Developmental improvements in reaching correction efficiency are associated with an increased ability to represent action mentally

Journal of Experimental Child Psychology 140 (2015) 74–91

Contents lists available at ScienceDirect

Journal of Experimental Child Psychology journal homepage: www.elsevier.com/locate/jecp

Developmental improvements in reaching correction efficiency are associated with an increased ability to represent action mentally Ian Fuelscher a, Jacqueline Williams b, Christian Hyde a,⇑ a b

Cognitive Neuroscience Unit, School of Psychology, Deakin University, Burwood, VIC 3125, Australia Institute of Sport, Exercise, and Active Living, College of Sport and Exercise Science, Victoria University, Footscray, VIC 3011, Australia

a r t i c l e

i n f o

Article history: Received 10 December 2014 Revised 29 June 2015

Keywords: Online control Motor imagery Action representation Internal modeling Double-step reaching Hand rotation

a b s t r a c t We investigated the purported association between developmental changes in the efficiency of online reaching corrections and improved action representation. Younger children (6–7 years), older children (8–12 years), adolescents (13–17 years), and young adults (18–24 years) completed a double-step reaching paradigm and a motor imagery task. Results showed similar nonlinear performance improvements across both tasks, typified by substantial changes in efficiency after 6 or 7 years followed by incremental improvements. Regression showed that imagery ability significantly predicted reaching efficiency and that this association stayed constant across age. Findings provide the first empirical evidence that more efficient online control through development is predicted, partly, by improved action representation. Ó 2015 Elsevier Inc. All rights reserved.

Introduction Development of online control The ability to correct one’s movement mid-flight in response to unexpected environmental changes has received much attention in the literature of late (e.g., King, Oliveira, Contreras-Vidal, & Clark, ⇑ Corresponding author. E-mail addresses: [email protected] (I. Fuelscher), [email protected] (C. Hyde). http://dx.doi.org/10.1016/j.jecp.2015.06.013 0022-0965/Ó 2015 Elsevier Inc. All rights reserved.

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

75

2012; Ruddock et al., 2014; Wilson & Hyde, 2013). This so-called online control of movement is seen as a marker of the nervous system’s capacity to interact effectively with an unpredictable and fluid environment, a critical feature of a mature motor system. Despite this, little is known about its development through childhood and beyond. The limited evidence that is available, however, suggests a nonlinear maturation. Specifically, during early childhood (prior to 8 years) mid-movement corrections to reaching following unexpected target perturbation are slow and highly variable. However, by mid-childhood (around 8 years), the efficiency with which a child can engage in these types of online corrective actions improves substantially. This period is followed by more subtle performance improvements into adolescence and beyond (King et al., 2012; Ruddock et al., 2014; Wilson & Hyde, 2013). For example, Wilson and Hyde (2013) recently compared the performance of younger (6–7 years), mid-aged (8–9 years), and older (10–12 years) children, as well as young healthy adults, on the well-validated double-step reaching task (DSRT). Participants were required to reach for one of three targets presented on a touch screen. For most trials the target remained stationary for the duration of movement, whereas for the remaining trials the target jumped at movement onset (jump trials). During this task, non-jump trials are thought to place few demands on online corrective systems because the target remains stationary throughout movement. That is, presuming that the initially programmed motor command is accurate, it can unfold unchanged. Conversely, the unexpected target perturbation that occurs during jump trials renders the initial motor command inaccurate. As such, successful trial completion is dependent on how efficiently an individual can update the motor command in real time and, hence, facilitate the timely redirection of the limb toward the newly cued target. Consequently, immature or impaired online control manifests as poor performance on jump trials relative to non-jump, a profile shown by patient groups where deficits in the online control of movement are core symptoms, including parietal lesion patients (Blangero et al., 2008; Gréa et al., 2002; Ochipa et al., 1997). Interestingly, Wilson and Hyde (2013) showed that whereas jump trial reaching speed (relative to non-jump) was relatively slow during early childhood (6–7 years), it decreased substantially by middle childhood (8–9 years) and remained relatively stable into older childhood. This age-related improvement in accounting for target perturbation was confirmed by kinematic analyses that showed a significant reduction in time to reach trajectory correction values between younger and middle childhood, where they then stabilized into later childhood (i.e., 10–12 years). Importantly, this developmental profile of online control is a consistent feature of the small number of developmental studies into the online control of reaching (King et al., 2012; Ruddock et al., 2014). From a computational perspective, the ability to engage in online reaching corrections is thought to depend heavily on an individual’s ability to represent action at an internal neural level. Specifically, the nervous system is thought to use an efferent copy of the impending motor command to anticipate the limb trajectory should the movement unfold as anticipated. On movement initiation, actual visual and proprioceptive in-flow becomes available and is compared with the predicted sensory information (as per the action representation) in real time. In case of a mismatch (e.g., following unexpected target perturbation), an error signal is generated which must then be integrated seamlessly with the unfolding motor command, affording fluent and efficient correction to the moving limb (Desmurget & Grafton, 2000). By anticipating the sensory consequences of movement, this predictive modeling system allows the nervous system to circumvent sensory processing delays (which can exceed 250 ms; Frith, Blakemore, & Wolpert, 2000) with minimum lag (Desmurget & Grafton, 2000; Shadmehr, Smith, & Krakauer, 2010). In neural terms, this system appears to be supported by a functional loop between connections across motor and frontal cortices and parietal and cerebellar networks (Andersen & Cui, 2009; Izawa & Shadmehr, 2011; Mulliken, Musallam, & Andersen, 2008). This neurocomputational modeling is supported by a strong body of indirect (Hyde & Wilson, 2011a, 2011b; King et al., 2012; Ruddock et al., 2014; Wilson & Hyde, 2013) and direct (Hyde, Wilmut, Fuelscher, & Williams, 2013) empirical evidence demonstrating that the efficiency with which individuals are able to implement online control is dependent on their capacity to generate and integrate internal ‘‘neural’’ action representations with incoming sensory information. Accordingly, it is generally argued that the nonlinear improvement in online control observed between the critical years spanning ages 6 to 12 is subserved by an improved ability to generate and/or use internal ‘‘action’’ representations

