We recorded topographic mapping of auditory evoked potentials (AEPs) to homogeneous series of tones delivered in the absence of any cognitive task ("neutral condition") and compared them with responses to identical stimuli when they acted as non-targets in an oddball paradigm. With respect to the neutral condition, non-target responses showed an increase in N100 amplitude which was found to depend on two different, but partially overlapping effects. The "early effect" had similar latency and topography as exogenous N100 and may reflect changes in the overall state of alertness since its amplitude declined at the end of the test. The "late effect" affected the N100s descending slope and had features consistent with those of a processing negativity. It was frontally distributed, and topographic mapping revealed a right-sided predominance. Finally, non-target responses exhibited a central positivity (P250) which did not appear in neutral AEPs. P250 shares several features, including latency, amplitude and scalp topography, with the endogenous ERP component commonly labelled "P3a". Non-target P250 could be the result of an attentional shift towards the stimuli, and reflect some aspects of the classification process. We suggest that non-target responses in simple oddball paradigms should be routinely studied along with target responses in order to improve the diagnostic capabilities of cognitive ERPs. Notably, non-target responses may help to decide whether an abnormal target-P300 is related or not to a deficit in the mobilization of attentional resources.