Brain potentials were recorded from 15 healthy young subjects during the performance of a word recognition task. During the study phase, subjects had to intentionally memorise a series of words. These words were presented again together with the same number of new words in a following test phase where the instruction was to discriminate between repeated words and new words. We compared event-related potentials (ERPs) evoked by correctly identified repeated words (hits) and ERPs evoked by incorrectly classified new words (false alarms). Although both types of words were thought to be repeated the ERPs indicated differences between these two conditions starting at about 450 ms after the stimulus onset. These differences were mostly pronounced over frontal scalp locations but occurred also over parietal scalp locations (false alarms produced significantly more negative going ERPs than hits). We interpret that frontal and parietal brain areas show greater activation during false recognition because of a more intensive search for item representations.