INTRODUCTION
Individuals with spina bifida meningomyelocele (SBM), the most common and severe form of spina bifida, have deficits in upper (Grimm, 1976) as well as lower limb control. Individuals with hydrocephalus, most with SBM, have impairments in persistent motor control, strength, balance, and fine motor skills (Hetherington & Dennis, 1999). Although it might be supposed that impaired motor performance would be associated with deficient motor learning, recent evidence suggests that SBM need not prevent successful acquisition of new motor skills.
According to Doyon and Benali (2005), motor learning involves motor sequence learning (i.e., gradual acquisition of movements into a smoothly executed behavior) and motor adaptation (i.e., the ability to respond to changes in the environment). From examinations of these two domains, it appears that the motor control deficits in individuals with SBM do not entail motor learning deficits. Edelstein et al. (2004) demonstrated preserved motor sequence learning in individuals with SBM on a mirror drawing task, even when task performance was poor. Intact motor adaptation in individuals with SBM was demonstrated on a task requiring prism adaptation to distorted visual information (Colvin et al., 2003). Children with SBM have difficulty visually tracking targets using smooth pursuit eye movements (Salman et al., 2005), conjugate eye movements serving to stabilize the image of a moving target on the fovea for high definition vision (Leigh & Zee, 1999). Nevertheless, the same children learn to adapt to error information by correcting saccades (Salman et al., in press).
The dissociations between motor performance and motor learning have been demonstrated, not only across tasks through a comparison of motor function (Grimm, 1976; Hetherington & Dennis, 1999) and motor learning studies (e.g., Colvin et al., 2003) but also within the same sample of participants (Edelstein et al., 2004; Salman et al., 2005, in press). Apparently, children with SBM have proficient nonreflexive motor learning skills regardless of the effector used (i.e., eyes, hands, or arms), even for tasks on which their motor performance is impaired.
For motor adaptation, the evidence for intact learning is clear, albeit incomplete. The key study of motor adaptation to prismatically distorted visual information (Colvin et al., 2003) studied adaptation to a new visual environment, measuring dominant arm accuracy in terms of absolute distance from the target. It is not known whether individuals with SBM show similar adaptation with a highly constrained movement under normal visual conditions with either the dominant or nondominant arm. It is unknown whether intact motor adaptation can be demonstrated, not only in terms of absolute accuracy but also in terms of movement speed and rate of adaptation on learning curves. There is no information about adaptation flexibility in terms of readaptation once the original environmental conditions are restored. In addition, the brain correlates of motor adaptation have been explored for eye movements but not for movements of the hand and limb. Although recent evidence suggests no correlation between linear measures of the cerebellum and saccadic adaptation (Salman et al., in press), there is no information about correlations between brain and motor adaptation for hand and upper limb.
The cerebellum has long been implicated in motor learning (see Albus, 1971; Marr, 1969). Adults with acquired cerebellar lesions have defective adaptation to prismatic distortion (Weiner et al., 1983). In a test of motor adaptation on a constrained ballistic arm movement task, Deuschl et al. (1996) used an elbow goniometer task in adult patients with cerebellar disease. In this task, the forearm and wrist are fixed to a lever such that the elbow joint can be moved freely, and participants make fast arm movements to match a target alternating between two positions on a computer screen. After baseline rate and accuracy are measured, the gain is changed, and the compensatory adaptations are tested. The procedure is similar to the “JUMP” task used by Thach and colleagues, which they used to demonstrate a role for the cerebellar cortex in motor adaptation in both humans and monkeys (Keating & Thach, 1990, 1991; reviewed in Thach, 1998). When applying the task of Deuschl et al., adult patients with cerebellar atrophy demonstrated less steep learning curves, providing support for cerebellar involvement in motor learning in the mature brain.
