Purchase this article with an account.
Todd Horowitz, Yoana Kuzmova; Predictability matters for multiple object tracking. Journal of Vision 2010;10(7):243. doi: https://doi.org/10.1167/10.7.243.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Most accounts of multiple object tracking (MOT) suggest that only the spatial arrangement of objects at any one time is important for explaining performance. In contrast, we argue that observers predict future target positions. Previously this proposition was tested by studying the recovery of targets after a period of invisibility (Fencsik, Klieger, & Horowitz, 2007; Keane & Pylyshyn, 2006). Here, we test the predictive hypothesis in a continuous tracking paradigm. In two experiments, we asked observers to track three out of twelve moving disks for three to six seconds, and varied the average turn angle. We held speed constant at 8°/s, but direction for each disk changed with probability. 025 on each 13.33 ms frame. Observers marked all targets at the end of the trial. Experiment 1 used turn angles of 0°, 30°, and 90°, while Experiment 2 used 0°, 15°, 30°, 45°, 60°, 75°, and 90°. Turn angle was fixed for all objects within a trial but varied across trials. In both experiments, accuracy was maximal at 0° and declined as turn angle increased (Exp 1: p = .001; Exp 2: p = .001). In Experiment 2, the steepest decline in accuracy was from 0° to 30°, while accuracy was roughly constant from 45° to 90°. These data demonstrate that it is easier to track predictably moving targets. Since velocity, density, and other factors known to affect MOT performance were constant, this suggests that observers predict target motion online to improve tracking. Furthermore, the pattern of data in Experiment 2 is compatible with a model in which the visual system assumes that target trajectories will vary only within a narrow 30° band.
This PDF is available to Subscribers Only