality of a stimulus hand that can be presented at varying degrees of angular rotation, as well as in different views (back versus palm) and sometimes in different postures (Ionta & Blanke, 2009; Ter Horst et al., 2010). Neuroimaging and self-report data indicate that the task can elicit the use of motor imagery when participants imagine moving their own hand into the position of the stimulus hand to aid in determining laterality (de Lange et al., 2006; Kosslyn et al., 1998; Parsons, 1987; Parsons & Fox, 1998). For many years, researchers interpreted linear increases in response time that occurred in line with rotation of the stimulus from 0° (fingers up) to 180° (fingers down) as evidence that motor imagery was occurring (Caeyenberghs, Tsoupas, et al., 2009; Lust, Geuze, Wijers, & Wilson, 2006; Mutsaarts, Steenbergen, & Bekkering, 2007; Williams et al., 2006, 2004). However, visual imagery, a functionally and neurophysiologically distinct form of imagery involving the rotation of objects, follows the same pattern of response (i.e., response increases with angular rotation of stimulus). For studies without corroborating neuroimaging data, the challenge was to demonstrate that the task was indeed
52
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
engaging participants in motor, rather than visual, imagery. To do this, the biomechanical constraints of real movements, which are evident in motor but not visual imagery, are used as evidence to support the use of motor imagery. For example, if the participant has their hands palm down on a computer keyboard, responses to stimuli in the back view will be faster than to those in the palm view due to postural congruence (Ter Horst et al., 2010). Further, as hands are physically easier to rotate in the medial direction versus the lateral direction, responses are usually faster and more accurate for hands presented in a medial rotation than those rotated laterally when motor imagery is used (Williams, Omizzolo, Galea, & Vance, 2013). As these task design and analysis techniques are a more recent addition to the hand rotation task, few studies with children have employed them. Hence, it is difficult to confirm the use of a specific motor imagery strategy during performance of the hand rotation task with many of these older studies. Accordingly, we are limited in our interpretation of these studies and, more specifically, what we can infer about the emergence and development of motor imagery. 1.2. At what age can children engage in the hand rotation task using motor imagery? To date, the limited evidence regarding this question has been equivocal. Importantly, a fundamental consideration when interpreting hand rotation task performance in children is that both cognitive development and motor imagery capacity need to be considered. Importantly, children need to have the cognitive capacity to determine left hands from right. This places a lower limit on the age at which the hand task is suitable for use and is a factor that has not been addressed in the studies described below. A recent review by Gabbard (2009) indicated that children aged 5–6 years may be capable of accurately imagining movement, though performance is very variable. By age 7 years, performance appeared to conform to similar constraints and patterns as adults, though quality at this age is reduced and improves through childhood. Generally, when considering the hand rotation task in typically developing groups, it appears that by 7 years of age, responses are generally above chance (Caeyenberghs, Tsoupas, et al., 2009; Williams, Thomas, Maruff, & Wilson, 2008; Williams et al., 2006), but only three studies have included children younger than 7 years when utilizing the hand rotation task. In the first, Funk et al. (2005) found that 12 out of 20 children, aged 5–7 years, who attempted a hand rotation task were able to perform at a level exceeding chance. Analysis of response time and accuracy demonstrated not only that the children were performing in a way that matched that expected on mental rotation tasks, but their responses conformed to the biomechanical constraints of real movement. That is, as rotation from the participants resting hand position to the stimulus hand position became more awkward, response time increased. Interestingly, a later study by Krüger and Krist (2009) found a rotation effect (i.e., response time increased with angular disparity), but only limited motor effects in a sample of children aged 5– 6 years. In contrast, motor effects were clear in a group of 7 year old children. It is important to note however, that the task used by the authors involved presentation of a stimulus image, along with two smaller images, one left hand and one right hand, at the bottom of the screen for comparison purposes. Such a presentation, while easier for children, can present an opportunity for participants to use a visual imagery matching strategy, rather than motor imagery per se. Finally, a recent study using the hand rotation task included children aged from 5 to 17 years of age (Dey et al., 2012). Analysis of age effects was limited to correlations between age and both response time and accuracy, with only the latter being significantly related to age. However, given the lower end of the age range, the failure of the authors to analyze age effects beyond correlational analysis makes further interpretation of this data difficult. Importantly, assessment of developmental shifts in task performance profile for the hand rotation task within this large and neuro-cognitively diverse age-span is not possible. Nonetheless, taken together it appears that children as young as five may be able to engage in motor imagery to solve the hand rotation task if they have the required cognitive capacity. However, evidence is mixed and a well-designed study exploring when the capacity for imagery emerges is currently lacking.