76

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

(King et al., 2012; Wilson & Hyde, 2013). Interestingly, this developmental progression coincides with the protracted maturation of the parietal cortices and their projections to frontal structures (see Casey, Tottenham, Liston, & Durston, 2005, for a review) and central nervous system more broadly (Kail, 1991). These neural systems are known to be critical to visually guided reaching in adults (Ferraina, Battaglia-Mayer, Genoveslo, Archambault, & Caminiti, 2009; Pisella, Binkofski, Lasek, Toni, & Rossetti, 2006), with the posterior parietal cortex in particular thought to be critical to integrating sensory (especially visual) inputs with predictive estimates of limb locations (Macuga & Frey, 2014) and possibly involved in generating and processing the error signals that arise following mismatches between the two (Reichenbach, Bresciani, Peer, Bulthoff, & Thielscher, 2011). Developmental improvements in online control coincide with a greater capacity for generating internal neural representations of action Empirical support for the argument that improvements in the proficiency of online control through development are supported, at least partly, by a greater capacity to generate internal representations of action can be found from evidence that the development of motor imagery (MI), an experimental protocol thought to elucidate the integrity of the action representation that ordinarily precedes movement (Decety, Jeannerod, & Prablanc, 1989; Parsons, 1994; Sirigu et al., 1996), follows a similar maturational timeline to that of online control. MI requires participants to mentally represent an action without overt movement taking place (Decety & Grèzes, 1999). Performance of typically developing children and adults has been shown to be subject to the same temporal and biomechanical constraints as actual movements. Indeed, the time taken to imagine a movement correlates closely with subsequent execution times (Decety et al., 1989; Sirigu et al., 1996), with awkward and more physically demanding actions taking longer to imagine (Butson, Hyde, Steenbergen, & Williams, 2014; de Lange, Roelofs, & Toni, 2008; Munzert, Lorey, & Zentgraf, 2009). This relative functional equivalence is coupled with corresponding neurophysiological similarities, with neuroimaging studies indicating that imagined movements activate similar neural networks to those activated in actual movement (Jeannerod, 2001; Munzert et al., 2009) and corticospinal pathways (e.g., Williams, Pearce, Loporto, Morris, & Holmes, 2012). Consequently, it is largely assumed that MI provides insight into one’s ability to accurately form and monitor the kinds of internal motor representations that support purposive action (de Lange et al., 2008; Jeannerod, 2001; Munzert et al., 2009). One of the more commonly adopted tasks of MI, the mental limb rotation task, requires participants to judge the laterality of a limb (usually a hand) presented at different rotation angles. Although participants often report imagining rotating their own hand to respond (de Lange, Helmich, & Toni, 2006; Kosslyn, Digirolamo, Thompson, & Alpert, 1998; Parsons & Fox, 1998), the use of an embodied MI strategy to complete the task is corroborated by neuroimaging evidence indicating that participants enlist fronto–parieto (dorsal) circuitry specific to MI (cf. visual imagery that typically enlists more ventral structures; see Kosslyn, Ganis, & Thompson, 2001, for a review). Behaviorally, performance of typically developing children (and adults) has been shown to conform to the kinesthetic and biomechanical constraints of action during MI; that is, children display faster response times when performing physically comfortable rotations compared with impossible or awkward rotations. Indeed, children as young as 5 years who are able to make ‘‘left/right’’ distinctions have been shown to display this profile (Butson et al., 2014). Consequently, it is widely argued that participants engage in an embodied (i.e., MI) strategy to perform the task (Gabbard, 2009; Munzert et al., 2009; ter Horst, van Lier, & Steenbergen, 2010), although this position is not without critique (see Grafton & Viswanathan, 2014). Although numerous developmental studies have explored MI performance in children aged 6 to 12 years using the hand rotation task (see Gabbard, 2009, and Butson et al., 2014, for reviews), few studies (e.g., Butson et al., 2014) have compared performance of younger children (i.e., 6–7 years) with that of older children (i.e., 8–12 years). Indeed, much of the research is drawn from comparisons of typically and atypically age-matched developing children (e.g., Deconinck, Spitaels, Fias, & Lenoir, 2009; Williams, Reid, Reddihough, & Anderson, 2011; Williams et al., 2011), thereby limiting the degree to which inferences can be drawn about the development of MI during the critical 7- and 8-year junction. Still, where available, developmental evidence from the hand rotation task points

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

77

toward a critical developmental period for MI ability between 6 and 12 years. For example, Caeyenberghs, Tsoupas, Wilson, and Smits-Engelsman (2009) reported nonlinear improvements in MI proficiency in typically developing children aged 7 to 12 years. Specifically, 7- and 8-year-olds were slower and less accurate than 9- to 12-year-olds on the hand rotation task, with no significant differences observed between 9- and 10-year-olds and 11- and 12-year-olds. This evidence was taken to suggest that imagery efficiency improves rapidly during early childhood, with only subtle improvements henceforth. It should be noted, however, that the authors did not analyze whether mental rotation performance was influenced by the biomechanical or postural constraints of real action (as would be expected if participants were using an MI strategy). Therefore, it is difficult to infer whether participants were indeed engaging in a specific MI strategy or an alternative nonmotoric imagery strategy to complete the task. In a more recent study, Butson et al. (2014) highlighted a similar developmental trajectory of MI capacity showing that 11-year-old children were significantly more accurate, but not significantly faster, than 7- and 8-year-olds on the hand rotation task. When accuracy was accounted for in response time, however, age differences appeared, with the 11-year-olds being significantly more efficient than the 8-year-olds. Notably, the authors tested a group of 5- and 6-year-olds but were unable to include this group in their main analysis because only five of the children met the minimum accuracy requirement. The descriptive data for these five children, however, indicated that performance was constrained by the biomechanical constraints of actual movement. This suggests that children under 7 years of age can perform MI, although with limited accuracy. At age 7, however, there was a considerable jump in accuracy levels, suggesting substantially improved MI performance. Taken together, this body of research points toward a critical period for the development of MI between 6 and 12 years of age. Given that MI is largely assumed to provide insight into one’s ability to accurately form and monitor mental representations of action, there is compelling evidence that children’s ability to represent action mentally develops in a strikingly similar nonlinear manner to online control of reaching. This hand rotation performance profile across development is unlikely to reflect age-dependent learning effects considering that no feedback on response accuracy was given and the different trial types (i.e., angular rotation, hand orientation, and direction of rotation) were randomized. Interestingly, the neural networks that support MI overlap considerably with those underpinning the online control of reaching, including the fronto–posterior–cerebellar structures such as the posterior parietal cortex (PPC), premotor cortex, and cerebellum. The cerebellum and PPC in particular are thought to play a critical role for MI (see Macuga & Frey, 2014). As noted, these neural structures, in particular those more posterior structures (e.g., the parietal cortices), undergo rapid maturation between 6 and 10 years of age, the product of a complex reciprocal interaction between endogenous (i.e., neurophysiological and genetics) and exogenous (i.e., environmental) factors (Butson et al., 2014; Caeyenberghs et al., 2009; Casey et al., 2005; Munakata, Casey, & Diamond, 2004). To summarize, there is a growing body of theoretical and empirical evidence that online control of reaching shows a nonlinear developmental progression characterized by rapid improvement during the critical 6- to 12-year period. Based on neurocomputational modeling, it is widely assumed that this development is subserved, at least in part, by an improved capacity to generate and/or use action representations. This suggestion is supported by evidence from MI studies showing a similar developmental profile. Despite a strong body of theoretical and indirect empirical evidence suggesting the importance of accurate action representation to individuals’ ability to efficiently correct their reaching online, no developmental study has measured both online control of reaching and MI ability in the same child at any age group. Accordingly, it is difficult to verify the degree to which the well-established trend of improved online control from 6 to 12 years and beyond is in fact associated with a greater capacity to generate and/or engage action representation. Clarifying this issue is critical to our understanding of the development of online control and the neurocognitive mechanisms that support it and to developing appropriate interventions when it develops atypically. To this end, the aim of this study was to test the purported association between the development of online control reaching efficiency throughout the critical 6- to 12-year period and the capacity to generate internal ‘‘neural’’ action representations; a group of adolescents and adults were included to ensure that child development was considered in the broader context of neuromotor maturation.

