中文摘要：

采用变化觉察范式,通过两个实验对不同类型视觉信息在工作记忆中的存储机制进行了探讨.实验一中要求记忆的客体由颜色和形状两基本特征构成;实验二中的客体为具有不同颜色及开口朝向的兰道环.通过观察检测项变化方式对记忆绩效的影响,获得以下结论：（1）对颜色和形状两基本特征的存储以整合客体为单位;（2）兰道环的颜色与开口朝向信息难以以整合客体方式存储.笔者推测,视觉工作记忆中存储的整合客体并非如现有理论假设的,是在集中注意参与下创建的,而是由并行加工阶段所获信息构成;客体中所含需集中注意提取的细节特征信息难以与其基本特征信息整合于该客体表征.

英文摘要：

Introduction During the past years, the storage mechanism of visual working memory （VWM） has been extensively studied. Luck ＆ Vogel （1997） indicated that the object stored in the visual working memory in an integrated manner, and that only 4 objects could be retained in VWM, regardless of the number of features containing in individual objects, which was called ＂strong object＂ hypothesis. However, more recent research revealed that object-based advantage was weakened when heterogeneous features were integrated in different parts of an object （Xu, 2002）. The advantage was further diminished when the conjunction of homogenous features served as the memory materials （Olson ＆ Jiang, 2002; Wheeler ＆ Treisman, 2002）. Based upon these results, Olson ＆ Jiang （2002） put forward the ＂weak object＂ hypothesis, suggesting that VWM is limited by both the number of objects and the composition of those object features. However, this hypothesis does not explain clearly what kind of information can be integrated into an object and what kind of information can not. Using two different sets of materials, we explored the storage mechanism of VWM. Inspired by the fact that the pattern of perceptual processing can be divided into parallel processing with spread attention and serial processing with focal attention, we hypothesized that attention would serve as the key factor in determining mechanism of storage, and that only information processed in parallel would be stored in an integrated-object retained in VWM. Method Using change detection paradigm, we ameliorated Wheeler ＆ Treisman＇s research （2002） by changing block design to random design so that the influence of memory strategy could be avoided. Two experiments were conducted, in which the subjects were required to judge whether the test item displayed in the memory item. In experiment 1, the objects were defined by color and shape, which could be processed in parallel. In experiment 2, the objects were defined by landolt tings w