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
53
1.3. Motor imagery in older children Motor imagery ability has been assessed using the hand rotation task more readily in older children, generally aged between 7 and 12 years (Caeyenberghs, Tsoupas, et al., 2009; Deconinck et al., 2009; Lust et al., 2006; Williams, Anderson, et al., 2011; Williams, Anderson, Reid, & Reddihough, 2012; Williams et al., 2006, 2008, 2004). Unfortunately, due to a variety of methodological and design issues, the studies provide neither a complete picture of performance nor imagery development. For example, improvements with age in children aged 7–12 years were demonstrated by Caeyenberghs, Tsoupas, et al. (2009), but the authors did not examine biomechanical effects in their data. As noted, analysis of biomechanical effects provides clear evidence as to whether motor imagery has been utilized to complete the hand rotation task. Other studies have examined biomechanical effects on both response time and accuracy (Deconinck et al., 2009; Williams, Anderson, et al., 2011; Williams, Reid, Reddihough, & Anderson, 2011), but these studies grouped children across the 7–12 years age span without examining changes with age. Notwithstanding these limitations, taken together, evidence from these studies demonstrates that children beyond 7 years of age are capable of engaging in motor imagery when completing the hand rotation task and that performance continues to improve throughout late childhood. However, without confirmation of these findings through analysis of both biomechanical and age effects on performance in a single sample, we must be circumspect when drawing conclusions regarding motor imagery performance in children based on the existing evidence. 1.4. What influence do motor skills have on imagined hand rotation performance in childhood? Remembering that theory suggests internal movement representations are updated in line with movement experience (Choudhury et al., 2007; Miall & Wolpert, 1996; Wolpert et al., 1995), it is not surprising that many studies have compared clinically motor impaired groups with non-impaired peers, with the expectation that poor motor skills result in a reduced ability to accurately update internal representations of movement. Studies have included children with Developmental Coordination Disorder (Deconinck et al., 2009; Maruff, Wilson, Trebilcock, & Currie, 1999; Williams et al., 2006; Williams et al., 2008; Wilson, Maruff, Ives, & Currie, 2001; Wilson et al., 2004), Attention Deficit Hyperactivity Disorder (Lewis, Vance, Maruff, Wilson, & Cairney, 2008; Williams et al., 2013) and congenital hemiplegia (Mutsaarts et al., 2007; Williams, Anderson, et al., 2011; Williams, Reid, et al., 2011). From these studies, we know that children between the ages of 7 and 11 years without motor impairment often perform significantly better than children with motor impairment. This suggests a relationship between motor imagery ability and motor skill level, which marries well with the expectation that internal representations of movement will be more accurate in individuals with increased exposure to a wider range of motor experiences. Still, only two studies have examined the relationship between motor imagery and motor skill level in children without motor skill impairment, with both supporting the link between motor skill level and motor imagery ability (Caeyenberghs, Tsoupas, et al., 2009; Gabbard, Cacola, & Bobbio, 2011a). Only Caeyenberghs, Tsoupas, et al. (2009) used the hand rotation task, identifying a significant positive correlation between performance accuracy and motor skills. However, the outcome of this study needs to be interpreted with some caution as the authors used unstandardized raw scores as their measure of motor ability. This makes it difficult to determine what proportion of the observed relationship resulted from improved motor performance compared with that resulting from broader developmental changes occurring with age. 1.5. Summary Our review of the literature indicates that the following points remain outstanding: (1) the age at which children can effectively engage in the hand rotation task; (2) whether the hand rotation task is an appropriate measure of motor imagery throughout childhood- i.e., whether performance on the hand rotation task throughout childhood is constrained by the biomechanical and postural constraints of real movement; (3) How motor imagery ability, as measured by the hand rotation task, changes during development and; (4) the influence motor skill level on performance on the hand rotation task.