In some theoretical accounts (e.g., Willingham, 1998), the cerebellum is involved in motor performance rather than in motor learning by virtue of its involvement in attention and motor timing (Seidler et al., 2002). In other accounts, some motor functions may be shared by the cerebellar and basal ganglia systems through a mutual pathway linking these two structures (Graybiel, 2005) or by a multipath system. In Doyon's model of motor skill acquisition (Doyon & Benali, 2005; Doyon et al., 2003; Doyon & Ungerleider, 2002), corticostriatal and corticocerebellar systems are recruited during the fast learning phase of acquisition, with only one circuit (the corticostriatal pathway, in performing a new motor sequence, or the corticocerebellar pathway, in motor adaptation) subserving later learning.
The brains of individuals with SBM are structured in a complex manner, rather than just being globally reduced in volume. Of particular interest to current theories of motor learning, they have relative disruption of the cerebellum, which is dysmorphic, herniated through the tentorial incisure and foramen magnum, and reduced in volume in the gray and white matter of the lateral cerebellar cortex (Barkovich, 2000; Dennis et al., 2006; Fletcher et al., 2005; Madsen et al., 2002). Children with SBM show impairments in delimited cerebellar functions such as short-duration timing (Dennis et al., 2004), similar to those in adults with cerebellar lesions (Ivry & Keele, 1989; Ivry et al., 1988; Mangels et al., 1998; Nichelli et al., 1996; Penhune et al., 1998; Rao et al., 1997). Some cerebellar regions are relatively preserved. For example, there are similar or larger volumes of gray and white matter in the medial cerebellum (Dennis et al., 2004; Edelstein et al., 2004) and enlarged linear measurements of the cerebellar vermis (Salman et al., 2006).
In this study, we revisit the issue of motor skill learning in children with SBM, using ballistic movement learning on the elbow goniometer task, which requires fast, ballistic, preprogrammed movements. We measure constrained movement under normal visual conditions, as well as adaptation rate compared with both baseline and readaptation. Our first aim was to establish whether motor learning was preserved, using a rigorous procedure for assessing ballistic arm movement. We hypothesize that intact motor learning will be demonstrated in individuals with SBM, regardless of spinal lesion level. Ballistic movements may require cerebellar involvement (Deuschl et al., 1996; Thach, 1998), but given the recent proposal by Desmurget et al. (2004) that the basal ganglia system is critical for the planning of movement amplitude, they may also require striatal function. We assess brain volumes and explore relations between motor adaptation and volumetric measures of the cerebellum and pericallosal gray matter volume, a region containing the basal ganglia. Our second aim was to relate ballistic movement learning to measures of voxel-based brain morphometry. We hypothesize that, as in the larger sample from which the present participants are derived (Fletcher et al., 2005), individuals with SBM will show volume reductions in cerebellar and pericallosal brain regions. Previous motor learning measures have not been related to cerebellar volume in this population (Edelstein et al., 2004), so we hypothesize no significant relation between cerebellar volumes and ballistic motor learning. To the extent that pericallosal volumes are a proxy measure for the basal ganglia, we hypothesize a relation between pericallosal volumes and ballistic motor learning.
METHODS
Participants
Participants were 137 children and adolescents between 8 and 19 years of age recruited from clinics at two sites: The Hospital for Sick Children in Toronto (n = 77) and the University of Texas Health Science Center—Houston (n = 60). The study was approved by the ethics boards at each site, and the data were obtained in compliance with institutional regulations and the guidelines of the Helsinki Declaration for human research. Participants gave informed assent or consent, and their parents gave informed consent before completing the study. One group (n = 102) had been diagnosed with SBM at birth and had been treated with a shunt shortly thereafter. Of those children, 27 had no shunt revision, 32 had 1 revision, 30 had 2–4 revisions, 11 had 5–9 revisions, and 2 children had more than 10 shunt revisions. The other group comprised typically developing, age-matched controls (n = 35), recruited through community advertisements, hospital newsletter advertisements, and word of mouth among staff at each testing facility. All participants had IQ scores ≥ 70 on at least one of the Verbal Reasoning or Abstract/Visual Reasoning subtests of the Stanford-Binet Test of Intelligence-Revised (Thorndike et al., 1986). Individuals were excluded from participation if upon interviewing and a review of medical records it was established that they had a diagnosis of neurological disorders unrelated to SBM, severe psychiatric disorder that precluded adequate cooperation (autism, psychosis, oppositional-defiant disorder), uncontrolled seizure disorder, or uncorrected sensory disorder, or if they were unable to control both upper limbs such that the task could not be completed reliably. The exclusions were ascertained by questionnaires completed by the parents (SNAP-IV; Swanson, 1992), a Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition checklist for autism and pervasive developmental disorders, a medical history chart reviewed by a research nurse, and observations of the child's behavior when evaluated.
