Application of formal modeling to understand limitations in visual working memory

描述(申请人提供)：我们在短时间内存储视觉信息的能力有限，这体现在视觉工作记忆(VWM)任务中的低准确性，该任务要求参与者将存储的表征与随后的显示进行匹配，以确定是否发生了变化。记忆中可以保存的表征的数量和这些表征与视觉感知的匹配程度都有限制。然而，VWM理论家通常将变化检测的错误归因于数量的限制。已经开发了一种称为混合建模的方法，该方法将参与者的反应建模为两个独立分布的总和，一个用于猜测，另一个用于非猜测反应。这种方法可以估计VWM表示的数量和保真度，并确定每个属性如何受到实验操作的影响。所提出的研究调查了几个对提供信息理论重要的问题：a)对VWM任务和对象属性的数量和保真度的估计是否一致？B)是否与其他能力受限的任务共享限制数量和保真度的流程？以及c)如何在最佳条件下限制VWM？这些问题将通过将混合建模和其他分析技术应用于各种实验操作下的VWM任务来研究。

DESCRIPTION (provided by applicant): Our limited ability to store visual information over short delays is demonstrated by poor accuracy in visual working memory (VWM) tasks that require participants to match stored representations to a subsequent display to determine whether a change occurred. There are limitations both in the quantity of representations that can be held in memory, and in how well those representations match visual perception. However, VWM theorists have generally attributed errors in change detection to limitations in quantity. A method has been developed, termed mixture modeling, that models participants' responses as the sum of two independent distributions, one for guesses and one for non-guess responses. This method can estimate the quantity and fidelity of VWM representations, and determine how each property is affected by experimental manipulations. The proposed studies investigate several issues important for informing theory: a) are estimates of quantity and fidelity consistent across VWM task and object properties? b) are processes that limit quantity and fidelity shared with other capacity-limited tasks? and c) how is VWM limited under optimal conditions? These issues will be investigated by applying mixture modeling and other analysis techniques to VWM tasks under various experimental manipulations.