54
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
Related to the first point, this study aimed to determine whether children below the age of 7 could effectively engage in the hand rotation task. Initially, this included assessing their performance on a baseline task which required them to identify the laterality of hands presented in the most basic view – back of hand at 0°. This allowed us to exclude children who were not yet able to accurately identify left from right hands. Based on the findings of Funk et al. (2005), we expected that approximately 50% of children in this age group who attempted the hand rotation task (after successfully completing the baseline task) would be able to complete the task with adequate levels of accuracy for analysis (operationalized as >50% correct to the back view stimulus at 0° of rotation) and their responses would conform to the biomechanical limitations of real movement. With respect to the second point, we explored whether the hand rotation task was an appropriate tool for investigating motor imagery in children aged 7–12 years by examining whether the postural constraints of real movements influenced hand rotation performance across the age groups. Specifically, we expected to see that responses would be faster and more accurate to hands rotated in the medial direction (compared to those rotated laterally) and to hands presented in the back view (posturally congruent) compared to those in the palm view (posturally incongruent). We expected these patterns to be relatively stable across age, which would be taken as an indication that the children were indeed engaging in motor imagery. After confirming the use of motor imagery, the third point was addressed by examining age-related changes in performance throughout childhood. Though categorizing age was not considered ideal (see Section 6), to enable comparison with previous research, participants were first grouped according to chronological age to examine task performance. In line with Caeyenberghs, Tsoupas, et al. (2009), we expected response times to decrease and accuracy to increase with age. To better understand development of performance on the task, we also created an ‘efficiency’ variable, by dividing response time by accuracy, to account for any speed-accuracy trade-off in the data. This variable has not been utilized previously, but we expect it to provide a more thorough description of the data than analysis of response time or accuracy alone. We also expected efficiency to improve with age. Finally, we examined the influence of age (in months, as a continuous variable) and motor skill level on performance efficiency. It was expected that age would be the biggest predictor of performance, but we expected motor skill level to further contribute to the variability of performance. 2. Method 2.1. Participants Initially, 101 children aged 5–12 years (52 girls) were recruited from a primary school in Melbourne, Australia, and by advertising throughout the university, including at university-run school holiday Sports Camps. Participants were required to have no physical or neurological condition causing motor skill impairment and no intellectual, visual, or hearing impairments. One 11 year old boy was excluded due to a diagnosis of high functioning autism and an 8 year old girl was excluded due to non-compliance throughout the assessment. Further descriptions by age group are included in the Section 7. 2.2. Measures 2.2.1. Motor skills The McCarron Assessment of Neuromuscular Development (MAND; McCarron, 1997) is a test-battery that assesses fine and gross motor skills in children from 3.5 years of age, with normative data provided through to adulthood. Age-scaled scores are used to calculate a Neuromuscular Development Index (NDI) score, with a mean of 100 and standard deviation of 15. 2.2.2. The Wechsler abbreviated scale of intelligence (WASI) The WASI (Wechsler, 1999) is an abbreviated scale of intelligence that provides an estimate of IQ. The two subtest version was used in this study, which includes the vocabulary and matrix reasoning
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
55
subtests. Unfortunately, WASI norms commence at age 6, leaving us without IQ data for the 5 year old children. 2.2.3. Motor imagery task – hand rotation Single hand stimuli were presented on a laptop computer using E-Prime software (Psychology Software tools, Pittsburgh, PA, USA). The participants were required to decide whether each stimulus, presented in the center of the screen and measuring 9 cm by 6 cm, was a left or right hand, as quickly and as accurately as possible. The stimulus remained on screen until the participant pressed one of two keys on the computer keyboard (d for left, k for right), or 10 s had elapsed. There was a random delay of 2–3 s between trials. E-Prime was used to record response time (RT) to the nearest 1 ms and response accuracy. Younger children (5–6 year olds) first completed a baseline version of the task. In this version, participants were presented with 10 stimulus hands at 0° of rotation – meaning stimulus fingers were pointing up. There were an equal number of left and right hands, which were presented in a random order. After each response, visual feedback was provided on the screen (either CORRECT in green font or INCORRECT in red font). This was provided to enable the researcher to mark down the total number correct in real time to determine whether the experimental task should be administered. Participants received two practice trials. Only participants scoring at least 6/10 for the baseline task were administered the experimental task. The experimental task was completed by all children aged 7 years and above, and those 5 and 6 year olds reaching the criterion level in the baseline task. Five practice trials preceded the presentation of 80 stimulus hands. An equal number of left and right hands were presented randomly in 45° increments between 0° and 315°, providing ten trials at each angle of rotation. These hands could be presented in either a palm or back of the hand view (five trials of each at each angle), with stimulus view (palm v back – five trials each at each angle), laterality (medial v lateral – 30 trials of each within the task) and angle were completely randomized. 