In a series of EP experiments, auditory temporal generalization in the range of milliseconds was compared to pitch generalization, with difficulty held constant across tasks. Participants had to identify a previously memorized standard stimulus among deviant stimuli, either with respect to pitch or duration. Identical sets of stimuli were presented in both tasks, using sine tones that varied independently along both the duration, and the pitch dimension. Two ranges of durations were investigated, 200 msec and 400 msec. For the 200-msec range, in three experiments inconsistent EP correlates of temporal vs. pitch generalization were found, whereas for the 400-msec range, there were no significant pitch - duration EP differences at all. However, within-task EP comparisons between stimuli of different duration yielded highly consistent results across the three experiments. Only when duration was attended to, variations in stimulus duration were accompanied by amplitude modulations of both a posterior P3b, and a fronto-central P500, but not during passive listening or while attending to pitch. P3b/P500 amplitude modulations, therefore, represent EP correlates of actively processed duration information. Furthermore, the nature of these EP correlates suggests real-time comparison of the current stimulus against a perceptual representation of the standard duration (200 msec). For the 400-msec range of durations, EP correlates of actively processed stimulus duration were qualitatively different. Stimulus duration was positively correlated with resolution time of a broad negative slow wave. Temporal processing in this longer time range appears to involve intense memory processes. It is concluded that durations below 300 msec (or, within the P300 span) are subject to time perception, whereas longer durations require memory-based time estimation.