Time-resolved associations among theta phase synchrony, power, and ERP amplitude in speech-sound change detection

Speech perception requires the rapid integration of acoustic input with emerging categorical representations, yet the relationship between time-locked evoked responses and underlying oscillatory dynamics remains unclear. In particular, it is unknown whether oscillatory measures reflect the same processes indexed by canonical event-related potentials (ERPs) or capture distinct aspects of neural activity. We address this question using an auditory oddball paradigm contrasting /ɛ/ and /ɪ/ in 45 participants, applying Bayesian generalized additive mixed models to time-resolved EEG data. Windowed ERP models confirmed canonical mismatch negativity (MMN; 100–250 ms) and a later positive response in the P300 range (250–500 ms). We then modeled ERP amplitude continuously as a function of three theta-band measures: inter-trial coherence (ITC), evoked theta power (phase-aligned), and induced theta power residualized with respect to evoked power (non-phase-aligned).Time-resolved analyses revealed partially distinct patterns of association. ITC showed a temporally structured relationship with ERP amplitude, with a broadly positive association spanning the MMN–P300 interval and attenuating thereafter. Evoked theta power exhibited weaker and less consistent effects. In contrast, residualized induced theta power showed a comparatively stable positive association across the epoch, although this effect was modest in magnitude and uncertain.These findings suggest that ERP amplitude reflects the combined influence of temporally structured phase alignment and more sustained oscillatory dynamics, rather than a single underlying process. More broadly, integrating ERP and time-frequency analyses—and distinguishing phase-aligned from non-phase-aligned dynamics—provides a more nuanced characterization of the relationship between neural mechanisms and auditory deviance detection.