Presenting segmented images in a rapid serial visual presentation stream improves search accuracy

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Visual search performance is a critical factor in many high-stakes duties, warranting the need for strategies to enhance target detection accuracy. Research using rapid serial visual presentation (RSVP) of stimuli shows that observers can detect categorically defined, pre-specified targets even when the presentation rate is rapid, suggesting RSVP as a viable strategy. To investigate how and how well RSVP can improve target detection in complex search arrays, five experiments were conducted to compare search performance between Full-Image search conditions and various RSVP-based conditions. Stimulus presentation time/total search time was the same across conditions. Experiment 1 demonstrated the utility of RSVP to enhance target identification in simple arrays (i.e., Landolt Cs). Experiment 2 involved more complex scenes and target-present/-absent judgments. Results showed that RSVP increased target detections due to both a liberal change in criterion and an increase in sensitivity. Experiment 3 provides some evidence against the reduction in peripheral clutter as the primary contributor to RSVP performance increases. Experiments 4 and 5 prompted and limited eye movements, respectively, to distinguish the role of eye movements in RSVP-based search. These two latter experiments imply that lower target detection performance under time constraints in whole image search conditions is attributable to time-wasting, irrelevant and inefficient eye movements. These experiments suggest that RSVP advantage occurs because the method maximizes time for inspecting and processing each search image/segment. Real-world visual search tasks may benefit from segmenting the search display and presenting images in an RSVP stream.
Funding Information
  • Office of Naval Research

This publication has 40 references indexed in Scilit: