Abstract
In a typical multiple-object tracking experiment, observers are shown moving target objects amidst similar distractors and are asked to identify which of the objects were targets at the end of each trial; these studies provide a snapshot of attentional performance at some fixed point in time. In our study, we instead had observers (4 naïve, 2 authors) track the perceived centroid of a number of moving targets in real time amidst visually identical distractors over the course of long (20 s) trials. This resulted in a continuous assay of attention vs. time within individual trials. All of the objects were bright squares (11x11 arcmin) against a middle gray background, and the "targets" temporarily turned purple and then returned to white before each trial began. Each object did an independent random walk with the constraint that objects were not allowed to collide. We varied both the number of targets and distractors across conditions. Our results show that the absolute Euclidean error of centroid estimates (the straight-line distance between the perceived and actual centroids) increased linearly with time; in other words, the longer observers had to track, the further afield their estimates became. Moreover, the "cost of time" (the slope of the error-vs.-time data) itself depended linearly on both the number of targets and the number of distractors. Our results support the notion that attention in multiple-object tracking tasks is a fluid resource that can be distributed amongst various numbers of objects, but further suggest that this resource is depleted over time at a rate that depends on task difficulty.
Meeting abstract presented at VSS 2017