A new method of measuring cerebral activation during perceptual and cognitive tasks is presented using the partial coherence. ERP averaging allows one to capture the underlying signal embedded in noise. The fidelity of signal averaging depends on the cumulative history of latency and amplitude fluctuations gathered from trial-to-trial. Signals that exhibit randomly modulated periodicities are said to show partial coherence. A partially coherent time series is one that is measurably synchronized or phase locked with a periodic signal representation such as the Fourier components. Partial coherence is measured by examining the fluctuations or variance in these components. We use partial coherence to provide a regional assessment of ERP strength recorded from a patient's subdural electrode grid during auditory stimulation. We show that partial coherence varies as a function of location along the grid and exhibits exceedingly large values in areas both on and near the superior temporal gyrus.