The baseline-shift mechanism (BSM) of evoked response (ER) generation. For a particular ER, probing the agreement with BSM would involve extracting both the ER and the oscillatory amplitude envelope. A. The single-trial broadband signal. B. The amplitude envelope of oscillations is extracted from a broadband signal of each trial. C. To get a high signal-to-noise ER, usually a few trials are acquired. Note that since oscillations have a negative mean, their attenuation would lead to the generation of an ER with a positive polarity (shown in E.). D. Similarly, for each trial, the amplitude envelope is extracted. E. Trials are averaged and, optionally, low-pass filtered to obtain an ER. F. Amplitude envelopes over trials are also averaged to obtain an estimate of the change in oscillatory amplitude in the poststimulus window. Here, we simulated the example of negative-mean oscillations giving rise to a positive-polarity ER.

Temporal similarity between P300 and the alpha amplitude envelope. A. Left panel—time course of P300 at the Pz electrode elicited by the target stimulus, and ER after a standard stimulus (sER) both averaged across the participants. Right panel—alpha amplitude envelope at Pz electrode averaged across participants for target and standard stimulus. The time courses of P300 and alpha amplitude display similarities in initial slope and peak latency. Alpha amplitude modulation is pronounced only for the target but not for the standard stimulus. Shaded areas display the standard error of the mean. B. A correlation between P300 and alpha amplitude. For grand averages at each electrode, the correlation between P300 and alpha envelope was computed with the Pearson correlation coefficient. Electrodes marked with “x” had significant correlation coefficients. The p-value was set at the Bonferroni corrected value of −4.

The difference in the strength of alpha amplitude modulation correlates with the difference in P300 amplitude. A. Alpha amplitude envelope sorted into 5 bins according to the depth of modulation in the poststimulus window. The bins were the following: (66, –25), (–25, –37), (–37, –47), (–47, –58), (–58,– 89) % change. Here, –100% corresponds to the deepest modulation, and 0% to the absence of a change in the amplitude. B. P300 responses are sorted into the corresponding bins. C. The spatio-temporal t-test reveals clusters of significant differences between the two most extreme bins—bin 1 and bin 5. The topography of t-statistics is sampled at 500 ms. The significant electrodes at this time point are marked with “x”.

Spatial similarity of topographies of P300 (A) and alpha amplitude (B) contrasted between the target and standard stimulus. The topographies are shown at the peak amplitude of P300, which was estimated from the averaged over trials ER for each participant within the time window of 200-1000 ms poststimulus at the Pz electrode (on average 509±171 ms). For ER, the contrast was built by subtracting the sER amplitude from the P300 amplitude. For alpha amplitude, the contrast was built by dividing values of the amplitude after the target stimulus onto values after the standard stimulus.

Spatial similarity between P300 and alpha amplitude in a source space. A. The difference between P300 and sER, after correction for multiple comparisons. The difference was estimated as the subtraction of averaged sER power from averaged P300 power in the time window of 300–700 ms. The colorbar thus indicates the difference in power. The black line outlines an overlap that is common for both P300 (top 10% of activity) and alpha amplitude (top 10% of activity). B. The difference in alpha amplitude envelope after standard and target stimuli with a correction for multiple comparisons (all dipole locations are significant). The difference was estimated as the target poststimulus alpha amplitude divided by the standard alpha amplitude. The poststimulus window was the same as for P300: 300–700 ms.

The baseline-shift index (BSI) explains the direction of ER based on the direction of alpha amplitude change. A. The average values of BSI at each electrode estimated from the resting-state data. Here, BSI is computed as the Pearson correlation coefficient (see Methods/The baseline-shift index). A positive deflection of P300 at posterior sites coincides with a decrease in alpha amplitude, a case that corresponds to negative mean oscillations. The sign of the mean can be estimated with the sign of BSI (Nikulin et al., 2010). B. BSIs at Pz were binned into 5 bins. The BSI bins were the following: (−0.99, −0.81), (−0.81, −0.46), (−0.46, 0.09), (0.09, 0.62), (0.62, 0.98). According to predictions of BSM, if BSI (and the oscillatory mean) was negative, then the attenuation of oscillations would lead to the upward direction of ER. C. P300 was binned into bins according to BSI. For bins with negative BSI, the amplitude of P300 is higher in comparison to bins with positive BSI. D. The evolution of the statistical difference between the amplitude of P300 in the first and fifth BSI-bins across time and space. The difference is prominent over the central and parietal regions. The cluster-based permutation test revealed significant clusters in central and parietal regions with a p-value 10−4. E. The topography of t-statistics is sampled at 500 ms (at the dashed line of the upper panel). The significant electrodes at this time point are marked with “x”.

P300 and alpha oscillations showed similar correlation profiles across cognitive processes. A. Attention, memory, and executive function scores correlate with P300 and the alpha envelope. Attention scores were computed from TMT-A time-to-complete and Stroop-neutral time-to-complete. Memory scores were computed from the CERAD word list (combined delayed recall, recognition, and figure delayed recall). Executive function scores were computed from TMT-B time-to-complete and Stroop-incongruent time-to-complete. P300 amplitude and latency were evaluated after spatial filtering with LDA. Alpha envelope amplitude and latency were evaluated after spatial filtering with CSP (see Methods/Spatial filtering). Correlation values are computed as a partial correlation with the Spearman correlation coefficient having age as a covariate variable. Sample size for this analysis is 1549. * p-value<0.05, ** p-value<0.001, *** p-value<0.0001. B. A spatial pattern corresponding to the LDA spatial filter that was applied to obtain high signal-to-noise P300. C. A spatial pattern corresponding to the CSP filter that was applied to obtain alpha oscillations.

The baseline-shift mechanism (BSM) summary. Two important prerequisites of the BSM—non-zero mean r and amplitude modulation A(t)—should occur together so the ER would be generated. A. Non-zero mean oscillations when modulated in amplitude generate an ER. B. If oscillations have a zero mean, then no ER is generated. C. If oscillations have a non-zero mean but do not systematically (trial-by-trial) experience modulation, then no ER is generated.

Overview of the previous findings concerning stimulus-related alpha amplitude decrease in an oddball paradigm and similar experiments. The search for this short review was completed via Google Scholar on 21-09-2021 using keywords: “evoked response”, “erp”, “erf”, “erd”, “ers”, and on 03-10-2021 using keywords: “P300”, “erd”, “ers”, “eeg”, “meg”. We picked only research or review papers written in the English language.

Time-space evolution of P300 and alpha amplitude envelope. A. P300 grand average over participants. B. Alpha amplitude grand average over participants.

The synchronisation in the population of neurons generating alpha rhythm affects the amplitude of the alpha rhythm but not the evoked response. Similarly as in Studenova et al. (2022), we performed simulations with evoked response and alpha rhythm resembling P300 and alpha amplitude envelope obtained from real recordings. The synchronisation strength was manipulated by the starting distribution of phases. The kappa parameter characterises the peakiness of phase distribution—with kappa = 5 the distribution has a peak and with kappa = 0.01 the distribution is close to uniform. In the real EEG data, the synchronisation of the underlying population is unknown. Therefore, a direct correspondence between the amplitude of the evoked response and the amplitude envelope of the alpha rhythm can be only approximately evaluated (for instance, with the baseline-shift index, BSI).