78

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

Online control of reaching was measured using the well-validated DSRT; the ability to generate internal movement representations was measured using a traditional MI task, the hand rotation task. As per earlier investigations, it was predicted that the ability to correct reaching in response to unexpected target perturbation would improve substantially from early childhood (6–7 years) across late childhood (8–12 years) and begin to plateau into adolescence (13–17 years) and early adulthood (18–24 years). Based on neurocomputational modeling suggesting that accurate and efficient internal action representation (i.e., MI) is fundamental to these types of corrective movements, it was expected that MI performance would show a similar developmental profile, indicated by faster response times and higher accuracy levels. For the same reason, it was expected that MI ability would positively predict the efficiency of online control throughout development. Method Participants The sample consisted of 115 participants, from which 13 children aged 6 or 7 years, 1 12-year-old, and 1 13-year-old were removed from the analysis because they failed to reach our minimum accuracy criterion on the hand rotation task (see ‘‘Design and analysis’’ section below). The final sample comprised 100 participants consisting of 15 younger children aged 6 or 7 years (5 boys and 10 girls, Mage = 7.04 years, SD = 0.61), 30 mid-aged children of 8 to 12 years (16 boys and 14 girls, Mage = 10.40 years, SD = 1.33), 36 adolescents aged 13 to 17 years (22 boys and 14 girls, Mage = 14.48 years, SD = 1.33), and 19 young adults aged 18 to 24 years (10 men and 9 women, Mage = 22.15 years, SD = 1.68). Although these age groups are common for studies investigating online control of reaching in children (Hyde & Wilson, 2011a, 2011b; Plumb et al., 2008), some studies report a further breakdown of the 8- to 12-year age group (e.g., Wilson & Hyde, 2013). In the current study, we chose to group 8- to 12-year-olds together because the majority of available research has failed to find appreciable difference within this range on our key metrics (Ruddock et al., 2014; Wilson & Hyde, 2013) and we were mindful of the effect of redundant multiple group comparisons on the family-wise error rate. For this latter reason, we also chose to group 13- to 17-year-olds together because the majority of available research has demonstrated only subtle improvements on our key metrics from late childhood into adulthood (e.g., Ruddock et al., 2014; Wilson & Hyde, 2013) or failed to find meaningful differences at all (e.g., King et al., 2012). In support of grouping 13- to 17-year-olds together here, preliminary analysis of participants within this age band failed to show a significant correlation between any of our key metrics and age, suggesting that chronological age was not linked to performance on our key metrics. Children and adolescents were recruited from two primary schools and two secondary schools in metropolitan Melbourne, Australia. Adults were undergraduate students attending a university in Melbourne. All participants had age-appropriate levels of motor skill, scoring above the 15th percentile on the McCarron Assessment of Neuromuscular Development (McCarron, 1997). Because children were recruited from mainstream primary schools, and adults were recruited from a mainstream university, it was assumed that participants included in the study were within the normal IQ range (Geuze, Jongmans, Schoemaker, & Smits-Engelsman, 2001; Hyde & Wilson, 2013). Measures and procedure Online control task Online control was assessed using the double-step reaching task. The DSRT was programmed using the Virtools software package (Dassault Systems, Velizy-Villacoublay, France) and was presented on a 40-inch touch-screen monitor (Samsung, Korea) mounted at an angle of 15° from horizontal on a table. A green circle 5 cm from the bottom center of the screen acted as a ‘‘home base.’’ Three yellow targets (each 2 cm in diameter) were presented in a semicircular formation across the top of the screen. Targets were located at 20°, 0° (sagittal from participants’ umbilicus), and 20° with respect to the home base. The center of the home base was located 30 cm away from each of the targets.

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

79

Participants completed the trial using their dominant index finger. Before the start of each trial, the home base was illuminated and participants were instructed to touch and hold the home base. Following a random interval between 500 and 1500 ms, the home base was extinguished and the central target was illuminated. Participants were informed that the central target would always illuminate first to ensure that visual attention was initially directed to the same spatial location. On target illumination, participants were required to reach and touch the middle of the target as quickly and accurately as possible. A short auditory tone indicated trial success (i.e., touching within the boundary of the target). For unsuccessful trials, participants were instructed to continue pressing the target until the auditory tone was emitted. At the completion of each trial, participants returned to the home base and waited for the next trial. For 80% of trials, the central target remained illuminated for the duration of the movement (non-jump trials). For the remaining 20% of trials, the central target was initially lit, but at movement onset (i.e., as soon as the finger left the home base) the central target was extinguished and reappeared immediately at one of the two peripheral locations (jump trials, see Fig. 5). Prior to testing, the assessor demonstrated a non-jump trial and a jump trial. Participants then completed 10 practice trials (8 non-jump trials and 2 jump trials). Following the practice trials, participants completed one block of 30 trials and one block of 40 trials, resulting in a total of 70 trials. Block 1 consisted of 24 non-jump trials and 6 jump-trials (3 to each side), and Block 2 consisted of 32 non-jump trials and 8 jump trials (4 to each side). Trials were presented in a pseudorandomized order. The reaching hand was tracked with a sensor placed on the back of participants’ index finger using an adhesive pad. Tracking was done in real time using an ultrasonic tracking system (Zebris CSM10, Aachen, Germany) sampled at 200 Hz. Chronometric measures taken were reaction time, measured by the time between target display onset and finger lift-off from the home base, and movement time (MT), measured by the interval between lift-off and placement of the finger on the target location. Two types of errors were recorded to verify trial viability; touchdown errors were recorded for trials where participants touched (i.e., center of pressure) outside the target boundary, and anticipatory errors were recorded for responses that were initiated (i.e., lift-off from the home base) before a target location was illuminated or for reaction times within 150 ms (see Hyde & Wilson, 2011a). Error trials were excluded from the analysis, resulting in an average of 46 (66%) valid trials for 6- and 7-year-olds, 50 (71%) for 8- to 12-year-olds, 61 (87%) for 13- to 17-year-olds, and 61 (87%) for adults. The kinematic measure recorded was time to correction (TC). TC was measured as the time that movement trajectory corrected away from the initial (central) target toward the correct target on jump trials. Specifically, the two-dimensional jump trial reaching trajectory was inspected for each jump trial to determine at which point in the reaching trajectory movement identifiably diverted from the (virtually) straight trajectory for non-jump trials. Movement trajectory was plotted on a two-dimensional Cartesian plane using MATLAB (Mathworks, Natick, MA, USA); TC was determined visually by two independent researchers to ensure reliability. This method of visual inspection has been reliably adopted by a number of studies that adopted the DSRT (Hyde & Wilson, 2011a, 2013; Pisella et al., 2000; Ruddock et al., 2014; Van Braeckel, Butcher, Geuze, Stremmelaar, & Bouma, 2007; Wilson & Hyde, 2013).

Hand rotation task The ability to engage in motor imagery was assessed using the hand rotation task. Single hand stimuli (9  8 cm, centered in the middle of the screen) were presented on a laptop computer using E-Prime software (Psychology Software Tools, Pittsburgh, PA, USA). Participants were instructed to decide whether each presented stimulus was a left hand or a right hand as quickly and accurately as possible (see Fig. 6). Hands were presented in either palm view (palm of the hand facing toward participants) or back view (back of the hand facing toward participants). Hands were presented randomly in 45° increments between 0° and 360° and remained on the screen until a response was recorded by pressing a designated key on the computer keyboard or 10 s had passed. Response time (RT) and accuracy for each stimulus were recorded; RT was recorded to the nearest 1 ms. Participants completed 5 practice trials, followed by 80 test trials, providing 8 trials at each angle. Thus, prior to group comparisons taking place, preliminary analysis of each group verified the use of an MI strategy.

