The reduction of long-latency auditory ERPs amplitude, including P300, to repeated stimuli has been well documented in the literature on habituation. The effect of block repetition on auditory ERPs recorded for clinical purposes, where interblock intervals are commonly longer than those employed in habituation studies, was studied in a sample of 38 adults submitted to two blocks of a counting oddball paradigm. Four different experimental conditions were considered, differing in target probability, delivery or not of a previous passive oddball tone sequence, and the performance or not of other oddball tasks requiring more complex discriminative responses between the two blocks. Results showed that: (1) N1 amplitude to the frequent non-target stimuli decreased in the second block under all the conditions; (2) when the two blocks were consecutive (separated by 2-3 min), P300 amplitudes were unaffected by block repetition, this whatever the probability of the target (25% vs. 10%) and whether or not a passive oddball sequence preceded the two active blocks; (3) P300 amplitude was only affected by stimulus repetition in those subjects who performed more complex cognitive tasks between the first and second blocks and; (4) latency values were unaffected by repetition. It is hypothesised that the N1 amplitude decline may be caused by a decrease in alertness or arousal level produced by stimuli repetition. Reduction in P3 amplitude only appeared when more difficult tasks had to be done between the two oddball blocks and may reflect a decrease in the amount of attentional resources allocated to the second block, due either to fatigue or over training. The practice of using a grand average of several repetitions of the oddball paradigm, as recommended for the clinical use of long-latency ERPs, seems to be adequate provided that long interblock intervals are used and that the subject is not engaged in tasks requiring a high mental workload between the trial blocks.