Motor imagery in children with DCD: A systematic and meta-analytic review of hand-rotation task performance

Highlights

• This is the first review summarizing HRT performance in children with DCD.

• Children with DCD engage in MI during HRT performance similarly to controls.

• However, they show consistent HRT performance deficits irrespective of outcome metric.

• Group difference effect sizes vary considerably depending on outcome measure.

• The efficiency metric offers the most sensitive measure for detecting group differences.

Abstract

This is the first review to quantitatively summarise evidence evaluating MI functioning in children with DCD compared to controls based on the hand rotation task (HRT). Specifically, MI performance was assessed using three different behavioural performance measures on the HRT (i.e., reaction time, accuracy and efficiency). Eight studies were included for quantitative analysis, yielding data for 176 and 198 children with and without DCD respectively. While children with DCD consistently used MI across all measures of the task, they continually demonstrated reductions in HRT performance relative to controls. Additionally, group differences appeared to be strongest and more commonly detected when using the IES (mean inverse efficiency-IES) metric on the HRT. These effects did not differ statistically as a function of instruction type. In support of the internal modelling deficit hypothesis, group effects suggested children with DCD demonstrate broad reductions in HRT performance relative to controls. However, consideration of effect size and study level analysis showed the ability for an individual study to detect these effects differs considerably depending on the outcome metric adopted.

Introduction

Developmental Coordination Disorder (DCD) is typified by poor motor coordination in the absence of any attributable medical or neurodevelopmental impairment (Zwicker et al., 2013; American Psychiatric Association, 2013). DCD affects approximately 5–6% of school-aged children (Zwicker et al., 2012). Individuals with DCD experience significant difficulty in executing the fundamental motor skills required for activities of daily living (e.g., getting dressed or eating meals with utensils) (Van der Linde et al., 2015; Bart et al., 2011; Zwicker et al., 2018), academic performance (e.g., handwriting) and social interaction (e.g., engaging in team activities at school) (Lingam et al., 2014; Gomez et al., 2017; Prunty et al., 2016; Leonard, 2016). Importantly, the impact of DCD extends beyond the motor domain and is associated with increased psycho-social difficulties (i.e., depression or anxiety) (Lingam et al., 2012; Chen et al., 2009; Dewey et al., 2002; Dewey and Volkovinskaia, 2018), and health complications such as obesity and cardiovascular disease (Rivilis et al., 2011; Cairney and Veldhuizen, 2013; Cermak et al., 2015; Cairney et al., 2017). These secondary issues may arise, in part, due to a commonly reported decrease in participation in physical activity (Zwicker et al., 2018; Cairney et al., 2017; Missiuna et al., 2007), particularly where group participation is required (Bouffard et al., 1996; Poulsen et al., 2007; Smyth and Anderson, 2000).

The last 20 years has seen considerable efforts to establish the underlying neuro-cognitive mechanisms responsible for DCD, in an effort to better understand atypical motor development and inform therapeutic interventions. This work has been the subject of several reviews (Zwicker et al., 2012; Wilson et al., 2013; Zwicker et al., 2009; Wilson and McKenzie, 1998; Wilson et al., 2017; Fuelscher et al., 2018). Given the substantial heterogeneity of DCD, it is perhaps unsurprising that no single aetiological account of DCD has been universally supported. Indeed, current work reports mixed descriptions of the neuro-cognitive processes that might be responsible for poor motor skill (Wilson et al., 2013). Nonetheless, several research groups have presented evidence suggesting that the motor impairments inherent to DCD may arise from a decreased capacity to mentally represent movement, often referred to as motor imagery (MI) (Adams et al., 2014; Gabbard and Bobbio, 2011; Williams et al., 2006; Hyde et al., 2014; Deconinck et al., 2009; Reynolds et al., 2015; Williams et al., 2008).

