Flexible resource allocation and prediction in attentive tracking

许多日常任务需要同时跟踪多个感兴趣的对象。例如，在交通拥挤的情况下，汽车司机必须跟踪移动的物体，如汽车、行人和骑自行车的人。这些任务的挑战在于在动态场景中保持视觉选择的稳定性，其中对象甚至可能暂时不可见。这种情况发生在驾驶中，例如，如果一辆大卡车暂时挡住了司机照顾的汽车。在该项目的第二个资助期，我们通过研究稳定选择和稳定机制的挑战，以第一个资助期的研究结果为基础。我们开发并测试了一个描述人类如何在动态场景中保持视觉选择稳定性的帐户。我们提出将视觉跟踪中的资源灵活地部署在运动目标上。这个假设的核心是观察者预测是为了跟踪动态对象。这种预测过程在感知动态场景时提供了稳定性。只要预测的视觉输入与实际的视觉输入没有偏差，视觉选择和注意跟踪是稳定的。预测过程使用目标位置的时空信息（速度、方向、纹理运动）和/或特征信息。此外，我们提出注意跟踪对社会线索敏感，这代表了一种特殊的对象特征，从使用凝视线索范式的研究中可以知道。我们提出了四个实验系列来检验这一说法。首先，我们使用违反目标运动信息可预测性的跟踪显示来验证目标运动信息有助于MOT预测过程的假设。其次，所提出的预测过程也可以评估目标特征信息。因此，我们采用了对原始MOT范式的修改，这使我们能够在对象上添加冲突的特征信息，并操纵目标和干扰对象之间的相似性。第三，我们提出视觉搜索过程涉及重新定位一个或多个丢失的目标物体。移动的物体要么在整个试验过程中可见，要么在短时间内不可见。这些实验的结果使我们能够指定特征信息如何有助于跟踪，并描述和指定视觉记忆如何有助于跟踪。在第四个实验系列中，我们使用社会凝视线索来测试凝视线索是否有助于在注意跟踪中提供方向信息，以及对注意跟踪中选择稳定性的影响。本提案的目标是开发一个全面的帐户来解释人类对多个移动物体的注意跟踪，并在行为和眼动追踪实验中测试该模型的预测。

Many everyday-tasks require the simultaneous tracking of multiple objects of interest. In heavy traffic, for example, car drivers have to keep track of moving objects such as cars, pedestrians, and bikers. The challenge in such tasks is to maintain stability of visual selection in dynamic scenes, in which the objects may even be temporarily invisible. This happens in driving, for example, if a large truck temporarily occludes cars the driver attends to. In the second funding period of this project we build on findings of the first funding period by studying challenges of stable selection and stability mechanisms. We develop and test an account that describes how humans maintain stability of visual selection in dynamic scenes. We propose that the resources in visual tracking are deployed flexibly on the moving objects. Central to this assumption is that observers predict in order to keep track of dynamic objects. This prediction process provides stability during perceiving dynamic scenes. As long as there is no deviation between predicted and actual visual input, visual selection and attentive tracking are stable. The prediction process uses spatio-temporal information (speed, direction, texture motion) and/or feature information available at target locations. Further, we propose that attentive tracking is sensitive to social cues, which represent a special case of object feature as known from research using the gaze-cueing paradigm.We propose four experimental series that test this account. First, we test the hypothesis that object motion information contributes to the prediction process in MOT with tracking displays that violate the predictability of object motion information. Second, the proposed prediction process might also evaluate object feature information. We therefore employ a modification of the original MOT paradigm, which enables us to add conflicting feature information on the objects and to manipulate similarity between target and distractor objects. Third, we propose that processes of visual search are involved in re-locating one or more lost target objects. The moving objects are either visible during the complete trial or invisible for a short period of time. Results of these experiments allow us to specify how feature information contributes to tracking and to characterize and specify how visual memory contributes to tracking. In the fourth experimental series we use social gaze cues both to test if gaze cues are helpful in providing direction information in attentive tracking and effects on the stability of selection in attentive tracking. The goal of this proposal is to develop a comprehensive account explaining human attentive tracking of multiple moving objects and to test the predictions of this model in behavioral and eye tracking experiments.