Abstract
When two target letters (T1 and T2) are embedded among distractors and the items presented sequentially at 100 ms each in a single-stream RSVP task, identification of the second target (T2) is poor if it lags the first (T1) by 200–500 ms. This attentional blink (AB) effect has been attributed to the processing demands of the T1. We present data from experiments that speak against this account. The main manipulation is an 8×8 noise matrix overlaying the RSVP letters. In Experiment 1, we contrasted (a) the standard single-letter stream against (b) a constant noise matrix, and (c) a noise matrix that changed every frame. AB was virtually eliminated in (c). Indeed, performance was better in the constant-matrix condition than the standard condition. In Experiment 2, we established that the critical frame was the T1-trailing (T1+1) frame. T2 performance was better when the noise matrix of the T1+1 frame was made distinct from the preceding noise matrices. In Experiment 3, we contrasted inserting a blank in the T1+1 frame with overlaying the T1+1 distractor with a noise matrix, and showed that T2 performance was better in the latter condition. We argue that the AB is caused when attention cannot disengage rapidly from the T1 stimulus. Visual cues that inform the visual system that the T1 episode is over facilitates attentional disengagement and thus modulates the blink.
supported by NUS-FASS Grant C-107-000-222-091