Two experiments are described in which the P300 component of the event-related potential was recorded during a modification of the Deese-Roediger-McDermott false-memory paradigm. P300 amplitudes and topographies were evaluated in both true recognition of previously presented (studied) words and in false recognition of associatively related, never presented (critical lure) words. P300 topography and amplitude did not appear to differ between true and false recognition. However, false recognition of critical lures produced substantially shorter P300 latencies than did the true recognition of studied words.