Figure 2.
Robust cortical response to silence during music listening. A, Experiment 1 setup. The EEG signal was recorded as participants listened to monophonic piano music. Univariate vectors were defined that mark with value 1 the onset of either notes (NT) or silent-events (SIL). A system identification procedure based on lagged linear regression was performed between each vector and the neural signal that minimizes the EEG prediction error. B, The regression weights represent the TRF describing the coupling of the EEG signal TRFNT and silent-events TRFSIL. TRFs at the representative channel FCz are shown (top), revealing significant differences (FDR corrected Wilcoxon test, *q < 0.001) between the neural signature of note and silent-event because of inverted polarities, as clarified by the topographies of the TRF components (bottom). C, D, Overall distribution of time intervals between notes and between silent-events and the immediately preceding note. The y-axis indicates the number of occurrences for a given bin of time intervals when considering all trials. The data show that a large number of silent-events occurred <200 ms after a note, implying that in experiment 1, TRFSIL could have potentially been affected by the late response to the previous note. E, The analysis from B was rerun by using multivariate TRF models, i.e., considering note and silent-event vectors simultaneously with multivariate lagged regression to account for possible interaction between the two. The figure shows the regression weights corresponding to the two regressors at the selected channel FCz, and the topographies show the regression weights. As for the univariate TRF result, significant differences were found between note and silent-event TRFs (FDR corrected Wilcoxon test, *q < 0.001). TRFNT showed qualitatively more pronounced early TRF components.