社会与工程心理学研究室知识库

    工作记忆(working memory)是一种容量有限的、能在极短时间内保存信息的记忆系统。插槽模型(slot model)认为，个体存储的是固定个数、固定精度(precision)、相互独立的表征。工作记忆的容量极限(capacity limit)是由插槽个数决定的。资源模型(resource model)则认为，工作记忆的容量确实是有限的，但不是以表征的最大个数来定义的。工作记忆是一池可以灵活分配的资源，个体用这些资源存储许多低精度的表征或者少量高精度的表征，存储个数与存储精度成反比。
    本研究以个体工作记忆容量和记忆负荷等自变量来检验两类模型。插槽模型认为个体的容量极限是工作记忆存储的一个重要转折点。一旦记忆序列呈现的个数超出容量极限后，多于个体容量极限的视觉刺激便不会被加工。被试记忆的CSD C circular standard deviation，表示圆分布数据的离散程度，与被试的记忆精度成反比。)曲线在被试的容量极限内随着记忆负荷的增加单调递增，而超容量极限后趋于平缓。同时，Pm(探测刺激存储在工作记忆中的概率)在被试的容量极限内随着记忆负荷的增加发生缓慢下降，超容量极限后发生急速下降。资源模型则认为个体的存储个数没有上限，个体记忆的CSD曲线应该一直随着记忆负荷的增加单调递增。我们在研究一中介入了个体容量这个差异变量，并观测了CSD曲线的变化情况，初步检验了两类模型。
    研究一中实验一的结果表明，不论个体容量高低，随着记忆序列呈现的视觉刺激个数(set size)从1升到6，被试对于整个记忆序列表征的CSD都呈显著上升趋势，相邻两种视觉刺激个数条件之间CSD的差异均显著，并且CSD曲线的斜率一直显著大于。。在所有刺激个数条件下两组被试之间CSD的差异均显著。另外，高容量被试在更大的视觉刺激个数条件下出现了Pm的急速下降，低容量被试在更小的视觉刺激个数条件下出现了Pm的急速下降。研究一中实验二的结果表明，不论个体容量高低，随着记忆序列呈现的视觉刺激个数从1升到16，被试对于整个记忆序列的CSD都呈显著上升的趋势，相邻两种视觉刺激个数条件之间差异均显著，并且CSD曲线的斜率一直显著大于。。另外，高低容量组被试的Pm曲线变化趋势一致，且Pm并未在超各自的容量极限后才发生急速下降。
    两类模型关于线索对个体表征精度的影响也有不同的预测。并且以往的研究表明，高容量被试比低容量被试更能抑制干扰子的影响。因此，研究二探索高低容量组被试的工作记忆精度(csD)是否受线索提示的不同影响。结果表明高低容量组被试的工作记忆精度受线索提示的影响没有差异，并且都表现为线索有效条件下csD显著低于中性条件和线索无效条件，同时中性条件下的csD显著低于线索无效条件。
    另外，以往的研究都是通过测试单个随机抽取的视觉刺激，来代表整个记忆序列的平均表征精度。这种测试方法会忽略整个记忆序列中不同空间位置上的视觉刺激可能存在的精度差异。所以，研究三检测了记忆序列当中每个视觉刺激的csD，以完善工作记忆精度的测量方法，并对存储模型进行理论补充。结果表明，在视觉刺激个数为2和6条件下，被试在所有位置上的csD和Pm是有差异的。且每个被试从CSD最低的位置到CSD最高的位置，两两位置之间的CSD差异均显著。但是每个被试CSD最低的位置和CSD最高的位置不完全一致。
    以上实验结果说明了以下三点:一，CSD曲线在容量极限前后的变化趋势和表征精度受线索影响的实验结果都表明，资源模型对被试工作记忆存储的解释度更高;二，高低容量被试组的差异还表现在资源量上，但并不表现在对于线索的利用程度上;三，在无线索提示的条件下个体的空间资源配置在所有视觉刺激之间是不均等的，即使在没有线索提示的情况下个体的资源分配依然很灵活。

    Working memory is a memory system of storage and encoding in a very short period of time with limited capacity. Slot model holds that individuals store a fixed number of representations generally independent of each other with fixed precision in WM. And the individual WM capacity limit is defined as the number of slots. Resource model also accepts the limitation of working memory capacity. However, it regards working memory as a pool of resource which can be flexibly assigned. Everyone will hold a small number of high-precision items or a lot of low-precision items. The number varies inversely with the resolution of stimulus.
    In this research two kinds of models were examined through the regulation of some independent variables such as working memory capacity and set size of display. As the prediction of slot model, the curve of CSD (circular standard deviation, it represents the standard deviation of the circular distribution and in turn reflects the precision of the item stored in WM) will reach a stable plateau once the set size of memory display exceeds the upper limit of one's working memory. Instead, the curve of CSD will never tend to be smooth under the theoretical background of resource model. There isn't any important turning point sense of individual capacity limit. Consequently, in study 1 two kinds of models were preliminarily tested through the observation of change in the CSD curve.
    The results showed that there was a significant increase of CSD from set size 1 to 6 in both two individuals groups. The slope of the CSD curve had always been significantly greater than 0. The differences of CSD in all set sizes between the two groups were always significant. While the Pm of high capacity decreased suddenly from set size 4 and the Pm of low capacity individuals declined suddenly from set size 2. The results of experiment 2 showed that there was a significant increase of CSD among set size 1，2, 3, 4, 6, 10, 16 in both two individuals groups. Additionly the slope of the CSD curve had always been greater than 0. Whereas the trends of Pm curve were same in two individual groups. Which meaned, the Pm of two capacity individuals didn't decrease suddenly from their own capacity.
    Furthermore, there had different predictions based on the influence of the cue on storage precision between these two models. Besides, previous studies had found that high capacity individuals were more efficient in inhibiting the irrelevant information than low capacity individuals. Then we explored whether the cue would have different impact on the precision of two individual groups in study 2. The results showed that CSD under the valid cue condition were both significantly lower than neutral condition and the invalid cue condition in two groups. Then CSD under neutral condition were also both significantly lower than the invalid cue condition in two groups.
    Previous studies of two models always used the precision of a randomly selected item to represent the precision of the whole memory display. We consider this method will conceal the variation of precision among all the items as well as the resource allocation in the whole memory display. Then we carried out study 3 to examine the precision of all the items in the memory display. The results showed that all bars were encoded to different levels of precision no matter whether 2 or 6 bars were in the display, which could be inferred from the differences of Pm and CSD. In addition, every subject had his own ranking from the mostly precise location number to the fuzziest location number, with no distinct rule.
    These experimental results illustrate the following points.  First of all, the processing CSD will steadily increase with the increase of set size, which suggests resource model can explain the storage of WM better. Additionly the results of study 2 further prove it. Second, the differences between the high and low capacity individual groups also include the quantity of resource, but not the ability of using cues. Third, the spatial resource allocation among all the items in the field of vision are not equal without cues, which suggests that individual distribution of resources is also very flexible in the absence of cues.