2.4 Data Analysis
Based on previous literature concerning adaptation and MMN (e.g., Bühler et al., 2017; Jaffe-Dax et al., 2017; Jost et al., 2015), a set of 25 frontocentral electrodes were pooled for analyses: E19, E11, E4, E20, E12, E5, E118, E13, E6, E112, E10, E16, E18, E30, E7, E106, E105, E37, E31, E129, E80, E87, E55, E36, E104 (see the top-left corner of Figures 3 or 4a for the electrode positions).
The time window for detecting the local peak amplitude of N1 was determined using the global field power (GFP) and global dissimilarity peaks from subsequent maps of the grand average data. GFP represents the root mean square (RMS) across the average-referenced electrode values, or the standard deviation of all electrodes at a given time point (Murray et al., 2008), while global dissimilarity quantifies configuration differences between maps, irrespective of their strength (Lehmann & Skrandies, 1980; Murray et al., 2008). The GFP and global dissimilarity were generated from Brain Vision Analyzer and Cartool (Brunet et al., 2011).
The N1 time windows for deviants, 2nd tones, and final tones were 70–218 ms, 76–118 ms, and 76–140 ms, respectively. In contrast, as there was no distinct GFP peak in the P2, its time windows were determined by peaks between N1 and N2. The P2 time windows of the corresponding conditions were 120–348 ms, 96–240 ms, and 98–254 ms, respectively (see Figure S1 in Supplementary Material II for illustrations of how these P2 time windows were defined).
In addition, the difference wave corresponding to the MMN component was computed by subtracting the amplitudes of the final tones in the 4th to 30th positions from those of the deviants in the same frontocentral electrodes (deviants–final tones), with the time window defined based on the GFP and the global dissimilarity peaks as 76–196 ms. The local peak MMN amplitude was automatically identified using Brain Vision Analyzer.
Similarly, for the adaptation effects, the local peak amplitude in each tone position and time window was automatically determined using Brain Vision Analyzer. We defined the initial adaptation as the peak amplitude decrease from the deviants to the tones in the 2
nd position
in the frontocentral electrodes (deviants–2
nd tones). On the other hand, the subsequent adaptation was defined as the subtraction of the peak amplitudes of the final tones in the 4
th to 30
th positions from those of the tones in the 2
nd position in the same electrodes (2
nd tones–final tones). Figure 1 depicts the calculation methods of initial adaptation, subsequent adaptation, and MMN.
The adaptation curves for the first ten tones of N1 and P2 were first plotted based on the peak amplitudes in the corresponding time window. Given that the trial number might not be sufficient (< 50 trials on average) after the 10th tones to generate reliable results, we only focused on the first ten tones for the adaptation curves. Paired sample t-tests were conducted between each consecutive tone pair, with Holm-Bonferroni corrections (Holm, 1979) applied to control for multiple comparisons. Additionally, repeated-measures analysis of variance (ANOVA) was performed on the amplitude employing an a priori linear trend analysis across the first ten tones in each stimulus sequence to elucidate the adaptation pattern observed in N1 and P2. To ascertain the presence of MMN, a one-sample t-test against zero was conducted for the amplitude of the difference wave.
Furthermore, the relationship between initial adaptation, subsequent adaptation effects, and MMN was examined through correlation and backward stepwise linear regression analyses. In the regression model, the independent variables (IVs) comprised N1 and P2 initial adaptations and subsequent adaptations, while MMN peak amplitude served as the dependent variable (DV). The elimination criterion for the backward regression was set at P > .10, meaning that in each step, the IV with the lowest partial correlation with the DV, meeting the criterion P > .10, was removed until all variables remaining in the model were P < .10. All statistical tests were performed using IBM SPSS Statistics Version 25 (IBM, Armonk, NY, USA).