Analyses before training

We examined task-related activation/deactivation during the illusion using a second-level group analysis⁶⁸ with a one-sample t-test to illustrate the patterns of task-related activation/deactivation across the entire brain in each age group (Fig. 3a). We also examined between-age-group differences by comparing the older group with the younger group (older vs. younger). We used a family-wise error rate [FWE]-corrected extent threshold of P < 0.05 across the entire brain for a voxel-cluster image generated at the uncorrected height threshold of P < 0.005, which was consistently used in the present whole-brain analyses. For the anatomical definition of the identified peaks, we referred to the cytoarchitectonic map implemented in the SPM Anatomy toolbox⁶⁹, which was also consistently used in the present study. To confirm the between-age-group differences in the ipsilateral (right) M1 region, we set an ipsilateral M1 ROI as a 4-mm radius sphere around the M1 peak (36, − 26, 66), in which the degree of deactivation is correlated with a better right hand/finger dexterity during childhood⁶. We extracted the parameter estimate of brain activity from the M1 ROI for each participant and plotted it against their individual ages (Fig. 3b).

We performed a regression analysis to test whether the degree of reduction/loss of ipsilateral M1 deactivation is associated with dexterity deterioration in the older group (n = 48). In this analysis, we hypothesized that a region with such a correlation could be observed in the M1 ROI, in which the degree of deactivation is correlated with a better right hand/finger dexterity during childhood⁶. Based on this strong anatomical hypothesis, we applied an SVC (P < 0.05)⁷⁰ with the M1 ROI (above). Finally, to illustrate the relationship between the individual degree of ipsilateral M1 activity and the individual peg time in the older group, we extracted the data from the M1 ROI and plotted them against the peg time (Fig. 4). We fitted a regression line to the data obtained from each group (BM and RH). We used an analysis of covariance for the statistical evaluation of the slopes of the regression lines and their intercepts.