Single-trial analysis of event-related potentials promises access to the trial-to-trial variability that averaging discards, and many studies report early-window summary measures that covary with later component amplitudes. Such couplings can, however, arise from the temporal autocorrelation of continuous EEG rather than from stimulus-locked processing. We asked whether the conventional family of endpoint-summary measures those that collapse a time window to a single value, including mean amplitude, root-mean-square, variance, signal-complexity measures (permutation entropy, sample entropy, Lempel-Ziv complexity), and Hjorth parameters, captures genuine stimulus-locked information about P300 amplitude in the active visual oddball once autocorrelation is controlled. Analyzing the ERP CORE visual P3 dataset (N = 27; 1,084 trials, 213 target and 871 standard, with experimental condition as a covariate), we related each early-window (0-150 ms) measure to P300 amplitude at Pz and re-estimated every model on pseudotrials placed at random latencies in the same recording; the direction of change under this substitution, not the raw effect size, is the diagnostic. Cross-channel amplitude and energy couplings strengthened under pseudotrial substitution, indicating dependence on background structure. Large same-channel coupling (R{superscript 2} {approx} 0.31) was unchanged under substitution and present at every electrode, including the eye channels, identifying it as general within-trial temporal continuity rather than a P300-specific process. Complexity measures carried near-zero population-level coupling but large, directionally split per-subject slopes. An independent dataset (different laboratory and hardware; same paradigm) reproduced the same-channel continuity result (N = 90 participants; 83 for pseudotrial fits) and the directionally split per-subject pattern across complexity measures. Endpoint-summary measures therefore do not capture consistent population-level P300 coupling once autocorrelation is controlled; the complexity family carries person-specific coupling that cancels at the population level, motivating analytic designs sensitive to individual differences.
Biber, E. et al. · CC-BY 4.0