80

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

Design and analysis Online control task Kinematic data were filtered offline using a fourth-order Butterworth filter with a cutoff of 10 Hz. For jump trials, timed responses were collapsed over target location (i.e., left and right). For each participant, mean values were recorded for each dependent measure; trials on which errors occurred were counted but not included in the analysis of chronometric or kinematic data. For each dependent measure, outliers were defined as responses ± 2.5 standard deviations from the mean. Separate mixed-design analyses of variance (ANOVAs) with trial type (non-jump vs. jump) as the repeated-measures factor and age group as the between-participants factor were conducted on mean values for reaction time and movement time. A one-way ANOVA with age group as the between-groups factor was conducted on mean TC values. Hand rotation task Mean RT and accuracy for each hand at each angle of rotation was calculated for each participant. RT was calculated using both correct and incorrect trials; in support of this approach, preliminary analysis indicated near perfect correlations between RTs using both correct and incorrect trials and RTs using correct trials only in each age group (all rs > .98, ps < .001), suggesting that participants were using a comparable mental imagery strategy irrespective of whether they responded correctly or not. Mean accuracy was calculated as the proportion of correct responses over all trials. A criterion of 60% accuracy for hands presented at 0° was set as a minimum requirement to include the data in the analysis. This criterion was used to ensure that participants were capable of differentiating hand laterality at the most basic level of stimulus presentation (see Butson et al., 2014) and resulted in the exclusion of 13 children aged 6 or 7 years, 1 8- to 12-year-old, and 1 13- to 17-year-old. Trials with RTs less than 250 ms were deleted from the analysis, resulting in an average of 77 (96%) valid trials for 6- and 7-year-olds and 79 (99%) for 8- to 12-year-olds, 13- to 17-year-olds, and adults. To confirm that performance of participants was constrained by the biomechanical constraints of action, we ran separate two-way mixed-design ANOVAs with RT and then accuracy as the dependent variables, age group (6–7, 8–12, 13–17, or 18–24 years) as the between-participants factor, and either direction of rotation (medial vs. lateral) or stimulus view (back vs. palm) as the within-participants factor. Medial rotation performance was determined as the average of responses for left hands presented at 45°, 90°, and 135° and for right hands presented at 315°, 270°, and 225°. Lateral rotation performance was determined as the average responses for left hands presented at 315°, 270°, and 225° and for right hands presented at 45°, 90°, and 135°. For each stimulus view condition (back vs. palm), performance was averaged across angular rotations. In line with previous research (see Butson et al., 2014), 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). Notably, although the effects for direction of rotation and stimulus view were of interest to draw inferences as to whether participants were using an MI strategy, the age group effect was considered redundant because it was presented in a subsequent analysis and, hence, is not reported here. That is, we conducted age group comparisons across angular rotation because it allowed us to simultaneously confirm the appropriateness of our predictor variable for the final regression model (see below). To investigate the developmental trajectory of MI performance (i.e., test for age-related group differences on MI performance), both RT and accuracy were submitted to a 5 (Angle)  4 (Age Group) mixed-design ANOVA. This analysis also allowed us to confirm whether mean RT and mean accuracy (i.e., averaged across angular rotation) provided an appropriate indirect measure of imagery performance on the hand rotation task. To this end, analysis needed to demonstrate a nonsignificant interaction effect between angle and age, that is, a constant association between these two factors. Developmental association between MI and online control measures To determine whether MI ability was a significant predictor of online control efficiency, an initial two-step hierarchical regression was conducted with TC as the dependent variable and RT and accuracy as the predictors while controlling for non-jump reaching efficiency. Although online control is

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

81

thought to be heavily dependent on the integrity of mental action representations, it is also influenced by an individual’s general manual reaching efficiency. Accordingly, we controlled for movement time on non-jump trials to partial out any effect of general reaching speed on reaching correction latency on jump trials (see Hyde et al., 2013). At Step 3, age was entered into the model to investigate whether the developmental progression of online control was influenced by general age-related neuromotor improvements over and above those attributed to action representation (and general reaching speed). Age was entered as a continuous predictor to avoid loss of power and residual confounding (Royston, Altman, & Sauerbrei, 2006). The final model then included age as a potential moderator to investigate whether the association between action representation (i.e., imagery performance) and online reaching correction efficiency (i.e., TC) varied according to age. Specifically, we created moderating terms for RT and accuracy separately and entered them as predictors into the model (see Aiken & West, 1991). Predictors were mean-centered prior to analysis. Data were cleaned and checked for violations of assumptions. Relevant assumptions were met unless otherwise stated. Preliminary analysis failed to reveal significant sex differences for performance on the online control and MI metrics. Furthermore, sex did not moderate the association between MI and online control performance. Accordingly, sex was not controlled for or included in the analyses. Results Developmental comparison of double-step reaching performance The two-way ANOVA on reaction time revealed a significant main effect for group, F(3, 96) = 14.64, p < .001, g2p = .31. No interaction effect and no main effect for trial were observed. Averaged across trials, younger children (663 ms) were significantly slower than mid-aged children (481 ms), p < .001, g2p = .33. No significant differences were observed between adolescents (465 ms) and adults (421 ms). The two-way ANOVA on movement time revealed a significant interaction effect, Wilks’ K = .68, F(3, 96) = 15.12, p < .001, g2p = .32. The interaction effect suggested that age group differences were more pronounced on jump trials, p < .001, g2p = .71, than on non-jump trials, p < .001, g2p = .43. Mean values are shown in Fig. 1. Analysis of movement time difference scores between jump and non-jump trials showed that 6- and 7-year-olds (345 ms) scored significantly higher than 8- to 12-year-olds (252 ms), p = .003, g2p = .18, who in turn scored significantly higher than 13- to

1200

Movement me (ms)

1000

800 6-7 years

600

8-12 years 13-17 years 18-24 years

400

200

0

Non-Jump

Jump

Fig. 1. Mean movement time values on the DSRT. Error bars represent standard errors.

82

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

17-year-olds (190 ms), p = .004, g2p = .13. No significant difference was observed between adolescents and adults (208 ms). The one-way ANOVA on TC revealed a significant difference among the age groups, F(3, 96) = 63.69, p < .001, g2p = .66. Younger children were significantly slower to change reaching trajectory than mid-aged children, p = .002, g2p = .21, who in turn were slower than adolescents, p < .001, g2p = .45. Adolescents were marginally slower than adults, p = .018, g2p = .10. Mean values are presented in Fig. 2. Development of MI: Hand rotation performance Testing for biomechanical effects on performance The 2 (Direction of Rotation)  4 (Age Group) mixed ANOVA on RT showed a significant main effect for direction of rotation, Wilks’ K = .60, F(1, 96) = 62.85, p < .001, g2p = .40, and a significant main effect for age group, F(3, 96) = 38.84, p < .001, g2p = .55. No significant interaction effect was observed. Averaged across the age groups, participants responded faster to medially rotated stimuli than to laterally rotated stimuli. Mean values for all biomechanical constraints analyses are presented in Table 1. The 2 (Direction of Rotation)  4 (Age Group) mixed ANOVA on accuracy showed a significant interaction effect, Wilks’ K = .78, F(3, 96) = 8.94, p < .001, g2p = .22. Averaged across the age groups, participants responded more accurately to medially rotated stimuli than to laterally rotated stimuli, Wilks’ K = .66, F(1, 96) = 49.17, p < .001, g2p = .34. This effect was present for all age groups. The interaction effect suggested that age group differences were more pronounced for lateral rotations, p < .001, g2p = .43, than for medial rotations, p < .001, g2p = .26. The 2 (Stimulus View)  4 (Age Group) mixed ANOVA on RT showed a significant main effect for direction of rotation, Wilks’ K = .48, F(1, 96) = 87.48, p < .001, g2p = .48, and a significant main effect for age group, F(3, 96) = 40.52, p < .001, g2p = .56. No significant interaction effect was observed. Averaged across the age groups, participants responded faster to stimuli presented in back view than to stimuli presented in palm view. The 2 (Stimulus View)  4 (Age Group) mixed ANOVA on accuracy showed a significant interaction effect, Wilks’ K = .83, F(3, 96) = 6.39, p = .001, g2p = .17. Averaged across the age groups, participants responded more accurately to stimuli presented in back view than to stimuli presented in palm view, Wilks’ K = .76, F(1, 96) = 30.58, p < .001, g2p = .24. The interaction suggested that this effect was most pronounced in the 6- and 7-year-olds.