2.3. Procedure The study protocol was approved by the Human Research Ethics Committee at Victoria University, Melbourne, Australia. Informed consent was obtained from participants and their parent/guardian, with each seen by a researcher one-on-one. Task order was randomized across participants. For the baseline and experimental tasks, participants were seated in front of a computer screen with their left hand on the d key and the right hand on the k key. They were instructed to imagine their own hand in the position of the stimulus that appeared and to use this to guide their decision as to whether the stimulus on the screen was a left or right hand. The participants were encouraged to respond as quickly and as accurately as possible. 2.4. Analysis Data were analyzed using SPSS 19.0 statistical software. Group means for motor and IQ scores were submitted to univariate ANOVA. For the experimental hand rotation task, each participant’s mean RT and accuracy (proportion correct) was calculated for each angle, as well as for palm and back view and medial and lateral rotations. Only children scoring above 50% (or 0.5) for accuracy when responding to hands presented in the back view at 0° (deemed the easiest stimulus presentation) were included in further analyses. This cut-off was chosen rather than a value that was significantly above chance level due to the variability in performance of some children – for example, some children performed very well when responding to hands in the back view versus hands in the palm view. This may have left them with overall values below that designated as significantly above chance, but was valuable data that we deemed important to include. After excluding children with poor accuracy, only limited data was available for the 5 and 6 year old children, thereby precluding statistical analysis. Instead, descriptive information was calculated for RT and accuracy for back (medial and lateral rotations) and palm view (medial and lateral rotations) to determine whether responses were constrained by biomechanical factors constraining real movement.
56
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
For older children, two repeated measures ANOVAs were used to examine the effect of biomechanical constraints on RT and accuracy individually (i.e., Direction (2: medial v lateral) View (2: palm v back) Age). Significant findings were followed up with pairwise comparisons with Sidak corrections for multiple comparisons. In line with other research (see, for example, Harris et al., 2000; Roelofs, van Galen, Keijsers, & Hoogduin, 2002; Williams, Reid, et al., 2011), general hand rotation performance was analyzed by combining palm and back views and collapsing medial and lateral rotations to provide mean values for responses from 0° to 180° (45° increments; 16 trials per angle). Both RT and accuracy were submitted to repeated measures ANOVAs with age group as a between-subjects measure. Finally, an efficiency measure was calculated by dividing individual RT values by accuracy and then calculating individual means at each angle from 0° to 180°. The resulting efficiency values were also submitted to repeated measures ANOVA with age group as a between-subjects measure. Of note, this type of efficiency measure, commonly referred to as the Inverse Efficiency Score (IES), has been criticised when accuracy levels are below 90%, as is the case in the current data set (Bruyer & Brysbaert, 2011). Bruyer and Brysbaert suggest careful inspection of the data when using IES to ensure that it accurately portrays the relationship between accuracy and response time. Inspection of Fig. 3 in the results indicates that the efficiency measure here does accurately reflect the relationship between accuracy and response time. In addition, we calculated efficiency using two other methods – one which transformed the variables and a second which standardized them, prior to performing an efficiency calculation. In both instances, the resulting efficiency values produced results that were very similar, though slightly stronger, to the IES. Thus, the more conservative IES measure was chosen to describe efficiency, as this is more commonly used to describe speed-accuracy trade-offs and will likely be more comparable in the future. While multiple comparisons increase the risk of Type I error, correcting the family-wise error rate unduly increases the likelihood of committing a Type II error (see O’Keefe, 2003; Saville, 1990). Accordingly, adjustments to the standard alpha level of p < .05 were not made during analysis. To determine whether age and motor skill level would predict efficiency on the hand rotation task, a two-step hierarchical regression analysis was employed. Age was entered as a continuous variable, to avoid the reduced power and residual confounding that occurs when artificially dichotomizing continuous variables (Royston, Altman, & Sauerbrei, 2006). In the first step, age in months, expected to be the strongest predictor of efficiency, was entered into the regression. In the second step, MAND NDI was added, allowing us to determine if it also made a unique contribution to efficiency after accounting for the effect of age.