Individuals with SBM have lesions at various levels of the spinal cord, providing a source of principled, within-group variability. Upper level lesions are associated with greater reductions in cerebellum volumes (Fletcher et al., 2005). To explore one source of biological variability associated with greater reductions in cerebellar volumes, participants with SBM were divided further into upper spinal lesion (T12 and higher; n = 21) and lower spinal lesion (L1 and lower; n = 81) groups, according to current taxonomies (Fletcher et al., 2005; Park et al., 1992).
Table 1 provides information on IQ and sociodemographic characteristics. As expected, children with SBM differed from controls in IQ [F(2,134) = 36.93, p < .001]: children with both lower and upper spinal lesions had lower average IQs compared with controls, although the two SBM lesion level groups did not differ significantly from one another (p > .05). As well, groups differed in socioeconomic status [F(2,134) = 5.69, p < .005], with children with lower spinal lesions showing lower socioeconomic status than controls (p < .005) but not those with upper level lesions. Differences in age, gender, and ethnicity (Table 1) were not statistically significant (p > .05).
Demographic information for control and upper and lower lesion SBM participants
Brain Imaging Procedures
Using structural magnetic resonance imaging (MRI) brain scans that were artifact-free, quantitative analysis of cerebellar volume was performed for 68 participants (42 children with lower lesions, 12 children with upper lesions, 14 controls) and of pericallosal gray matter volume, a proxy for the regional cerebral volumes that would include the basal ganglia, for 63 participants (34 children with lower lesions, 10 children with upper lesions, 19 controls).
The participants providing the cerebellar volume data had a mean age of 12.54 years (SD = 2.63), a mean IQ of 93.15 (SD = 14.45), and a mean socioeconomic score of 43.28 (SD = 13.60). Differences between the lesion level groups in age, gender, and ethnicity were not statistically significant (p > .05), as in the larger sample. Groups differed in IQ [F(2,65) = 15.98, p < .001], with both groups of participants with SBM scoring lower than controls (p < .001). As well, groups differed in socioeconomic status [F(2,65) = 3.86, p < .05], with a trend for the lower spinal lesion level group to score lower than the upper lesion level group (p = .05). Of the 54 participants with SBM, 12 had no shunt revision, 17 had 1 revision, 15 had 2–4 revisions, 9 had 5–9 revisions, and 1 child had more than 10 shunt revisions.
The participants providing the pericallosal gray matter volume data had a mean age of 12.45 years (SD = 2.65), a mean IQ of 94.57 (SD = 13.83), and a mean socioeconomic score of 43.83 (SD = 12.60). Differences between the lesion level groups in age, gender, and ethnicity were not statistically significant (p > .05). Groups differed in IQ [F(2,60) = 24.66, p < .001], with both groups of participants with SBM scoring lower than controls (p < .001). Groups also differed in socioeconomic status [F(2,60) = 5.89, p < .01], with the lower spinal lesion level group scoring lower than both the upper lesion level group and the controls (p < .05). Of the 44 participants with SBM, 9 had no shunt revision, 15 had 1 revision, 14 had 2–4 revisions, and 6 had 5–9 revisions.
Image Acquisition
Three sets of images were acquired with external fiducial markers to coregister and position normalize the scans, including a T1-weighted coronal series for assessment of white and gray matter and a T2-weighted coronal series for assessment of cerebrospinal fluid (CSF). An initial series [spin-echo T1-weighted sagittal localizer, field of view (FOV) 24, TR 500, TE14, 256 × 192 matrix, 3 mm skip 0.3, two repetitions] was used for anatomical landmark identification. One whole-brain coronal series (of contiguous 1.5-mm slices across the whole brain) consisted of a fast spin-echo proton density and heavily T2-weighted images (FOV 20, TR 4000, TE1 15, TE2 112, 256 × 192 matrix, with two repetitions). Another whole-brain coronal series consisted of a three-dimensional spoiled grass gradient echo contiguous 1.5-mm coronal series (TR21, TE4, Flip angle 35°, 124 locations, 256 × 192 matrix, one repetition).
Quantitative Measures
Segmentation was based on a fully automated fuzzy cluster analysis (Pao, 1989) developed specifically for children with hydrocephalus (Brandt et al., 1992, 1994, 1996). The T1-weighted scan volume, which provides superior white–gray contrast compared with the T2-weighted scan, was used to obtain white and gray matter tissue volumes. The T2-weighted scan was fuzzy clustered separately from the T1-weighted scan to extract CSF volumes, and this was used to adjust the white and gray matter volume measures obtained from the T1-weighted volume.
All slices for which the cerebrum or cerebellum could be visualized were segmented. For the cerebrum, three front-to-back regions within each hemisphere were measured: precallosal, pericallosal, and retrocallosal (Filipek et al., 1992). The pericallosal region subtended the coronal brain volume extending from the most anterior to the most posterior part of the corpus callosum. The precallosal region extended fully frontally from the pericallosal region, whereas the retrocallosal region extended fully posteriorly from the pericallosal region. The gray matter volume in the pericallosal region served as a proxy for the regional cerebral volumes, including the basal ganglia.
Separate tissue volumes (white matter, gray matter, CSF) were obtained for the whole cerebellum, medial cerebellum, and lateral cerebellum. An algorithm was developed to estimate cerebellar volumes that would correspond to medial and lateral cerebellar regions by identifying the midsagittal cerebellum slice from the coronal series and the primary fissures to the left and right of the middle cerebellar MRI slice. In typically developing children, the vermis represented on average 11% of the total cerebellum. This estimate was used to define a medial cerebellar volume by identifying the areas 5.5% on either side of the midline, with the remainder being defined as the left or right lateral regions. The medial cerebellar volume is, therefore, a proxy for the vermis volume and may be subject to some error of measurement in precisely defining the vermis across individual cases.
Elbow Goniometer Task
The elbow goniometer task (Deuschl et al., 1996) was administered using an IBM-compatible personal computer, with participants facing the monitor at a 55-cm distance. The wrist and forearm (right and left hand with testing order assigned randomly based on the odd/even counterbalance condition for another task) were comfortably fixed to a lever with Velcro adhesive, and the participant was asked to make a fist (see Figure 1). The lever was attached to a table such that the elbow joint only could be moved freely over an angle of up to 180° (i.e., arm fully extended) with a starting elbow angle of approximately 90° (i.e., forearm parallel to the edge of the table). The displacement of the elbow was continuously monitored by a goniometer attached to the base of the lever. The lever position was sampled at 100 Hz and smoothed with a five point moving average.
Elbow goniometer task apparatus.
On each trial, a 1.5-cm2 red square target appeared on the left or right side of the screen, with these positions separated by a distance of 15 cm. Elbow joint angle was represented by a red circular cursor with a diameter of 1.3 cm, which was controlled by moving the lever. The participant was required to move the cursor to the target position by making a single, rapid arm movement either toward or away from the body. Each trial involved a flexion and extension movement component, with the destination angle being equally large (15° or 21°, depending on the phase condition) but opposite in direction for each component. Once the cursor achieved the target position, the target alternated to the other side of the screen for the second component of the movement or initiated a new trial if both components had been completed.
The emphasis during the task was on speed rather than accuracy. Participants were told to plan the movement in advance and then to complete the movement as quickly as possible. The movement was considered started when the lever velocity first exceeded 20°/s and was considered ended when it subsequently first changed direction. Although participants could correct the arm movement to reach the target (and needed to do so to initiate the next trial), only the initial ballistic movement was recorded.
The participant began the test trials following a practice session of six trials (12 flexion and extension movement components). The task was divided into three blocks of 20 trials (including both flexion and extension components) each. The first block of trials (learning phase) involved a gain (relation between cursor movement and angle change with arm movement) of 1.0 cm/°. For the second block of 20 trials (adaptation phase), the gain was changed to 1.4 cm/°, such that the participant had to learn to readjust elbow movement. Finally, for the third block of trials (learning reactivation phase), the gain was restored to 1.0 cm/°. After completing all 60 trials with one arm, the participant repeated the practice and test trials with the other arm. Hand dominance was assessed by noting the hand used to complete the Developmental Test of Visual Motor Integration (Beery, 1989), a task that was part of a larger test battery. Of the 137 participants, 2 controls and 22 SBM participants were left-hand dominant; all other participants were right-hand dominant. Individuals were excluded from the study if they could not complete the task with both arms (n = 2) or if their data were not reliable (n = 1). An additional eight participants were excluded because their time to complete the movement exceeded 5 s on one or more trials.
Data Analysis
The time to complete the movement and the elbow displacement error (i.e., difference between the elbow angle at the end of the movement and the destination angle) were recorded for both the flexion and extension components of each trial for each arm. Time was measured to assess whether participants with SBM took longer to complete the ballistic movement; collapsed over hand, no significant differences between groups were evident [F(2,134) = 2.39, p > .05]. Overshoots of the target were considered positive errors, and undershoots were negative errors. Flexion and extension errors were pooled for analysis in accordance with Deuschl et al. (1996). Mean error values were calculated for each trial for each group of participants.
The errors within each block of 20 trials were modeled as an exponential decay for each participant. The starting error rate (κ), decay rate (τ), and asymptotic error rate (δ) parameters were chosen to minimize the sum of squares error between the flexion and extension errors and the following function:
in which yx is the error on trial x. That is, these parameters were chosen to minimize the discrepancy between the errors predicted by Equation 1 and the actual flexion and extension errors; this discrepancy is represented by the following function:
in which yx(f) and yx(e) are the actual flexion and extension errors, respectively, on trial x.
RESULTS
Adaptive Learning and Performance
Mean values for observed error across trials are shown in Figure 2. The derived outcome measures for these analyses were starting error rate (κ), decay rate (τ; the number of trials to reach the asymptote), and asymptotic error rate (δ) obtained for each participant (Figure 3).
Elbow goniometer raw error scores from children with spina bifida meningomyelocele (SBM; upper and lower spinal lesions combined) and from typically developing controls. The mean errors (undershoot or overshoot measured in degrees) on each trial are averaged over all participants.
Elbow goniometer derived measures from children with spina bifida meningomyelocele (SBM; upper and lower spinal lesions combined) and from typically developing controls. Top: Starting error rate during the learning, adaptation, and learning reactivation phases. Middle: Decay rate during the learning, adaptation, and learning reactivation phases. Bottom: Asymptotic error rate during the learning, adaptation, and learning reactivation phases (see the Methods section for more information on the derived measures).
The derived measures were entered into separate repeated measures analyses with the within-subject factors Phase (learning, adaptation, learning reactivation) and Hand (dominant, nondominant) and the between-subject factor Group (no lesion, lower lesion, upper lesion). No main effects of Group achieved statistical significance: F(2,134) = 1.34, p > .05; F(2,134) = .10, p > .05; F(2,134) = .47, p > .05, for starting error rate; decay rate; and asymptotic error rate, respectively, nor were any interactions with Group significant. To protect against Type II errors, we computed the effect sizes (d) for the mean differences between the control and SBM groups (collapsing the lesion level groups to maximize power) for each phase of each of the three derived measures. As reported in Table 2, all effects are negligible (< .2), with the exception of the effect size for asymptotic error rate (−.33) during the learning phase, which is in the small (.2–.5) range. This value relates to the relatively peripheral issue of performance deficits, rather than motor learning, which is represented by decay rate in the derived measures. We note that the original paradigm with 10 adult control participants and 10 cerebellar patients of heterogeneous pathology reported significant differences in motor learning (Deuschl et al., 1996). Thus, it is unlikely that increasing the number of participants in the present study would alter the present pattern of results.
Effect sizes for the mean differences (d) in children with SBM (n = 102) and in typically developing controls (n = 35)
The only effect to attain significance in each analysis was Phase, which was significant for starting error rate [F(2,268) = 114.74, p < .001], decay rate [F(2,268) = 12.37, p < .001], and asymptotic error rate [F(2,268) = 6.69, p < .005; see Figure 3]. As the sphericity assumption was violated for both the starting error rate and asymptotic error rate analyses, but the effect remained significant with the Greenhouse–Geisser correction, paired t tests with a Bonferroni correction were used to explore these main effects. These post hoc tests indicated that starting error rate was higher in the adaptation phase following the gain change than during the learning or learning reactivation phases (p < .05). As well, starting error rate was higher during the learning phase than during learning reactivation (p < .005). By contrast, decay rate was higher in the learning phase than in the adaptation and learning reactivation phases (p < .005), and was equivalent for the latter two phases (p > .05). Performance with respect to asymptotic error rate was lower during the adaptation phase than the other two phases (p < .005) and was not different during the learning or learning reactivation phases (p > .05). Thus, participants demonstrated the highest starting error rate but the lowest asymptotic error rate during the adaptation phase and the greatest number of trials to reach asymptote during the initial learning phase. Further analyses indicated similar findings with respect to decay rate regardless of whether starting error rate and/or asymptotic error rate were held constant to control for differences across participants and phases, respectively, or were free to vary. As well, additional analyses were done with Hand (left, right) as a within-subject factor and returned essentially the same results as those investigating dominant versus nondominant hand as well as a trend for a main effect of Hand for asymptotic error rate [F(1,134) = 3.68, p = .057]. All individuals tended to have a higher asymptotic error rate for their left hand than their right hand.
To assess the possible impact of shunt revisions on motor adaptation, separate analyses of variance with shunt revision group as a between-subject factor were conducted on starting error rate, decay rate, and asymptotic error rate. Shunt revision groups were categorized as listed in the Methods section, with the 2 participants with more than 10 revisions being grouped with the 11 participants with 5–9 revisions for the purposes of these analyses. No significant effects or interactions with learning phase were obtained from these analyses, indicating that the number of shunt revisions has little impact on this motor learning task.
Cerebellar and Pericallosal Volumes
Measures of cerebellar volume by region and tissue type are given in Table 3. Comparisons of regional differences in cerebellar volumes were made using Group (no lesion, lower lesion, upper lesion) by Region (medial, lateral-left hemisphere, lateral-right hemisphere) by Tissue Type (CSF, gray, white) repeated measures analysis of variance (ANOVA). Groups differed in overall cerebellar volume [F(2,65) = 31.29, p < .001], but this effect was not explored because the significant three-way interaction [F(8,260) = 36.86, p < .001] indicated that the differences between the groups were not simply overall volume reductions but instead reflected a different pattern of tissue type organization in the cerebella of children with SBM. This interaction remained significant when corrected for a violation of the sphericity assumption, but because of this violation, t tests appropriate for independent samples with a Bonferroni correction (p < .017) were used to compare differences between the groups.
Cerebellar volumes (cm3) and standard deviations by region and tissue type in children with upper (n = 12) and lower (n = 42) lesion SBM and in typically developing controls (n = 14)
As is displayed in Table 3, both SBM lesion groups had a significantly smaller volume of medial region CSF, smaller white and gray matter volumes in both the lateral left and right cerebellar hemispheres, and a significantly smaller volume of lateral CSF in the left hemisphere compared with controls. Additionally, participants in the lower level SBM group had significantly larger white matter volumes than controls in the medial cerebellum.
Differences in cerebellar volume were also evident as a function of lesion level in the participants with SBM. Differences were not evident in medial CSF volume, but gray matter volume in the medial cerebellum was larger for individuals with lower level lesions than those with upper lesions. Children with upper lesions additionally had smaller white and gray matter volumes than those with lower lesions in both lateral hemispheres.
