Shifting attention in a rapid visual search paradigm.

A method is introduced for studying shifts of attention in semantic space, testing 56 subjects in four experiments on a semantic monitoring task based on rapid, serial, visually presented (RSVP) word-sequences. Following a cue to shift attention, accuracy of semantic monitoring drops abruptly to a low level, then gradually recovers to reach preshift levels over successive stimuli in the RSVP sequence. Using this method, we compared two kinds of criterion-shifts, one requiring a set-reversal ('reversal shifts'), the other involving a shift between orthogonally defined categories ('orthogonal shifts'); no differences were found. We have also examined the difference in a shift between two different processing domains (semantic vs typographic) compared with a shift of criterion within the same processing domain. The results showed no differences for within- vs between-domain shifts. Finally, we studied the time-course of a semantic attention shift. Execution of a semantic shift did not follow an internally controlled time-course but was a direct function of the rate of stimulus presentation. No evidence was found for the operation of a 'supervisory attentional system' independent of external stimulus triggering.