Structural and functional bases for individual differences in motor learning

Abstract

People vary in their ability to learn new motor skills. We hypothesize that between-subject variability in brain structure and function can explain differences in learning. We use brain functional and structural MRI methods to characterize such neural correlates of individual variations in motor learning. Healthy subjects applied isometric grip force of varying magnitudes with their right hands cued visually to generate smoothly-varying pressures following a regular pattern. We tested whether individual variations in motor learning were associated with anatomically colocalized variations in magnitude of functional MRI (fMRI) signal or in MRI differences related to white and grey matter microstructure. We found that individual motor learning was correlated with greater functional activation in the prefrontal, premotor, and parietal cortices, as well as in the basal ganglia and cerebellum. Structural MRI correlates were found in the premotor cortex [for fractional anisotropy (FA)] and in the cerebellum [for both grey matter density and FA]. The cerebellar microstructural differences were anatomically colocalized with fMRI correlates of learning. This study thus suggests that variations across the population in the function and structure of specific brain regions for motor control explain some of the individual differences in skill learning. This strengthens the notion that brain structure determines some limits to cognitive function even in a healthy population. Along with evidence from pathology suggesting a role for these regions in spontaneous motor recovery, our results also highlight potential targets for therapeutic interventions designed to maximize plasticity for recovery of similar visuomotor skills after brain injury.

Figure 1

(A) Mean and standard error of the 95th percentile (p₉₅) of the tracking error in the Sequence (red) and Random (light blue) conditions at each block across all subjects (n = 12). (B) The p₉₅ of the tracking error in the Sequence (left) and Random (right) conditions calculated at block 1 and at Block 10 (Δp₉₅) in all subjects (n = 12). There was a significant reduction in Δp₉₅ for the Sequence (paired t test: t(11) = 4.3, P < 0.001) but not for the Random (paired t test: t(11) = −0.56, P > 0.1) condition. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 2

Task‐related functional networks (n = 12). Group mean brain activation for Sequence vs. Rest (A), Random vs. Rest (B) and Sequence vs. Random (C) contrasts, as well as linear reduction in Sequence (vs. Rest) related activation contrast (D), Z > 2.3, P < 0.05 corrected. R: right; L: left. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 3

Regions where BOLD signal change correlates positively with individual learning scores (Δp₉₅), Z > 2.3, P < 0.05 corrected. Higher BOLD signal changes in the Sequence vs. Rest (A) and Random vs. Rest (B) contrasts are associated with steeper learning curves (i.e., higher Δp₉₅ values). R: right; L: left. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 4

(A) Regions of the cerebellum and precentral gyrus showing covariation between FA and motor learning scores (Δp₉₅) in the Sequence task, t > 2, P < 0.05 corrected. Light blue shows the WM skeleton in which statistical analysis was carried out; dark blue indicates regions where higher FA correlates with steeper learning curves (i.e., higher Δp₉₅ values). (B) Regions of the cerebellum and the temporo‐occipital cortex showing covariation between GM density and learning scores (Δp₉₅) in the Sequence task, t > 2, P < 0.05 corrected. In these regions, higher GM density co‐varies with steeper learning curves. R: right; L: left. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 5

Multivariate analysis of behavioral and imaging measures using multidimensional scaling (MDS). Distances between elements of the two dimensional representation of behavioral vs. functional vs. structural measures reflect the overall similarity between their properties. Abbreviations: BStemF: brainstem functional; LAmgF: left amygdala functional; LCbF: left cerebellar functional; LCbGM: left cerebellar grey matter; LCbWM: left cerebellar white matter; LCingF: left cingulate functional; LHippoF: left hippocampus functional; LPallF: left palludum functional; LParacingF: left paracingulate functional; LParOpercF: left parietal opercular functional; LPostCF: left postcentral gyrus functional; LPreCF: left precentral functional; LpSMAF: left preSMA functional; LPutF: left putamen functional; LSFGF: left superior frontal gyrus functional; LThalF: left thalamus functional; RCbF: right cerebellar functional; RCbGM: right cerebellar grey matter; RCbWM: right cerebellar white matter; RCingF: right cingulate functional; RDentF: right dentate functional; RParacingF: right paracingulate functional; RPreCWM: right precentral gyrus white matter; RpSMAF: right preSMA functional; RSFGF: right superior frontal gyrus functional; RVisGM: right visual cortex grey matter. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]