Abstract
Research on the limits of visual tracking abilities —typically studied with multiple object tracking (MOT) — has seen a shift, recently, from fixed-resource views to flexible-resource views. Flexible resource views have been especially successful explaining the limits of tracking computationally, in terms of specific spatial interactions between tracked and untracked objects. But all extant flexible-resource models share the assumption that while participants may not always track the right individuals (i.e. they can confuse targets and non-targets), they always track the right number of individuals. Here we show that they do not. In Experiment 1, we reduced MOT to a simple spatial working memory task: participants had to remember between three and twelve static targets among a group of identical non-targets. After a delay, they were instructed to click on all the targets and to press an ‘OK’ button when done. Thus the number of clicks they could provide was unconstrained. With as few as six targets, participants misperceived the number of targets on 5.3% of trials. With increasing numbers of targets, enumeration became increasingly worse, jumping to 20% with just seven targets. These results demonstrate that participants do not always select the right number of targets at the start of a trial, though flexible-resource models explicitly assume that they do. Additionally, a set of two experiments demonstrated that longer tracking durations increase enumeration errors. In other words, participants do not only confuse targets and non-targets while tracking, they may also lose targets entirely. The faithful tracking of the right number of items has been an unsupported assumption in current flexible-resource models. Further experiments explored the role of speed in enumeration. Overall, we suggest that current models must make adjustments to accommodate the fact that participants do not always know how many targets are in a display.
Meeting abstract presented at VSS 2012