UC Riverside

P300-based brain–computer interface (BCI) systems have been widely investigated for assistive communication and neurotechnology applications. However, most existing studies primarily emphasize classification performance within individual stimulation paradigms, providing limited understanding of whether electrophysiological response characteristics remain consistent across heterogeneous experimental conditions. This study investigates the cross-paradigm organization of P300 responses using the publicly available bigP3BCI dataset through an integrated analysis of temporal ERP morphology, spectral characteristics, time–frequency dynamics, statistical variability, session-level reproducibility, and low-dimensional feature representations. Rather than proposing a new decoding algorithm, the study examines whether common electrophysiological response characteristics remain observable across the Row–Column (RC), Checkerboard (CB), and Random (RD) paradigms despite differences in stimulus presentation and EEG morphology. Experimental results indicate that the CB paradigm produced the largest target-related ERP amplitude ( $3.66~\mu $ V) and the earliest dominant positive response (192 ms), whereas the RD paradigm achieved the highest signal-to-noise ratio (19.0 dB). Session-level analyses demonstrated consistent waveform morphology across repeated recordings, with mean correlation coefficients exceeding $r_{\mathrm {sess}}=0.90$ for all paradigms. Continuous wavelet transform analysis revealed temporally localized theta-band activity primarily within the 4–8 Hz range around the dominant ERP interval. Exploratory PCA-based feature space analysis showed partial overlap among paradigm-specific representations (silhouette score = 0.144; mean cosine similarity = 0.74), suggesting that common electrophysiological response characteristics coexist with paradigm-dependent variability. Overall, the findings provide a multidomain characterization of heterogeneous P300 responses and support the use of physiologically interpretable representation analysis for investigating cross-paradigm ERP characteristics under realistic EEG acquisition conditions.