The major intent of this study was to compare the role of stimulus repetition and change in the elicitation of the MMN, an ERP component specific to stimulus change, and N2b, usually partially overlapping the MMN when stimuli are attended. Event-related potentials were recorded in one set of conditions where subjects ignored the stimuli and read a book, and in another set of conditions where subjects counted stimuli designated as targets. Stimuli were delivered in 4 ways, the common feature between all these conditions being the occurrence of infrequent events at a probability of 0.20: (1) an oddball paradigm with 1 deviant, (2) an oddball paradigm with 2 deviants, each with a probability of 0.10, (3) a regular alternation of tones of 2 pitches where either of the 2 tones infrequently repeated (P = 0.20), and (4) a random presentation of tones of 5 different pitches, where any of the 5 tones infrequently repeated (P = 0.20). In the count conditions, the infrequent events were designated as targets. It was found that the MMN was elicited by stimulus change and not stimulus repetition in the ignore and count conditions, whereas the N2b was elicited by both stimulus changes and repetitions in the count conditions. It was also possible, in the count conditions, to disentangle the part of the late positive complex which is related to stimulus deviation and the part which is related to stimulus significance (target).