Figure 4. The non-zero mean property of resting-state neural oscillations is a prerequisite for the baseline-shift mechanism.

The late component of the ERP is generated by a baseline shift occurring during event-related desynchronization (ERD) of non-zero mean oscillations. (A) The amplitude fluctuation asymmetry index ($AFAI$) is quantified as the normalized difference between the variance of the peaks and troughs of an oscillatory signal during resting state. ${\displaystyle AFAI<0}$ (blue) indicates a stronger modulation of the troughs relative to the peaks, consistent with a negative mean. The topography represents the group-level t-statistics of $AFAI$: a significant negative cluster was found at frequencies between 5 and 30 Hz and with an occipito-parietal peak. The baseline shift index ($BSI$) is quantified as the correlation between the oscillatory envelope at a certain frequency and low-pass filtered EEG signal at 3 Hz during resting-state. ${\displaystyle BSI<0}$ (blue) indicate a negative relationship between oscillatory envelope and low-pass filtered signal, consistent with negative mean. The topography represents the group-level t-statistics of $BSI$: a significant negative cluster was found at frequencies between 5 and 21 Hz and with an occipito-parietal peak. Black dots represent electrodes comprising the significant clusters. (B) Comparison between group-level $AFAI$ (black) and $BSI$ (blue), averaged across respective cluster electrodes, and shown for each frequency. Shaded areas indicate ± SEM. These results indicate the presence of a negative oscillatory mean, consistent with the baseline-shift account.