A one-way ANOVA with a between-subject factor of Group (no lesion, lower lesion, upper lesion) was used to compare differences in pericallosal gray matter volume (averaged across hemisphere) and yielded a significant main effect of Group [F(2,60) = 10.78, p < .001]. A Tukey post hoc test revealed that controls had a significantly larger pericallosal gray matter volume (M = 219.19 cm3; SD = 21.59) than both SBM groups (p < .05), who also differed from each other in volume (lower lesion M = 203.25; SD = 20.80; upper lesion M = 182.49; SD = 16.15).
To assess whether participants with the greatest degree of cerebellar compromise also had the most reduced pericallosal gray matter volume, correlations were computed between the six combinations of tissue-type (CSF, gray, white) and region (medial, lateral) and the pericallosal gray matter volume using the data from participants with SBM only. Volumes of medial and lateral CSF, lateral white matter, and medial and lateral gray matter were significantly correlated with pericallosal gray matter volume [r(36) = .33, .46, .32, .56, and .57, respectively, p < .05]. Thus, children with SBM had reduced lateral but not medial cerebellum and less volume in the pericallosal gray matter that includes the basal ganglia. Furthermore, reductions in cerebellar volume corresponded with decreased pericallosal gray matter volume.
Correlations Between Adaptive Learning and Cerebellar and Pericallosal Volumes
Partial correlations controlling for age and comparing the individual cerebellar and pericallosal gray matter volumetric data with the decay rates (τ) from the elbow goniometer task were calculated for upper and lower lesion SBM groups combined (n = 54 and 44 for the cerebellar and pericallosal gray matter analyses, respectively). Data from one SBM individual were subsequently dropped from the cerebellar analyses due to extreme values (well below 2 standard deviations) on all cerebellar volume measures. Decay rate, the measure of learning, did not correlate with the cerebellar volume measures (p > .05), although decay rate during the adaptation phase was negatively correlated with pericallosal gray matter volume [r(41) = −.37, p < .05]. Learning was faster for those individuals with more pericallosal gray matter volume, the area including the basal ganglia.
DISCUSSION
In initial learning of a ballistic movement, recalibration of the learned movement after a gain change, and learning reactivation when the initial conditions were restored, children with SBM and controls performed comparably with respect to initial error rate, decay rate, and asymptotic error rate. Upper and lower spinal lesion level groups showed similar motor adaptation during all phases. Children with SBM had smaller cerebellar hemispheric volumes that were correlated with smaller pericallosal gray matter volumes, although only the latter volumes were negatively correlated with adaptation decay rate.
Intact motor adaptation in individuals with SBM has clinical implications. Despite dramatic lower limb dysfunction and upper limb motor deficits, these children show intact motor learning over different domains, with different effectors, and with different forms of movement. Rehabilitation experts may be able to exploit preserved motor learning in motor rehabilitation programs for which these children previously might have been considered ineligible.
The present data are relevant to several issues, including the distinction between motor performance and control, on the one hand, and motor adaptation and learning, on the other; the preservation of motor adaptation in children with SBM and upper spinal lesions; brain dysmorphology in individuals with SBM; the relative roles of the cerebellum and basal ganglia in motor learning and adaptation; and the question of age-based functional plasticity in motor adaptation following congenital cerebellar compromise.
Motor performance of individuals with SBM was better in a ballistic task than in a motor sequence learning task in which they made more errors than controls (Edelstein et al., 2004), perhaps because the former requires execution of a single planned movement, whereas motor sequence learning requires ongoing performance adjustments and revisions. Although learning assessed by improvement over trials seems intact for both motor adaptation and motor sequence learning, accuracy of motor performance is relatively better in the former. In general, children with SBM have difficulty with adaptive control tasks that involve constant performance adjustments (Dennis et al., 2006).
Spinal lesion level was associated with differences in cerebellar hemispheric volumes, although not with differences in motor learning. This finding means that the significant motor performance impairments of children with upper spinal lesions (Fletcher et al., 2005) do not include specific deficits in motor adaptation.
