Previous research has shown that negative stimuli elicit more attention than do positive stimuli. However, this research has relied on response-based measures to assess attention. The current research uses the P1 component of the event-related brain potential (ERP) as a proximal index of attention allocation to valenced stimuli. In two studies, P1 amplitude was measured while participants evaluated positive and negative pictures. In both studies, principal components analysis showed that P1 amplitudes to frequent stimuli and to rare negative stimuli were larger than P1 amplitudes to rare positive stimuli. This is (a) evidence for the extremely rapid (<120 ms) differentiation of positive and negative stimuli and (b) process-based evidence for a negativity bias in attention allocation.