The Costs of Multiple Concurrent Tasks in Scene Perception

Document Type


Publication Date


Digital Object Identifier (DOI)



Motivation. Everyday scene perception often involves attending to a variety of component processes and relations that change over seconds. Most research in scene perception fails to capture this richness because researchers have focused on single tasks with brief time courses. We designed experiments to address the complexity of natural scenes, involving multiple tasks that evolve and overlap across time. This research presents a novel paradigm for assessing the perceptual demands of simultaneously monitoring multiple scene dimensions and allows us to examine an important theoretical contrast: The attentional set hypothesis is that task set is an important determinant of perception, and that the need to repeatedly change task set will interfere with performance. In contrast, if scene perception is efficient and primarily bottom-up, then changes in task set should cause minimal interference.

Methods. We created 4 animated processes (tasks) involving targets and distractors defined by dynamic changes in one property: motion, location, color, or shape. The display screen was divided into 4 quadrants. In the single task conditions, one task filled each quadrant; in the multitask condition, there was a different task filling each quadrant. Each trial consisted of 30 seconds of animations involving a total of 136 distractors and 8 targets. 36 observers were required to monitor all four quadrants for targets.

Results and discussion. Performance is reasonably summarized by overall accuracy in each condition. Accuracy averaged 79.4% across the four single task conditions. In the multitask condition, overall accuracy was 5.3% lower. The reduction in performance was statistically reliable (p [[lt]].01); however, the ability to achieve multitask levels near single task levels is impressive given the requirement to simultaneously perform four distinct tasks. Impressive scene perception abilities found in previous research with single task paradigms appear to extend multitask scene perception with only small decrements.

Was this content written or created while at USF?


Citation / Publisher Attribution

Journal of Vision, v. 9, issue 8, art. 962