MI refers to the mental simulation of action in the absence of any evident motor output (Guillot et al., 2012; Munzert et al., 2009). This cognitive perceptual process is known to recruit similar, though not entirely analogous, neural structures to overt movement (e.g., parietal cortex, primary motor cortex, basal ganglia and premotor cortex (Hétu et al., 2013)). Further, internal representations of action appear to be bound by many of the same temporal and biomechanical constraints as overt movement (see below paragraph) (Butson et al., 2014; Parsons, 1994; Spruijt et al., 2015). Based on the functional and neural parallels between imagined and overt movement (Munzert et al., 2009), it is generally accepted that MI provides insight into an individual’s ability to generate forward models of action that subserve purposive movement (Guillot et al., 2012; Kilteni et al., 2018; Ridderinkhof and Brass, 2015). In short, prior to movement, the nervous system uses a copy of the impending motor command to predict the sensory consequences of movement should it unfold as expected (Guillot et al., 2012; Desmurget and Grafton, 2000; Franklin and Wolpert, 2011; Pickering and Clark, 2014; Gentsch et al., 2016). These estimates are then compared to sensory-inflow in real-time as the movement occurs. If a discrepancy is detected between the actual and predicted outcomes of an action (as per the forward model), an error signal is generated to revise the motor command in real-time. Accordingly, forward models provide stability for the nervous system beyond that offered by relatively slow cortico-sensory processing (Desmurget and Grafton, 2000). Further, if motor predictions prove to be inaccurate across successive trials, the error signal serves to adjust the forward model, which facilitates more accurate motor planning and control for future actions (Desmurget and Grafton, 2000; Y-w et al., 2007). The latter is particularly important in childhood, where constant neuro-muscular development renders predictive estimates of limb dimensions, strength, and speed obsolete on a virtually continuous basis (Wilson et al., 2013; Gredebäck et al., 2018). In short, forward modelling is a central pillar of typical motor development and is critical for efficient motor planning, control and learning.

While a variety of behavioural paradigms have been used to measure MI in samples of individuals with DCD, including reaching estimation (Caçola et al., 2014) and the visually guided pointing task (Adams et al., 2018; Lewis et al., 2008; Wilson et al., 2001), the hand rotation task (HRT) has been most commonly adopted (see below). Briefly, the former mental chronometry tasks usually involve participants performing a movement, and then subsequently being asked to imagine the movement they performed (Spruijt et al., 2015). These particular tasks are said to invoke an explicit form of MI, which activates different neural networks to the HRT (Hétu et al., 2013; Spruijt et al., 2015). Indeed, it has been argued that the HRT engages a distinct form of MI to most others (see Spruijt et al., 2015), and hence is the primary focus of the present review. The HRT is a well-validated measure of MI that involves participants identifying the laterality of hand stimuli (left vs. right hand) that appear on a screen at different angles. Self-reports and behavioural performance profiles typically indicate that participants engage in MI during the HRT. With respect to the latter, participant performance generally conforms to the same biomechanical constraints of actual movement, as per MI (Butson et al., 2014; Spruijt et al., 2015). That is, responses are often slower and/or less accurate when undertaking biomechanically more challenging rotations (i.e., lateral) compared to simpler ones (i.e., medial). Similarly, where stimuli are presented in back and palm view orientation, participants are faster to respond when the orientation of the hand stimuli is congruent to that of their own (de Lange et al., 2006; Hoyek et al., 2014; Ionta et al., 2007, 2012; Sirigu and Duhamel, 2001). These biomechanical effects are unique to MI performance strategies and are absent from non-motoric forms of imagery (e.g., visual imagery) (Sirigu and Duhamel, 2001).

In the case of DCD, most studies show that individuals with DCD display performance profiles on the HRT that are consistent with a MI strategy, suggesting that like healthy controls they tend to engage in MI while performing the HRT (Deconinck et al., 2009; Reynolds et al., 2015; Adams et al., 2016; Fuelscher et al., 2015a; Williams et al., 2011, 2013). However, this finding has not always been supported (Lust et al., 2006; Wilson et al., 2004), leaving some question as to whether children with DCD can, or do, engage in MI during the HRT.

