Subjects discriminated whether two sequentially presented human face pictures (S1 and S2) were identical while event-related potentials were recorded to explore the neural basis of visual completion for occluded objects. Four kinds of stimulus trials were employed: S2 and S1 were identical (match-complete); S2 and S1 were identical, but S2 was partially occluded (match-incomplete); S2 was a different face from S1 (mismatch-complete); S2 was different from S1, and was partially occluded (mismatchincomplete). Incomplete faces enhanced the amplitude of N1 compared with complete faces. The peak latency of N2 elicited by incomplete faces was delayed about 20 ms compared with complete faces. The enhancement of N1 is related to visual completion processing that requires additional time.