Children with SBM have significant reductions in cerebellar volumes (see also Edelstein et al., 2004; Fletcher et al., 2005, for reports based on the same cohort) without concomitant deficits in motor adaptation. Individuals with SBM have lateral cerebellar volume reductions in the range of 24–43%, which may be sufficient to support a mediated role for motor learning, but which, of interest, is not sufficient to support normal core cerebellar functions such as short-duration timing (Dennis et al., 2004). Other areas of the cerebellum may contribute to motor adaptation; specifically, medial portions of the cerebellar gray and white matter, which were actually comparable or enlarged in volume in the participants with SBM, relative to controls, may subserve motor adaptation on a ballistic task. The cerebellar vermis, for example, which is enlarged in participants with SBM compared with controls but does not differ in size based on lesion level (Salman et al., 2006), may contribute to successful learning and performance on this task.
That children with SBM have significant pericallosal gray matter volume loss shows that SBM is not exclusively a white matter disorder and is consistent with animal models of hydrocephalus (del Bigio, 1993), which report effects of hydrocephalus on gray matter due to cell death, migrational defects, and other factors. Although cerebellar and pericallosal volumes were correlated with each other, only the latter were correlated with adaptation, which suggests the hypothesis that both cerebellum and basal ganglia contribute to motor adaptation in SBM but that the former involves a less direct relationship than the latter, being mediated through the basal ganglia. Further evidence supporting a mediated role of the cerebellum is the absence of motor adaptation differences between the SBM lesion level groups, despite significant differences in cerebellar volume.
Whereas the voxel-based morphometry measures of the cerebellum provide relatively direct indices of cerebellar volumes, the pericallosal gray matter volume was a proxy for basal ganglia volumes. The correlations of brain volumes with adaptive learning were exploratory and based on relatively small samples that were heterogeneous with respect to lesion level. Despite these limitations, our data are strikingly consistent with current views of the role of the basal ganglia in motor learning (e.g., Graybiel, 2005) and in the planning of movement amplitudes (Desmurget et al., 2004). Much of the neurobiology of motor learning still remains to be understood, and future studies using methods such as diffusion tensor imaging to look at structural connectivity and functional neuroimaging will be useful in providing direct brain measures to test causal models of the brain–behavior relationships in adaptive motor learning. Motor learning involves, not only acquisition, but also stages of consolidation, automaticity, and very long-term retention without additional practice (Doyon & Benali, 2005). It remains to be understood whether intact ballistic motor learning in individuals with SBM is sustained in the face of explicit motor interference and over very long periods of time.
Motor adaptation, which is not dependent on conscious recall of the initial learning exposure, is a form of implicit learning, or learning without the intention to learn. Recent findings showing intact motor sequence learning (Edelstein et al., 2004) and motor adaptation (Colvin et al., 2003; Salman et al., in press) suggest that implicit motor learning may be appropriately developed in children with SBM, in contrast to adults with cerebellar lesions. The present study adds two pieces of new information to this literature: learning, adaptation, and readaptation are preserved for ballistic movements, and adaptation is related to pericallosal gray matter, a proxy for basal ganglia volumes. In both children and adults, motor learning may be subserved by a neural system that includes the basal ganglia as well as the cerebellum. Any functional plasticity in this system based on chronological age may concern a more mediated relation between the cerebellum and motor adaptation in the younger brain.
ACKNOWLEDGMENTS
This work was funded by the National Institute of Child Health and Human Development Grant P01 HD35946 “Spina Bifida: Cognitive and Neurobiological Variability”. No financial conflicts of interest exist with respect to this manuscript. The authors thank Kim Copeland, Paul Cirino, Joanne Robitaille, Jennifer Janes, Andrea Martin, Amy Walker, Irene Townsend, and Susan Inwood. These data were presented as a poster at the Symposium on Spina Bifida conference held May 10–12, 2006 in Chicago, Illinois. The information in this manuscript and the manuscript itself is an original work that is not currently under review elsewhere and has not been published previously in any form.
