Auditory evoked potentials (AEPs) were used to examine selective stimulus processing in sleep. In waking, repetitive stimuli generate exogenous P1, N1 and P2 components of the auditory evoked potential (AEP). Deviant stimuli generate endogenous cognitive components including the mismatch negativity (MMN), N2 and P3 components. We examined long-latency auditory evoked potentials elicited by repetitive and deviant stimuli during waking and stage II-IV sleep to assess whether stimulus deviance is detected during sleep. The waking P1, N1b and P2 had maximal amplitudes at fronto-central scalp sites, with additional peaks (N1a, N1c) at temporal sites. Deviant tones generated a frontal maximal MMN, and complex novel tones generated an additional P3 component maximal at centro-parietal sites. During stages II-IV sleep N1a, b, c amplitudes were reduced. During stage II sleep all stimuli generated increased P2 amplitudes and a late negative component (N340). Deviant stimuli generated greater P2 and N340 amplitudes than frequent stimuli in stage II sleep, as well as an additional P420 component. In stage III-IV sleep the P420 was absent and the AEP was dominated by a negativity of long duration whose amplitude increased in response to deviant stimuli. These data indicate that auditory evoked activity changes from wakefulness to sleep. The differential response to deviant sounds observed during waking and all sleep stages supports the theory that selective processing of auditory stimuli persists during sleep.