3. Results 3.1. Sample characteristics Participant information is provided Table 1. Neither NDI nor IQ differed between ages (p = .79 and .66, respectively). Due to the small N for the 12 year olds, these children were excluded from further analysis.
Table 1 Sample characteristics by age. Age group
N % Male NDI M SD IQ M SD
5 years
6 years
7 years
8 years
9 years
10 years
11 years
12 years
7 71.4 102.86 16.01 – –
11 36.4 104.50 16.64 101.44 12.43
15 60 104.07 19.61 105.85 8.12
11 45.5 97.73 13.61 99.00 10.94
16 43.8 94.69 15.60 107.10 15.01
20 45 96.95 18.19 106.41 9.61
17 52.9 96.41 13.54 104.79 12.60
2 0 95.00 8.49 107.50 2.12
Note: NDI = Neurodevelopmental Index from MAND; IQ = estimate of intelligence from WASI (norms begin at age six).
57
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
3.2. Motor imagery performance – 5 and 6 year olds Five and six year old children initially completed the baseline hand task. Data was missing for one 5 and one 6 year old due to computer malfunction. Of the 6 remaining 5 year olds, three (50%) reached the criterion of an accuracy score of P6/10 correct. Of the 10 remaining 6 year olds, eight (80%) reached the same level. These 11 children then attempted the hand rotation task. One 5 year old and three 6 year olds commenced the task, but it was later aborted as it was clear that they were not responding to the stimuli. Of the seven children who completed the task, two 6 year olds failed to score above 50% for the trials where the hand was presented in back view at 0° (accuracy levels of 38% and 50%; see Fig. 1 for percentage of children completing the task in each age group). The accuracy of the remaining five children ranged from 75% to 100% (one 5 year old and one 6 year old achieved 100%), but only two (one of each age) scored above 50% on the task when hands from both views and all angles of rotation were combined. The children completing the hand rotation task had a similar MAND NDI (M = 102) to those who did not (M = 104). Table 2 presents descriptive data for these five children, comparing RT and accuracy of the different views and rotation directions. Response accuracy is very low except for when hands were presented in the simplest view (back view and medial rotations), but there are clear differences in participants responses that appear to reflect biomechanical constraints. Palm view responses, for example, are less accurate than back view (remembering that participant’s hands were in a posture compatible with back view) and medially rotated hands were more accurately recognized than laterally rotated hands. RT showed similar effects, with the exception of the lateral rotations in palm view – the reduced RT here might reflect the difficulty of this view and an increased likelihood of participants simply guessing.
Fig. 1. Percentage of children in each age group who reached the inclusion criterion of accuracy greater than 50% when responding to hands in the back view at 0° of rotation.
Table 2 Hand rotation task performance for five children aged 5 and 6 years who reached the designated level on the baseline hand task (N = 5). View/rotation direction
Response time (ms)
Accuracy (% correct)
Back view, medial rotation Back view, lateral rotation Palm view, medial rotation Palm view, lateral rotation
1904.25 2400.82 2241.82 1732.32
79.90 56.70 46.70 23.40
(825.50) (1115.89) (953.34) (552.32)
(13.79) (27.34) (21.75) (15.01)
58
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
Fig. 2. Response time (left) and accuracy (right) to medially versus laterally rotated stimuli (top) and palm versus back view stimuli (bottom). Error bars represent ±2 SD.