Further, the vast majority of these studies show that children with DCD perform differently to neurotypical children on the HRT, with few exceptions (e.g., Lust et al., 2006), likely indicating a reduced capacity to engage MI (Adams et al., 2014; Gabbard and Bobbio, 2011; Deconinck et al., 2009; Williams et al., 2008). Assuming that MI provides insight into one’s ability to engage the forward model of action, it is often argued that children with DCD may have a deficit implementing this forward modelling system (Adams et al., 2014). However, while atypical HRT performance is commonly reported in children with DCD, the profile varies considerably across studies. For example, where some work indicates preserved reaction time (RT) yet decreased accuracy (e.g., Williams et al., 2006, Williams et al., 2011, Williams et al., 2013), others state the reverse (e.g., Wilson et al., 2004; Fuelscher et al., 2016), and others still have discovered broad deficits in both performance metrics (e.g., Deconinck et al., 2009; Reynolds et al., 2015; Williams et al., 2008; Adams et al., 2016; Fuelscher et al., 2015a; Noten et al., 2014). It is important to note that there are several sources of heterogeneity in the HRT protocol employed by studies when assessing MI in those with DCD; namely, the number of angles and viewpoints used to present the stimuli, the occlusion of the participant’s hands during task completion, and the response method utilised. These factors may have implications for HRT performance by altering the complexity of the task (please see the discussion for a detailed account). Given that deficits in MI appear to be more pronounced as task difficulty increases (Williams et al., 2008; Caçola et al., 2014; Noten et al., 2014; Adams et al., 2017), these methodological differences in HRT protocol are likely to produce variability in the nature of group differences reported in earlier studies. In short, the profile of HRT performance in children with DCD remains unclear, and by extension, so does the status of MI ability. Therefore, given that MI capacity provides a window into the ability to generate and employ internal models of action that are critical for the regulation of movement, elucidating the nature of MI in DCD may offer valuable insight into the neurocognitive underpinnings of this disorder.

We argue that one factor contributing to within and between group variability in HRT profiles may be the outcome measures that are reported, which are by no means consistent. Indeed, given these inconsistencies, our own research group have recently adopted an efficiency index (mean inverse efficiency-IES) to characterize HRT performance in individuals with DCD (Hyde et al., 2014; Fuelscher et al., 2015a, 2016; Hyde et al., 2018). Assuming that relevant statistical assumptions are satisfied (these assumptions are canvassed briefly in the ‘methods’ section here, and discussed in greater detail in Hyde et al., 2014), we have argued that incorporating response time and accuracy into a single metric (as per the mean IES metric) can provide a more sensitive measure of HRT performance than RT or accuracy considered in isolation. In support, using this IES measure we have observed reduced HRT performance in both children (Fuelscher et al., 2016, 2015b) and young adults (Hyde et al., 2014, 2017) with DCD, with remarkable consistency in the magnitude of effect and group descriptive statistics obtained across samples. Given the small sample sizes typical of experimental studies in DCD, the sensitivity of outcome measures is paramount to accurately characterizing differences between patient and control populations, should they exist.

Finally, there is some variability in the instruction type given to participants in earlier accounts of HRT performance in those with DCD. That is, where some studies give specific instructions prompting the use of MI (e.g., Noten et al., 2014; Reynolds et al., 2015; Williams et al., 2006, Williams et al., 2008, Williams et al., 2011, Williams et al., 2013), others do not (Deconinck et al., 2009; Fuelscher et al., 2016; Adams et al., 2017; Lust et al., 2006). Further, where studies have provided participants with MI instructions on a second attempt of completing the HRT (Williams et al., 2006; Reynolds et al., 2015; Williams et al., 2008), it is possible that any following improvements in performance may be a consequence of practice effects instead of the provision of instructions. Still, given that preliminary evidence suggests instruction may influence MI performance in those with DCD differently to controls (Williams et al., 2008), it is possible that instruction type may further contribute to variability in the profile of MI ability in samples of those with and without DCD. Clarifying this issue may also help to inform clinical interventions, particularly those employing MI.

Taken together, a strong body of work suggests that MI performance on the HRT is atypical in individuals with DCD. However, there still remains substantial variability in the profile of group differences reported. The aim of the present study was to provide the first meta-analytic review to quantitatively summarise the literature speaking to the performance of MI in children with DCD using the classic HRT. In light of the current interest in the viability of MI therapies as a means of treatment for children with DCD (Adams et al., 2016, 2017; Wilson et al., 2002), our study provides a timely account of HRT performance in children with DCD, and by extension, the status of MI in this group. More specifically, the present meta-analysis aimed to: (1) investigate whether participants with DCD at the meta-analytic level show evidence of engaging in MI during the HRT similar to controls; (2) characterize the profile of HRT performance, and by extension, MI functioning in children with DCD according to the three commonly used performance measures (i.e., RT, accuracy and mean IES); and (3) investigate whether the profile of MI performance in DCD is moderated by instruction type provided (i.e., with or without instructions). While these aims are somewhat intertwined, they are nonetheless distinct and critical to the valid interpretation and comparison of available and future accounts of HRT performance in DCD.