Srivastava Nisheeth, Vul Ed
Department of Psychology, University of California, San Diego.
Top Cogn Sci. 2016 Jan;8(1):335-48. doi: 10.1111/tops.12189. Epub 2016 Jan 8.
We present a computational model of multiple-object tracking that makes trial-level predictions about the allocation of visual attention and the effect of this allocation on observers' ability to track multiple objects simultaneously. This model follows the intuition that increased attention to a location increases the spatial resolution of its internal representation. Using a combination of empirical and computational experiments, we demonstrate the existence of a tight coupling between cognitive and perceptual resources in this task: Low-level tracking of objects generates bottom-up predictions of error likelihood, and high-level attention allocation selectively reduces error probabilities in attended locations while increasing it at non-attended locations. Whereas earlier models of multiple-object tracking have predicted the big picture relationship between stimulus complexity and response accuracy, our approach makes accurate predictions of both the macro-scale effect of target number and velocity on tracking difficulty and micro-scale variations in difficulty across individual trials and targets arising from the idiosyncratic within-trial interactions of targets and distractors.
我们提出了一种多目标跟踪的计算模型，该模型可对视觉注意力分配以及这种分配对观察者同时跟踪多个目标能力的影响进行试验级别的预测。该模型基于这样一种直觉，即对某个位置的注意力增加会提高其内部表征的空间分辨率。通过结合实证实验和计算实验，我们证明了在此任务中认知资源与感知资源之间存在紧密耦合：目标的低级跟踪会生成自下而上的错误可能性预测，而高级注意力分配会有选择地降低被关注位置的错误概率，同时增加未被关注位置的错误概率。早期的多目标跟踪模型预测了刺激复杂性与反应准确性之间的大致关系，而我们的方法既能准确预测目标数量和速度对跟踪难度的宏观影响，也能准确预测因目标与干扰项在试验内的特殊相互作用而导致的单个试验和目标之间难度的微观变化。