450 400

Time of correcon (ms)

350 300 250 200 150 100 50 0

6-7 years

8-12 years

13-17 years

18-24 years

Fig. 2. Mean time of correction values on the DSRT. Error bars represent standard errors for each age group.

83

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91 Table 1 Descriptive statistics for biomechanical constraints analyses in each age group.

Medial rotation Lateral rotation Back view Palm view

6–7 years (n = 15)

8–12 years (n = 30)

13–17 years (n = 36)

RT

ACC

RT

ACC

RT

ACC

18–24 years (n = 19) RT

ACC

2915.48 (898.64) 3268.35 (929.96) 2971.85 (809.43) 3242.47 (863.02)

.80 (.12) .56 (.26) .78 (.16) .58 (.20)

2117.46 (535.62) 2475.89 (560.65) 2118.07 (461.90) 2470.41 (516.60)

.91 (.09) .83 (.12) .88 (.09) .84 (.12)

1550.79 (358.24) 1958.29 (562.55) 1638.57 (482.04) 1995.49 (456.01)

.95 (.06) .91 (.08) .93 (.07) .90 (.07)

1187.69 (277.82) 1393.46 (381.93) 1213.46 (342.12) 1406.28 (329.72)

.94 (.07) .88 (.12) .92 (.09) .89 (.12)

Note. RT, mean response time (ms); ACC, mean accuracy (proportion correct). Standard deviations are in parentheses.

Age effects The 5 (Angle)  4 (Age Group) mixed ANOVA on RT revealed a significant interaction effect, Wilks’ K = .78, F(12, 246.35) = 2.03, p = .023, g2p = .08, and a significant main effect for group, F(3, 96) = 39.02, p < .001, g2p = .55. Averaged across angle, younger children responded significantly slower than mid-aged children, p < .001, g2p = .30, who in turn responded slower than adolescents, p < .001, g2p = .22, who in turn responded slower than adults, p < .001, g2p = .22. This group effect was present at each angle. The interaction suggested that greater angular rotation resulted in longer RTs for all age groups, with this effect being less pronounced in 6- and 7-year-olds. Mean values are presented in Fig. 3. The 5 (Angle)  4 (Age Group) mixed ANOVA on accuracy revealed a significant main effect for group, F(3, 96) = 24.57, p < .001, g2p = .43, and a significant main effect for angle, Wilks’ K = .73, F(4, 93) = 9.12, p < .001, g2p = .28. No significant interaction effect was observed. Averaged across angle, younger children were significantly less accurate than mid-aged children, p < .001, g2p = .37, who in turn were less accurate than adolescents, p = .002, g2p = .14. No significant difference in accuracy was observed between adolescents and adults. Averaged across group, accuracy decreased with increasing angular rotation. Mean values are presented in Fig. 4. Importantly, the significant group main effects across both analyses confirmed mean RT and mean accuracy as appropriate indirect indicators of imagery performance on the hand rotation task. These measures were consequently used in the regression model. 4000

Mean response me (ms)

3500 3000 2500 6-7 years

2000

8-12 years 13-17 years

1500

18-24 years

1000 500 0

0

45

90

135

180

Angular rotaon (°) Fig. 3. Mean response time values on the hand rotation task. Error bars represent standard errors at each angle.

84

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

Mean accuracy (proporon correct)

1.0 0.9 0.8 0.7 0.6

6-7 years

0.5

8-12 years

0.4

13-17 years

0.3

18-24 years

0.2 0.1 0.0

0

45

90

135

180

Angular rotaon (°) Fig. 4. Mean accuracy values on the hand rotation task. Error bars represent standard errors at each angle.

Association between MI and online control of reaching Results of the four-step hierarchical regression are displayed in Table 2. The initial two-step model (Model 2) demonstrated that, after controlling for general reaching efficiency, both RT and accuracy were significant predictors of TC, accounting for just under 20% (DR2 = .18, p < .001) of the variability in TC over and above non-jump reaching efficiency. Both RT (sr = .33, p < .001) and accuracy (sr = .19, p = .005) made significant unique contributions to TC. The inclusion of age in the model (Model 3) significantly improved prediction of TC, explaining an additional 6% (DR2 = .06, p < .001) of the variability in the dependent variable. Even after the inclusion of age in the model, both RT (sr = .17, p = .007) and accuracy (sr = .13, p = .031) continued to make significant unique contributions to TC. The purpose of the final model (Model 4), which included age as a potential moderator, was to investigate whether the association between imagery performance and TC varied according to age. Results demonstrated that age did not significantly moderate the associations between imagery performance and TC (DR2 = .01, p = .157). Neither the interaction term for RT nor the interaction term for accuracy made a significant unique contribution to the prediction of TC. Following the inclusion of the interaction terms, neither RT nor accuracy continued to make significant unique contributions predicting variance in TC.

Table 2 Summary of hierarchical moderating regression analysis for variables predicting TC (N = 100). Model 1

Model 2

Model 3

Model 4

Variable

B

SE

b

B

SE

b

B

SE

b

B

SE

b

MTNJ RT ACC Age

0.39

0.05

.64**

0.21 0.04 114.28

0.05 0.01 39.70

.35** .41** .20**

0.14 0.02 83.18 4.96

0.05 0.01 37.99 1.28

.22** .24** .15* .37**

0.13 0.13 63.00 6.38

0.05 0.01 41.68 1.48

.21* .14 .11 .48**

0.00 1.29

0.00 6.91

.14 .01

RT  Age ACC  Age R2 Fchange

.41 69.10**

.60 21.81**

.65 14.95**

.67 1.89

Note. MTNJ, movement time, non-jump; RT, reaction time; ACC, accuracy. RT, ACC, and age were centered at their means. * p < .05. ** p < .01.

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

85

Given that Model 4 did not significantly improve prediction of TC, Model 3 provided the model with the best fit, explaining just under two thirds of the variance in TC. A comparison between Models 1 and 3 (see Lange-Küttner, 1997, 2004, for a description of the statistical analysis) suggested that age and imagery performance alone explained approximately 24% of the variance and that 38% of the variance accounted for by these factors was shared variance with non-jump reaching. Discussion Recent research suggests a critical period of development for the online control of reaching between 6 and 12 years of age characterized by nonlinear improvements in efficiency. Neurocomputational modeling suggests that these improvements may arise, at least in part, as a result of an increased ability to generate and/or implement internal action representations. To date, however, no study has measured both action representation and online control in the same sample of children across development, limiting the degree to which we are able to draw inferences about this purported association. Accordingly, we measured the efficiency with which children between 6 and 12 years of age could engage in both online control of reaching and action representation; separate groups of adolescents (13–17 years) and young adults (18–24 years) were also included to ensure that childhood performance could be interpreted in the broader context of typical development. As predicted, the ability of children to correct reaching online improved substantially from 6 to 12 years of age, characterized by a significant improvement in reaching efficiency from early (6–7 years) to late (8–12 years) childhood. Although performance continued to improve into adolescence, only subtle improvements were observed between adolescents and adults. Interestingly, participants’ capacity to generate and/or engage action representation followed a similar, although not identical, developmental progression. Furthermore, regression analysis demonstrated that even after controlling for general reaching ability, participants’ capacity to internally represent movement significantly predicted the efficiency with which they were able to correct their reach trajectory on the DSRT (i.e., online control); this association did not differ significantly across age. Taken together, these data are the first of their kind to provide direct empirical evidence that the nonlinear trend in improved online control through the critical 6- to 12-year period (and beyond) may be associated with more efficient action representation. Online control performance Consistent with earlier work (Ruddock et al., 2014; Wilson & Hyde, 2013), there was evidence of a significant reduction in movement time over childhood and into adulthood. This developmental change seen here is consistent with the documented pattern of central nervous system and neuromotor maturation (e.g., Casey et al., 2005) throughout this period, which sees incremental improvements in general movement efficiency. Movement time difference scores between non-jump and jump trials decreased significantly from early childhood (6–7 years) to late childhood (8–12 years) and into adolescence (13–17 years), with no significant difference between adolescents and young adults. Specifically, effect sizes indicated that the observed decrease in movement time difference scores was strongest between early and late childhood. To reiterate, the increase in movement time observed from non-jump trials (assumed to place few demands on online control) to jump trials (considered to place considerable demands on online control) is thought to reflect the time taken for the nervous system to account for unexpected target perturbation; the more efficient one’s ability to engage in online control of reaching, the smaller this difference is expected to be (see Hyde & Wilson, 2011a, 2011b). Thus, these results are consistent with the view that the ability to account for target perturbation improves substantially from early to late childhood, where incremental improvements are henceforth observed. Interestingly, the chronometric values reported here were largely consistent with the bulk of earlier work where comparable DSRTs had been adopted in pediatric samples (Hyde & Wilson, 2011a, 2011b, 2013; Wilson & Hyde, 2013), although marginally slower than in those instances where shorter distances between the home base and targets (i.e., display scaling) were used (Ruddock et al., 2014). Notably, however, the pattern of nonlinear performance improvements observed in the current

