To elucidate the brain mechanisms to encode sequential events, event-related potentials (ERPs) were recorded during a sound-sequence discrimination task using young and middle-aged adult subjects. In the task, a series of six or 12 kinds of natural sounds were sequentially presented; 70-80% of the stimuli were presented in a fixed order (Non-target), but the remaining stimuli, in a random order (Target). The subjects were instructed to detect the Targets and press a button at the end of each Target. In a control task, the same sounds were randomly presented (Control), and they were instructed to press the button at the end of each sound. Behavioral results indicated that the young subjects learned the task faster than did the middle-aged subjects. Positive ERP waves were evoked by Targets and Non-targets in the parieto-occipital area around 300-700 ms after stimulus onset. The mean amplitudes during this period in the young subjects were larger in Target than Control conditions, and those in Target condition were larger in the young than middle-aged subjects. Furthermore, the mean amplitudes in the Target condition were significantly correlated with behavioral performance. Equivalent dipoles for the ERPs evoked by Targets were estimated in the medial temporal lobe including the hippocampal formation and parahippocampal gyrus. The results suggest that the ERPs around 300-700 ms latency are involved in sound-sequence information processing. Furthermore, decrease in amplitudes of this positivity in the middle-aged subjects suggests that age-related memory decline is associated with deficits in encoding and retrieval of unfamiliar sequence.