Separating stimulus-locked and unlocked components of the auditory event-related potential.

A new method is evaluated for separating stimulus-locked and unlocked components of auditory event-related EEG activity. The new method uses a regression based subtraction procedure as a way to account for latency and amplitude variability within individual trials. It was applied using the oddball paradigm under conditions of active and passive listening and analyzed as spectral correlations (normalized differences) between post-stimulus epochs of standard, target or deviant stimulus conditions and silent pre-stimulus baseline activity. The regression-subtraction procedure accounted for a greater amount of variance than a method that uses linear subtraction alone. The major component of the response to auditory stimulation was an event-related synchronization in the delta and theta (2-4 Hz) frequency range. Event-related desynchronizations were also observed in the 10 Hz (alpha/mu) and in the 20-30 Hz (beta) frequency range. The regression based subtraction procedure provides better separation of stimulus-locked and unlocked components of event-related EEG activity then linear subtraction alone. Stimulus-locked and unlocked components show different patterns and topographies of effects related to attention and active discrimination. Studying both stimulus-locked and unlocked components of event-related EEG reactivity in the frequency domain provides a more comprehensive account of dynamic brain activity subserving auditory information processing.