A rapid serial visual presentation technique was used to display sequentially two targets, T1 and T2, and monitor P3 amplitude and latency variations associated with the attentional blink (AB) effect. A red T1 digit was embedded on each trial in a sequence of black letters. T2 was either masked by a trailing stimulus or not masked. T1 had to be identified on a proportion of trials, or ignored in other trials. T2 was the black letter 'E' on 20% of the trials, or any other non-'E' black letter in the other 80% of the trials. A delayed 'E' detection task was required at the end of each trial. An AB was observed when T1 had to be reported and T2 was masked. The AB effect was associated with a sizable amplitude reduction of the P3 component time locked to T2 onset. When T2 was not masked, no AB or P3 amplitude variations were observed. When T1 had to be reported, a delayed P3 peak latency was observed at short compared to long T1-T2 intervals. No effect of T1-T2 interval was observed on the T2-locked P3 peak latency when T1 could be ignored. Taken together these findings provide converging evidence in support of temporal attention models bridging behavior and electrophysiology that postulate a direct link between the cause of the AB effect and the sources of both amplitude and latency variations in the T2-locked P3 component.