A strong subjective tendency exists for simultaneous sound frequencies forming an harmonic series (integer multiples of the fundamental) to "group" together into a unified auditory percept whose pitch is similar to that of the fundamental. The aim of the study was to determine whether cortical auditory evoked potentials (AEPs) to complex tones differ according to whether the component frequencies of the stimuli are harmonically related or not. AEPs were recorded to continuous complex tones comprising four or more sinusoids. The vertex-maximal "change-potentials" (CP1, CN1, CP2), recorded to a stimulus cycle comprising one harmonic and five inharmonic complexes changing every second, showed no sensitivity to harmonicity, although an additional mismatch negativity was possibly present to the harmonic complex. In a second study the CP2 was significantly attenuated when an harmonic complex changed to a new one in the presence of an unchanging sinusoidal background tone, harmonically related to the first complex but not to the second, and thus becoming perceptually distinct. This, however, might be caused by lateral inhibitory effects not related to harmonicity. In a third experiment, when four concurrent sinusoidal tones came to rest on steady frequencies after a 5-s period of 16/s pseudo-random frequency changes, fronto-centrally maximal "mismatch-potentials" (MN1, MP2), were recorded. Both the MN1 and the MP2 were significantly shorter in latency when the steady frequencies formed an harmonic complex. Since the harmonic complex had a short overall periodicity, equal to that of the fundamental, while that of the inharmonic complex was much longer, the effect might be explained if the latencies of the mismatch-potential are related to periodicity. The perceptual grouping of harmonically related frequencies appears not to be a function of spectral domain analysis, reflected in the change-potentials, but of periodicity analysis, reflected in the mismatch-potentials