3.3. Motor imagery performance – 7 to 11 year olds Of the 79 children aged 7–11 years, 8 were excluded after failing to score above 50% when responding to hands presented in the back view at 0° of rotation. This included three 7 year olds, two 8 year olds, one 9 year old and two 10 year olds (see Fig. 1). Biomechanical constraints are demonstrated in Fig. 2. A repeated measures ANOVA on RT identified a significant effect of view, Wilks’ K = .46, F (1,66) = 76.79, p < .001, gP2 = .54, with responses significantly faster to hands presented in the back, compared with palm, view (p < .001). There was also a significant interaction between direction and age, Wilks’ K = .82, F (4,66) = 3.54, p = .011, gP2 = .18. Pairwise comparisons demonstrated that this interaction resulted from the significant effect of direction, where responses to stimuli rotated in the medial direction were faster than to stimuli rotated laterally, in the 8, 9 and 11 year old groups (p = .006, .003 and .05). For accuracy, repeated measures ANOVA identified a significant effect of view, Wilks’ K = .55, F (1,66) = 53.66, p < .001, gP2 = .45, and direction, Wilks’ K = .74, F (1,66) = 23.30, p < .001, gP2 = .26, as well as for age, F (4,66) = 5.25, p = .001, gP2 = .24. There were no significant interactions among
59
M.L. Butson et al. / Human Movement Science 35 (2014) 50–65
Fig. 3. Response time (left), accuracy (middle) and efficiency (right) averaged across views and angles for the hand rotation task for children aged 7–11 years.
variables. Responses were more accurate to hands presented in the back, compared with palm, view (p < .001), and to those rotated in the medial, compared to lateral, direction (p < .001). Regarding age, the 11 year olds were significantly more accurate than the 7 and 8 year olds (p = .006 and .003, respectively). General hand rotation task performance (combining both back and palm view and collapsing medial and lateral stimuli) for the children aged 7–11 years can be seen in Fig. 3. Repeated measures ANOVA demonstrated that there was a significant effect of angle of rotation on response time, Wilks’ K = .34, F (4,63) = 30.60, p < .001, gP2 = .66, but there was no effect for age (p = .12) or interaction between the two (p = .76). For accuracy, a repeated measures ANOVA also demonstrated a significant effect for angle of rotation, Wilks’ K = .60, F (4,63) = 10.44, p < .001, gP2 = .40, and also for age, F (4,66) = 5.85, p < .001, gP2 = .26, but there was no interaction between the two (p = .94). The 11 year olds were significantly more accurate than the 7 and 8 year olds (both p = .002). Finally, a repeated measures ANOVA demonstrated that there was a significant effect for both angle, Wilks’ K = .44, F (4,63) = 20.14, p < .001, gP2 = .56, and age, F (4,66) = 4.23, p = .004, gP2 = .20, on efficiency, but there was no interaction between the two (p = .99). Across groups, efficiency scores were significantly lower (indicating greater efficiency) at 0° and 45° compared to other angles (all p < .05) and also at 90° compared to 180° (p < .001). Across angles, the 11 year olds were more efficient than the 8 year olds (p = .022) and there was a trend toward the same when the 11 year olds were compared to the 7 and 9 year olds (p = .050 and .051, respectively). A two-step hierarchical regression analysis was performed to determine whether age and motor skill level would predict efficiency on the hand rotation task. In the first step, age was demonstrated to be a significant predictor of performance, accounting for 29% of the variance in hand rotation efficiency, F (1,66) = 26.93, p < .001, R2 = .29. Next, MAND NDI was added as a predictor variable in a second step to determine whether, after accounting for age, motor skill level would influence hand rotation efficiency. This model accounted for 38% of the variability in hand rotation efficiency, F Table 3 Summary of hierarchical regression results with hand rotation efficiency as the outcome variable and Age (in months), and Neurodevelopmental Index (NDI) as the predictor variables (N = 71). Step
Predictor variable
1
DR2
ANOVA DF
Sign. DF
df
.29
26.93
.000
1,66
.089
9.29
.003
1,65
Age (in months) 2 Age (in months) NDI Note: S2 = Squared semi-partial correlations.
b
Regression SE b
t
b
p
S2
31.96
6.16
5.19
.54
