Twelve-picture RSVP sequences support feed-forward models of detection at 75 Hz

Abstract

Potter, Wyble, and McCourt previously reported that viewers show above-chance detection of a picture in a rapid serial visual presentation (RSVP) when they are given a verbal title (e.g., cut-up fruit) just before or immediately after the sequence, even at a duration as short as 13 ms/picture (75 Hz). We took these results as support for a feed-forward model of detection. However, given that there were only six pictures in each sequence, was detection only possible because of the minimal input load or incomplete masking? Here we report the results for twelve-picture sequences. We presented color photographs of a wide variety of scenes at rates of 80, 53, 27, and 13 ms/picture, with no ISI. The target was never the first or the last picture. In one group the target name was presented before the sequence and in the other group, immediately after the sequence. All pictures were new to the participants and none were repeated in the experiment. Comparing hit and false alarm rates, performance in both groups was significantly above chance at all durations. Whether or not the participant reported detection of the target, each target trial was followed by a forced choice test between two pictures, both of which matched the target title but only one of which had been in the sequence. Given correct detection, forced choice was above chance. The finding that both pre- and post-detection are robust with 12-picture sequences gives further support for feed-forward models of high-level perception. Meeting abstract presented at VSS 2013