86

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

Fig. 5. Graphical representation of a non-jump trial (left) and a jump trial (right) on the DSRT.

study was remarkably consistent with these earlier developmental studies (Hyde & Wilson, 2013; Ruddock et al., 2014; Wilson & Hyde, 2013). It should be noted that each of these studies used a variant of the DSRT; hence, it is possible that the consistent profile of development observed across studies may reflect a task-specific effect rather than the broader development of online control. We argue, however, that this is unlikely given that the DSRT has shown to be a valid measure of online control of reaching, differentiating between patients groups where online control of reaching is a core feature of symptomatology (including posterior parietal patients; see Gréa et al., 2002) and healthy controls. The developmental trend in online control shown by movement time data was largely reinforced by kinematic analysis of performance. Consistent with earlier research, TC decreased significantly from early childhood to mid childhood, where it continued to undergo refinement into adolescence; only subtle improvements between adolescents and young adults were observed. This developmental progression was similar to the developmental progression observed in comparable studies (e.g., Hyde & Wilson, 2013; King et al., 2012; Wilson & Hyde, 2013). Effect sizes indicated that the strongest decrease in TC occurred between late childhood and adolescence. This pattern of performance appears somewhat inconsistent with the observed trend on movement time difference scores, which showed similar age-related decrements, although the largest decrement was from between younger and older children. However, as our own analysis has shown, although TC is dependent on one’s ability to accurately generate action representations, it is also influenced by other factors, including general reaching speed. In the case of 6- and 7-year-olds compared with 8- to 12-year-olds, only subtle differences in general reaching speed (i.e., non-jump reaching) were observed (g2p = .10). Hence, the difference in TC values observed between these two groups appears to be largely independent of general reaching efficiency and, as we argue below, largely reflects an improved capacity to generate internal action representations. Moreover, our sample of 8- to 12-year-olds was substantially slower on non-jump trials than 13- to 17-year-olds (g2p = .20). Thus, it may be that the large difference in TC values observed between these two groups reflects a degree of age-related improvements in general motor ability as well as improved action representation. Despite this, consistent with earlier work (King et al., 2012; Wilson & Hyde, 2013), the observed developmental progression of TC here suggests rapid improvements in children’s ability to correct reaching online (and by extension to represent action mentally) from early to late childhood, where incremental improvements are observed into adolescence and early adulthood. Taken together, chronometric and kinematic performance on the DSRT was largely consistent with earlier work demonstrating nonlinear improvements in the ability to account for unexpected target perturbation mid-flight from early to late childhood, where incremental improvements are observed

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

87

into adolescence and early adulthood (King et al., 2012; Wilson & Hyde, 2013). Based on the assumption that the ability to correct individuals’ reaching efficiently in response to unexpected target location changes is dependent on their capacity to generate and implement internal action representations, we argue that the largely nonlinear developmental trajectory of online control can be explained, at least partly, by the more efficient generation and implementation of action representations. Next, we consider MI performance to test the viability of this suggestion. MI performance The performance profile of all age groups on the hand rotation task was consistent with the use of an embodied MI strategy. Similarly to the biomechanical constraints associated with real movement, participants responded faster and more accurately when required to medially rotate (cf. laterally rotate) stimuli and displayed faster RTs and higher accuracy levels when responding to stimuli presented in the posturally congruent back view compared with the posturally incongruent palm view. The developmental trajectory of RT showed that response latency to stimuli decreased with age, which is consistent with the documented developmental progression of MI (Caeyenberghs et al., 2009; Toussaint, Tahej, Thibaut, Possamai, & Badets, 2013). Results revealed significant differences between each of the age groups, with the strongest decrease in RT between 6- and 7-year-olds and 8- to 12-year-olds. The developmental trajectory of RT was largely paralleled by improvements in accuracy across age. Here, analyses indicated that accuracy levels increased significantly from 6- and 7-year-olds to 8- to 12-year-olds to adolescents, with no significant differences observed between adolescents and adults. Again, the strongest increase in accuracy levels was observed between 6- and 7-year-olds and 8- to 12-year-olds in this study. Taken together, both chronometric and accuracy data suggest critical improvements in task performance after 7 years of age and more subtle improvements from there onward. This performance pattern is consistent with earlier developmental work using the hand rotation task (Butson et al., 2014; Caeyenberghs et al., 2009; Toussaint et al., 2013), demonstrating that MI shows a critical period of development after 7 years of age, where imagery performance has been shown to improve considerably with further incremental improvements observed into adolescence and adulthood (Conson, Mazzarella, & Trojano, 2014). Based on the assumption that performance

Fig. 6. Hand rotation task stimulus: Right hand in palm view at 135°.

88

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

on the hand rotation task provides insight into the integrity of the internal neural action representation (de Lange et al., 2008; Jeannerod, 2001; Munzert et al., 2009), we argue that this developmental profile is subserved, at least in part, by an improved capacity to generate and/or implement internal representations of action. Critically, the observed nonlinear improvements in MI largely parallel the developmental progression of online control in that participant performance on both metrics improved considerably at around 8 years of age, with only incremental improvements from childhood to adolescence and relative stability between adolescence and adulthood. Given that (a) current neurocomputational modeling posits that efficient online control of reaching is dependent on one’s ability to generate and monitor internal representation of movement and (b) MI is largely assumed to provide insight into one’s ability to accurately form and monitor these representations, the considerable overlap in developmental profile of performance on the DSRT and MI tasks in the current study provides preliminary evidence that there may be a functional association between one’s ability to generate internal representations of action (i.e., MI) and the efficiency with which one is able to implement fast corrective response to action. A direct statistical test of this proposed association is presented next.

