Dynamic in-flight shifts of working memory resources across saccades.
Dynamic in-flight shifts of working memory resources across saccades.
复制标题
DOI：
10.1037/xhp0000960
复制
发表时间：
2022-01
期刊：
影响因子：
--
通讯作者：
Husain M
中科院分区：
文献类型：
--
作者：
Udale R;Tran MT;Manohar S;Husain M
当人们从视觉环境中主动提取信息时，关于自然视觉中的记忆资源如何在连续的眼球运动和注视中分配的知之甚少。在这里，我们使用凝视视情眼动追踪来研究这些资源是如何从旧信息动态重新分配到进入工作记忆的新信息的。当参与者按顺序看项目时，我们在不同的时间中断了这个过程，在开始扫视时熄灭显示器。在短暂的间隔之后，参与者被问及所呈现的一个项目。奇怪的是，在所有的实验中，当更多的项目之前被固定时，最终（未固定的）扫视目标被更精确地回忆起来，也就是说，具有更长而不是更短的扫视序列。这一结果很难用目前的工作记忆模型来解释，因为回忆错误，即使是最后一个项目，通常也会随着记忆负荷的增加而增加。然而，这些发现可以用一个模型来解释，这个模型描述了资源是如何在每时每刻动态重新分配的。在每次扫视期间，通过消耗来自有限工作记忆的当前可用资源的一部分以及通过重新分配资源远离先前编码的项目来对目标进行编码。这些发现揭示了工作记忆资源是如何在主动视觉中跨记忆转移的。我们不断地进行快速的眼球运动，称为扫视，以检查我们周围的世界。短期存储需要暂时保存这些信息，以建立视觉场景的内部表示，或指导我们在视觉环境中的行动。众所周知，视觉短期记忆的容量是非常有限的，这就给我们带来了一个挑战：当我们在扫视序列中存储信息时，我们如何重新分配记忆资源来接收新的信息？在这里，我们表明，资源被动态地重新分配远离以前固定的项目存储在内存中，使新的眼跳目标的空间。即使在进行扫视之前，关于该目标的大量信息也被存储。我们的研究结果表明，有限的内存资源是如何动态分配，以支持我们的能力，从视觉场景中记住信息的机制。
Little is known about how memory resources are allocated in natural vision across sequential eye movements and fixations, as people actively extract information from the visual environment. Here, we used gaze-contingent eye tracking to examine how such resources are dynamically reallocated from old to new information entering working memory. As participants looked sequentially at items, we interrupted the process at different times by extinguishing the display as a saccade was initiated. After a brief interval, participants were probed on one of the items that had been presented. Paradoxically, across all experiments, the final (unfixated) saccade target was recalled more precisely when more items had previously been fixated, that is, with longer rather than shorter saccade sequences. This result is difficult to explain on current models of working memory because recall error, even for the final item, is typically higher as memory load increases. The findings could however be accounted for by a model that describes how resources are dynamically reallocated on a moment-by-moment basis. During each saccade, the target is encoded by consuming a proportion of currently available resources from a limited working memory, as well as by reallocating resources away from previously encoded items. These findings reveal how working memory resources are shifted across memoranda in active vision. We continuously make rapid eye movements called saccades to inspect the world around us. A short-term store is needed to hold this information temporarily, to build an internal representation of the visual scene, or guide our actions within the visual environment. It is well established that visual short-term memory is highly limited in capacity, which confronts us with a challenge: When storing information across saccade sequences, how do we reallocate memory resources to new incoming information? Here, we show that resources are dynamically reallocated away from previously fixated items stored in memory to make room for the new saccade target. A large amount of information is stored about this target, even before the saccade is made. Our results suggest a mechanism of how limited memory resources are dynamically allocated in order to support our ability to remember information from the visual scene.
登录
查看更多内容
影响因子：
1.8
作者：
Bays PM;Gorgoraptis N;Wee N;Marshall L;Husain M
通讯作者：
Husain M
影响因子：
1.9
作者：
Bridgeman B
通讯作者：
Bridgeman B
影响因子：
1.8
作者：
Bays PM;Catalao RF;Husain M
通讯作者：
Husain M
DOI：
10.3758/bf03206794
发表时间：
1995-08-01
期刊：
PERCEPTION & PSYCHOPHYSICS
影响因子：
--
作者：
HOFFMAN, JE;SUBRAMANIAM, B
通讯作者：
SUBRAMANIAM, B
影响因子：
2.6
作者：
Donkin, Chris;Kary, Arthur;Taylor, Robert
通讯作者：
Taylor, Robert