The durations of food retrieval events in 75 trials of each left- and right-hand use in each experimental session were sampled. For normalizing the sample distributions, the Box–Cox transformation with λ = −1 was applied for each session data. The homoscedasticity among the transformed distributions was examined with Bartlett’s test. Differences among the sample averages of different sessions and used hands were compared using analysis of variance (ANOVA) with correction for multiple comparisons using the Tukey–Kramer method.
The fNIRS data in each of the 150 trials (75 right hand, 75 left hand) was time-locked (t = 0) to the onset of food retrieval movement for that trial. Then the values of ΔHbO and ΔHbR for each time point (i.e., a 130 ms period, starting from t = 0 for each trial) were averaged and plotted in Fig. 4 (and in Fig. 5 for additional tasks). For each time point, a paired t-test between the left-hand trials and right-hand trials was also conducted. Figure 6 plots the t-score values for the contrast of left versus right hand movement, for both ΔHbO and ΔHbR, at each time point in each of the eight sessions. When significant amplitude (p < 0.05) in the t-value curves was detected, the maximum amplitude in the ΔHbO and ΔHbR curves and the time elapsed from onset to maximum amplitude was calculated. These calculations at each channel were performed for each experimental session. Maximum amplitude and elapsed time in each session were averaged between channels for each brain hemisphere separately and the Pearson’s product-moment correlation coefficient of them against the time required for food retrieval were calculated.