A greater capacity for generating internal representations of action predicts developmental improvements in online control As expected, the initial two-step hierarchical regression model demonstrated that, after controlling for general reaching efficiency (i.e., non-jump reaching movement time), MI performance was a significant predictor of the efficiency with which participants were able to correct their reaching trajectory. Specifically, both RT and accuracy made significant unique contributions to TC, together accounting for approximately 20% of the variability in time to correction over and above general reaching speed. These findings mirror those of our earlier preliminary investigation of the association between imagery ability and reaching correction efficiency in healthy young adults, which showed that after controlling for general reaching speed, accuracy and efficiency on the hand rotation task predicted just under 25% of the variability in TC (Hyde et al., 2013). More specifically, the results of our analysis here demonstrate that faster and more accurate performance during the hand rotation task was associated with faster corrections to reaching following unexpected target perturbations. This pattern of results is consistent with neurocomputational modeling proposing that internal representations of movement are critical to the efficient in-flight correction of reaching (Izawa & Shadmehr, 2011; Shadmehr & Krakauer, 2008; Wolpert, Diedrichsen, & Flanagan, 2011). The inclusion of age into the model improved the prediction of TC values by a small (DR2 = .06) but significant amount, suggesting that TC was marginally influenced by age-related improvements over and above those attributed to general reaching speed and one’s ability to generate action representations. That age added only a small amount of explained variance to the model was not surprising considering the large amount of shared variability expected between general reaching speed (i.e., non-jump reaching), MI ability (i.e., hand rotation performance), and age. Indeed, approximately 50% of the variability explained in TC was shared variance between age and these earlier predictors; this likely reflects the age-related improvements in non-jump reaching and imagery performance documented earlier. However, both RT and accuracy made significant unique contributions to TC even after accounting for age in the model. The purpose of the final model, which included age as a potential moderator, was to investigate whether the observed association between imagery performance and TC varied according to age. Results showed that age did not significantly moderate the associations between either RT or accuracy and TC. Thus, results indicated that the association between each of the MI measures and the efficiency with which participants were able to correct reaching was similar across the age range in this sample and suggest that MI performance and age, while controlling for general reaching speed, provided the model with the best fit for explaining variability in TC. In line with recent computational modeling (Izawa & Shadmehr, 2011; Shadmehr & Krakauer, 2008; Wolpert et al., 2011) and our earlier pilot study using healthy adults (Hyde et al., 2013), our results provide qualified support for the view that the ability to generate action representations is a strong predictor of individuals’ ability to correct

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

89

their reaching in-flight and that this association is largely consistent across the critical period of neurocognitive development during the primary school years and beyond. Implications and limitations This study is only the second one to directly test the purported association between action representation and online control and the first one to do so in a developmental setting (see also Hyde et al., 2013). It is the first of its kind to provide empirical evidence that there is a predictive association between the documented nonlinear developmental improvement in online control through the critical 6- to 12-year period (and indeed beyond) and the capacity for generating internal neural representations of movement. This knowledge is critical to our understanding of online control and the neurocognitive mechanisms that support it. Despite the promising results obtained in this study, it is not possible to infer a causal association between developmental improvements in action representation and online control efficiency. Indeed, research has demonstrated that MI performance might reflect a degree of visuospatial working memory performance (see Paus, 2005, for a review). Given the documented developmental improvements in visuospatial working memory throughout childhood and into adolescence (e.g., Spencer-Smith et al., 2013), it is reasonable to presume that a degree of variability in the hand rotation performance in the current sample was due to this factor rather than individual differences in action representation. Similarly, in the case of time to reach trajectory corrections (i.e., jump trial performance on the DSRT), these corrections are likely influenced by factors other than the action representation, including basic cognitive and morphological factors. Where possible, we have attempted to control for these confounds experimentally; however, it is difficult to control for their cumulative effect. Still, in light of earlier computational and indirect empirical work suggesting a direct association between action representation and online control, and considering the consistent findings to this effect here, our results provide compelling empirical evidence for the importance of accurate internal representation of movement to the typical development of online control. Longitudinal evidence adopting combined behavioral and neuroimaging measures of performance in typically and atypically developing populations will likely shed further light on the purported causal association between action representation and online control of reaching and provide insight into the effects of potential developmental confounds. Conclusion The current research is the first study to provide empirical evidence that the well-documented nonlinear developmental progression of online control during the critical 6- to 12-year period is associated, at least in part, with a greater capacity to generate and/or engage internal neural action representations. Our findings support current neurocomputational theories of human reaching and extend preliminary evidence from healthy adults. Although further longitudinal research incorporating behavioral and neurophysiological measures is required in order to determine the causal association (if any) between the action representation and online control, results here are critical to our understanding of the development of online control in typically (and atypically) developing children. Acknowledgments Our sincere gratitude extends to the students, parents, and staff of those schools that participated in this research. In addition, we thank the Department of Education and Early Childhood Development (DEECD) for its support. Finally, we thank Tim Miles for his valued assistance during data collection. References Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage. Andersen, R., & Cui, H. (2009). Intention, action planning, and decision making in the parietal–frontal circuits. Neuron, 63, 568–583.

