Two-process recognition memory: Evidence from ROC curves

描述（由申请人提供）：识别记忆的单过程模型假设判断是基于对测试探针的熟悉程度。双过程模型假设基于熟悉的过程可以由回忆过程的输出来补充。双过程模型的一种证据来自于主观报告，即一个特定的测试项目是基于回忆还是熟悉，使用记忆-知道（R-K）程序（Tulving, 1985）。定量建模直到最近才应用于R-K判断，并且出现了许多候选模型。像识别模型一样，R-K模型也有一个和两个过程的形式。R-K判断的一维信号检测（SDT）解释最早由Donaldson（1996）提出，Wixted和Stretch（2004）对其进行了扩展。我们提出了一个二维SDT模型，称为STREAK （Rotello, Macmillan, & Reeder, 2004）。Reder等人（2000）、Yonelinas（2002）和Murdock（出版中）也提出了其他双过程模型。我们专注于区分这些替代模型，使用ROC曲线和分析RT分布作为主要的方法工具。目标1和目标3评估SDT模型中假设的变异性的来源。在识别任务中，可变性源于刺激之间的差异，而模型通常对所有类型的可变性都一视同仁。在R- k范式中，这是一个明显的过度简化：当然，有些项目比其他项目更有可能导致R反应。我们通过在典型的单词识别实验（目标1）中改变刺激特征来解决这个问题，并通过研究人脸记忆（目标3）来解决这个问题。Aim 2采用了一种不同的方法来评估模型，重点关注“旧”响应的RT分布，然后是R或K判断。这些分布的统计数据为判断是否基于相同的过程提供了信息。我们利用这一事实来测试R-K决策模型。这显然是一个基础研究项目。然而，了解健康成年人做出识别判断的机制对于理解脑损伤后或以后生活中可能出现的问题至关重要。这项研究将对失忆症、认知减缓等衰老效应和面孔失认症（无法识别面孔）产生影响；它也与目击者的记忆有关。

DESCRIPTION (provided by applicant): One-process models of recognition memory assume that judgments are based on the familiarity of the test probe. Two-process models assume that the familiarity-based process may be supplemented by the output of a recollective process. One type of evidence for two-process models has come from subjective reports of whether a particular test item is recognized on the basis of recollection or familiarity, using a remember-know (R-K) procedure (Tulving, 1985). Quantitative modeling has been applied to R-K judgments only recently, and a number of candidate models have emerged. Like models of recognition, models of R-K come in one- and two-process forms. The one-dimensional signal detection (SDT) explanation of R-K judgments was first proposed by Donaldson (1996) but has been extended by Wixted and Stretch (2004). We proposed a two-dimensional SDT model known as STREAK (Rotello, Macmillan, & Reeder, 2004). Other two-process models have been suggested by Reder et al. (2000), Yonelinas (2002), and Murdock (in press). We focus on distinguishing these alternative models, using ROC curves and analysis of RT distributions as the primary methodological tools. Aims 1 and 3 evaluate the source of the variability assumed in SDT models. In recognition tasks, variability arises from differences among the stimuli, and models typically treat all kinds of variability alike. That this is an obvious oversimplification in the R-K paradigm: surely some items are more likely to lead to R responses than others. We address this issue by varying stimulus characteristics in our typical word-recognition experiments (Aim 1), and by studying memory for faces (Aim 3). Aim 2 takes a different approach to model evaluation, focusing on the RT distributions for "old" responses that are followed by R or K judgments. The statistics of these distributions are informative about whether the same process(es) underlie the judgments. We take advantage of this fact to test models of R-K decisions. This is clearly a basic research program. However, understanding the mechanisms by which recognition judgments are made in healthy adults is critical to understanding the problems that can develop after brain injury or in later life. This research will have implications for amnesia, aging effects such as cognitive slowing, and prosopagnosia (the inability to recognize faces); it is also relevant for eyewitness memory.