Found A, Müller H J
Birkbeck College, University of London, England.
Percept Psychophys. 1996 Jan;58(1):88-101. doi: 10.3758/bf03205479.
Search for odd-one-out feature targets takes longer when the target can be present in one of several dimensions as opposed to only one dimension (Müller, Heller, & Ziegler, 1995; Treisman, 1988). Müller et al. attributed this cost to the need to discern the target dimension. They proposed a dimension-weighting account, in which master map units compute, in parallel, the weighted sum of dimension-specific saliency signals. If the target dimension is known in advance, signals from that dimension are amplified. But if the target dimension is unknown, it is determined in a process that shifts weight from the nontarget to the target dimension. The weight pattern thus generated persists across trials, producing intertrial facilitation for a target (trial n + 1) dimensionally identical to the preceding target (trial n). In the present study, we employed a set of new tasks in order to reexamine and extend this account. Targets were defined along two possible dimensions (color or orientation) and could take on one of two feature values (e.g., red or blue). Experiments 1 and 2 required absent/present and color/orientation discrimination of a single target, respectively. They showed that (1) both tasks involve weight shifting, though (explicitly) discerning the dimension of a target requires some process additional to simply detecting its presence; and (2) the intertrial facilitation is indeed (largely) dimension specific rather than feature specific in nature. In Experiment 3, the task was to count the number of targets in a display (either three or four), which could be either dimensionally the same (all color or all orientation) or mixed (some color and some orientation). As predicted by the dimension-weighting account, enumerating four targets all defined within the same dimension was faster than counting three such targets or mixed targets defined in two dimensions.
当目标可能出现在几个维度之一而不是仅一个维度时，寻找与众不同的特征目标所需的时间更长（Müller、Heller和Ziegler，1995年；Treisman，1988年）。Müller等人将这种代价归因于辨别目标维度的需要。他们提出了一种维度加权理论，其中主地图单元并行计算特定维度显著信号的加权和。如果目标维度事先已知，来自该维度的信号会被放大。但如果目标维度未知，则在一个将权重从非目标维度转移到目标维度的过程中确定它。由此产生的权重模式在各次试验中持续存在，为与前一个目标（试验n）维度相同的目标（试验n + 1）产生试验间促进作用。在本研究中，我们采用了一组新任务，以便重新审视和扩展这一理论。目标沿着两个可能的维度（颜色或方向）定义，并且可以具有两个特征值之一（例如，红色或蓝色）。实验1和2分别要求对单个目标进行缺失/存在判断和颜色/方向辨别。结果表明：（1）两项任务都涉及权重转移，不过（明确地）辨别目标维度需要一些除了简单检测其存在之外的额外过程；（2）试验间促进作用实际上（在很大程度上）是维度特定的，而不是特征特定的。在实验3中，任务是计算显示中的目标数量（三个或四个），这些目标可以在维度上相同（全是颜色或全是方向）或混合（一些颜色和一些方向）。正如维度加权理论所预测的，枚举在同一维度内定义的四个目标比计算三个这样的目标或在两个维度中定义的混合目标要快。