90

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

Blangero, A., Gaveau, V., Luauté, J., Rode, G., Salemme, R., Guinard, M., et al (2008). A hand and a field effect in on-line motor control in unilateral optic ataxia. Cortex, 44, 560–568. Butson, M., Hyde, C., Steenbergen, B., & Williams, J. (2014). Assessing motor imagery using the hand rotation task: Does performance change across childhood? Human Movement Science, 35, 50–65. Caeyenberghs, K., Tsoupas, J., Wilson, P. H., & Smits-Engelsman, B. C. M. (2009). Motor imagery development in primary school children. Developmental Neuropsychology, 34, 103–121. Casey, B. J., Tottenham, N., Liston, C., & Durston, S. (2005). Imaging the developing brain: What have we learned about cognitive development? Trends in Cognitive Sciences, 9, 104–110. Conson, M., Mazzarella, E., & Trojano, L. (2014). Developmental changes of the biomechanical effect in motor imagery. Experimental Brain Research, 226, 441–449. de Lange, F. P., Helmich, R. C., & Toni, I. (2006). Posture influences motor imagery: An fMRI study. NeuroImage, 33, 609–617. de Lange, F. P., Roelofs, K., & Toni, I. (2008). Motor imagery: A window into the mechanisms and alterations of the motor system. Cortex, 44, 494–506. Decety, J., & Grèzes, J. (1999). Neural mechanisms subserving the perception of human actions. Trends in Cognitive Sciences, 3, 172–178. Decety, J., Jeannerod, M., & Prablanc, C. (1989). The timing of mentally represented actions. Behavioural Brain Research, 34, 35–42. Deconinck, F. J. A., Spitaels, L., Fias, W., & Lenoir, M. (2009). Is developmental coordination disorder a motor imagery deficit? Journal of Clinical and Experimental Neuropsychology, 31, 720–730. Desmurget, M., & Grafton, S. (2000). Forward modeling allows feedback control for fast reaching movements. Trends in Cognitive Sciences, 4, 423–431. Ferraina, S., Battaglia-Mayer, A., Genoveslo, A., Archambault, P., & Caminiti, R. (2009). Parietal encoding of action in depth. Neuropsychologia, 47, 1409–1420. Frith, C., Blakemore, S., & Wolpert, D. (2000). Abnormalities in the awareness and control of action. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 355, 1771–1788. Gabbard, C. (2009). Studying action representation in children via motor imagery. Brain and Cognition, 7, 234–239. Geuze, R., Jongmans, M., Schoemaker, M., & Smits-Engelsman, B. (2001). Clinical and research diagnostic criteria for developmental coordination disorder: A review and discussion. Human Movement Science, 20, 7–47. Grafton, S. T., & Viswanathan, S. (2014). Rethinking the role of motor simulation in perceptual decisions. Advances in Experimental Medicine and Biology, 826, 69–90. Gréa, H., Pisella, L., Rossetti, I., Desmurget, M., Tilikete, C., Grafton, S., et al (2002). A lesion of the posterior parietal cortex disrupts online adjustments during aiming movements. Neuropsychologia, 40, 2471–2480. Hyde, C., Wilmut, K., Fuelscher, I., & Williams, J. (2013). Does implicit motor imagery ability predict reaching correction efficiency? A test of recent models of human motor control. Journal of Motor Behavior, 45, 259–269. Hyde, C., & Wilson, P. H. (2011a). Dissecting online control in developmental coordination disorder: A kinematic analysis of double-step reaching. Brain and Cognition, 75, 232–241. Hyde, C., & Wilson, P. H. (2011b). Online motor control in children with developmental coordination disorder: Chronometric analysis of double-step reaching performance. Child Care Health and Development, 37, 111–122. Hyde, C., & Wilson, P. H. (2013). Impaired online control in children with developmental coordination disorder reflects developmental immaturity. Developmental Neuropsychology, 38, 81–97. Izawa, J., & Shadmehr, R. (2011). Learning from sensory and reward prediction errors during motor adaptation. PLoS Computational Biology, 7, e1002012. Jeannerod, M. (2001). Neural simulation of action: A unifying mechanism for motor cognition. NeuroImage, 14, 103–109. Kail, R. (1991). Developmental change in speed of processing during childhood and adolescence. Psychological Bulletin, 109, 490–501. King, B., Oliveira, M., Contreras-Vidal, J., & Clark, J. (2012). The development of state estimation explains improvements in sensorimotor performance across childhood. Journal of Neurophysiology, 107, 3040–3049. Kosslyn, S. M., Digirolamo, G. J., Thompson, W. L., & Alpert, N. M. (1998). Mental rotation of objects versus hands: Neural mechanisms revealed by positron emission tomography. Psychophysiology, 35, 151–161. Kosslyn, S. M., Ganis, G., & Thompson, W. L. (2001). Neural foundations of imagery. Nature Reviews Neuroscience, 2, 635–642. Lange-Küttner, C. (1997). Development of size modification of human figure drawings in spatial axes systems of varying complexity. Journal of Experimental Child Psychology, 66, 264–278. Lange-Küttner, C. (2004). More evidence on size modification in spatial axes systems of varying complexity. Journal of Experimental Child Psychology, 88, 171–192. Macuga, K. L., & Frey, S. H. (2014). Differential contributions of the superior and inferior parietal cortex to feedback versus feedforward control of tools. NeuroImage, 92, 36–45. McCarron, L. T. (1997). McCarron assessment of neuromuscular development: Fine and gross motor abilities. Dallas, TX: Common Market Press. Mulliken, G., Musallam, S., & Andersen, R. (2008). Forward estimation of movement state in posterior parietal cortex. Proceedings of the National Academy of Sciences of the United States of America, 105, 8170–8177. Munakata, Y., Casey, B. J., & Diamond, A. (2004). Developmental cognitive neuroscience: Progress and potential. Trends in Cognitive Sciences, 8, 122–128. Munzert, J., Lorey, B., & Zentgraf, K. (2009). Cognitive motor processes: The role of motor imagery in the study of motor representations. Brain Research Reviews, 60, 306–326. Ochipa, C., Rapcsak, S., Maher, L., Rothi, L., Bowers, D., & Heilman, K. (1997). Selective deficits of praxis imagery in ideomotor apraxia. Neurology, 49, 474–480. Parsons, L. M. (1994). Temporal and kinematic properties of motor behavior reflected in mentally represented simulated action. Journal of Experimental Psychology: Human Perception and Performance, 20, 709–730. Parsons, L. M., & Fox, P. T. (1998). The neural basis of implicit movements used in recognising hand shape. Cognitive Neuropsychology, 15, 583–615.

I. Fuelscher et al. / Journal of Experimental Child Psychology 140 (2015) 74–91

91

Paus, T. (2005). Mapping brain maturation and cognitive development during adolescence. Trends in Cognitive Science, 9, 60–68. Pisella, L., Binkofski, F., Lasek, K., Toni, I., & Rossetti, Y. (2006). No double dissociation between optic ataxia and visual agnosia: Multiple sub-streams for multiple visuo-manual integrations. Neuropsychologia, 44, 2734–2748. Pisella, L., Gréa, H., Tilikete, C., Vighetto, A., Desmurget, M., Rode, G., et al (2000). An ‘‘automatic pilot’’ for the hand in human posterior parietal cortex: Toward reinterpreting optic ataxia. Nature Neuroscience, 3, 729–736. Plumb, M., Wilson, A., Mulrou, A., Brockman, A., Williams, J., & Mon-Williams, M. (2008). Online corrections in children with and without DCD. Human Movement Science, 27, 695–704. Reichenbach, A., Bresciani, J.-P., Peer, A., Bulthoff, H. H., & Thielscher, A. (2011). Contributions of the PPC to online control of visually guided reaching movements assessed with fMRI-guided TMS. Cerebral Cortex, 21, 1602–1612. Royston, P., Altman, D. G., & Sauerbrei, W. (2006). Dichotomizing continuous predictors in multiple regression: A bad idea. Statistics in Medicine, 25, 127–141. Ruddock, S. R., Hyde, C. E., Piek, J. P., Sugden, D., Morris, S., & Wilson, P. H. (2014). Executive systems constrain the flexibility of online control in children during goal-directed reaching. Developmental Neuropsychology, 39, 51–68. Shadmehr, R., & Krakauer, J. W. (2008). A computational neuroanatomy for motor control. Experimental Brain Research, 185, 359–381. Shadmehr, R., Smith, M., & Krakauer, J. (2010). Error correction, sensory prediction, and adaptation in motor control. Annual Review of Neuroscience, 33, 89–108. Sirigu, A., Duhamel, J. R., Cohen, L., Pillon, B., Dubois, B., & Agid, Y. (1996). The mental representation of hand movements after parietal cortex damage. Science, 273, 1564–1568. Spencer-Smith, M., Ritter, B. C., Mürner-Lavanchy, I., El-Koussy, M., Steinlin, M., & Everts, R. (2013). Age, sex, and performance influence the visuospatial working memory network in childhood. Developmental Neuropsychology, 38, 236–255. ter Horst, A. C., van Lier, R., & Steenbergen, B. (2010). Mental rotation task of hands: Differential influence number of rotational axes. Experimental Brain Research, 203, 347–354. Toussaint, L., Tahej, P. K., Thibaut, J. P., Possamai, C. A., & Badets, A. (2013). On the link between action planning and motor imagery: A developmental study. Experimental Brain Research, 231, 331–339. Van Braeckel, K., Butcher, P. R., Geuze, R. H., Stremmelaar, E. F., & Bouma, A. (2007). Movement adaptations in 7- to 10-year-old typically developing children: Evidence for a transition in feedback-based motor control. Human Movement Science, 26, 927–942. Williams, J., Anderson, V., Reddihough, D. S., Reid, S. M., Vijayakumar, N., & Wilson, P. H. (2011). A comparison of motor imagery performance in children with spastic hemiplegia and developmental coordination disorder. Journal of Clinical and Experimental Neuropsychology, 33, 272–282. Williams, J., Pearce, A. J., Loporto, M., Morris, T., & Holmes, P. S. (2012). The relationship between corticospinal excitability during motor imagery and motor imagery ability. Behavioural Brain Research, 226, 369–375. Williams, J., Reid, S. M., Reddihough, D. S., & Anderson, V. (2011). Motor imagery ability of children with congenital hemiplegia: Effect of lesion side and functional level. Research in Developmental Disabilities, 32, 740–748. Wilson, P. H., & Hyde, C. (2013). The development of rapid online control in children aged 6–12 years: Reaching performance. Human Movement Science, 32, 1138–1150. Wolpert, D. M., Diedrichsen, J., & Flanagan, J. R. (2011). Principles of sensorimotor learning. Nature Reviews Neuroscience, 12, 739–751.