Abstract
Attentional tracking is traditionally assumed to be capacity-limited in that only a few discrete items can be tracked (e.g., Pylyshyn’s FINST-based visual index theory). However, some recent studies suggested that attention to spatiotemporal objects is only limited by a pool of resources that can be allocated flexibly to track a small number of items with high resolution or a large number of items with low resolution.
In a multi-object tracking task, observes tracked a variable number of targets among distractors in random motions for 4.5 seconds, and then reported whether a probed item was one of the tracked items on a 6-point confidence scale. Receiver operating characteristic (ROC) curves were constructed from the confidence data. Two independent parameters, modeled from the slot+averaging hypothesis (Zhang & Luck, 2008), were estimated from ROCs, to represent resolution of the tracked items (d prime) and the probability that the probed item is tracked (PT). We found that both PT and resolution can be quantitatively accounted for by the slot+averaging model. That is, the maximum number of items that can be tracked is capped at about 3.5 at set size four and six, suggesting limited capacity. In addition, resolution of tracking is significantly better at set size two, when more than one attentional indices can be allocated to track a single item, compared to set size four & six where only one spatial index can be allocated to one item. In Experiment 2, we manipulated the speed of the random motion. PT significantly decreased from slow speed to intermediate speed, but not from intermediate speed to fast speed. In contrast, resolution of the tracked item remained constant across motion speeds. Together, these results suggest that FINST-based visual indices are based on a limited set of fixed-resolution representations.
Meeting abstract presented at VSS 2012