Moussa Kousa¹ (), Brad Wyble¹; ¹The Pennsylvania State University

A picture flashed for only a few milliseconds can still be clearly perceived if nothing follows it. However, in a rapid serial visual presentation (RSVP) of photographs, each picture masks the one before it, yet even under such rapid masking, people can still detect a briefly presented target scene when given only a conceptual cue, such as “picnic” (Potter 1976). To explore the limits of this selection, we tested whether searching for “relatedness” allows participants to detect a pair of semantically related objects (e.g., a golf club and a golf ball; depicted as line drawings) embedded within an RSVP of unrelated pairs (e.g., a shoe and a plant), as assessed by a memory probe. We also tested the reverse: detecting an unrelated pair within a stream of related pairs. These tasks were assessed with four RSVP SOAs: 800, 400, 200, and 106 ms. At longer SOAs, performance in both conditions was near ceiling. As SOA decreased, accuracy declined drastically for both conditions, although detecting related pairs was slightly easier than detecting unrelated pairs. This poor performance at short SOAs contrasts with excellent performance for using conceptual cues with natural images at SOAs of 100ms in prior work, suggesting that “relatedness” is not an effective cue for selecting targets in the same way that a broad scene label is. To verify that subjects were able to connect our stimuli to conceptual labels even during high-speed RSVP, we conducted a variant of the experiment in which participants were given a conceptual label (e.g., chopping for axe–tree) instead of searching for related or unrelated pairs. At SOAs of 400, 200, and 106 ms, performance at the lower SOAs was much higher than in the first experiment. These results suggest a specific inability to select based on relatedness as a general property.