Temporal Retrieval Effects in Recognition Memory

We examine whether temporally defined associations play a role in item recognition. The role of these associations in recall tasks is well known; we demonstrate an important role in item recognition as well. In this study, subjects were significantly more likely to recognize a test item as having been previously experienced if the preceding test item was studied in a temporally proximal list position than if the preceding test item came from a more distant list position. Further analyses showed that this associative effect was almost entirely due to cases in which the preceding test item received a highest-confidence recognition judgment. In old/new recognition, a subject studies a series of items and is then shown a series of test probes. Subjects judge whether each probe is ‘‘old’’ (a repetition of an item that was presented on the studied list) or ‘‘new’’ (an item that had not appeared during the experiment). Widely studied for almost 100 years (Strong, 1912), the old/new recognition task is believed to measure itemspecific memory, devoid of interitem associations (Humphreys, 1978; Murdock, 1974). In contrast, recall tasks rely on strong temporal associations among items (Raskin & Cook, 1937). For example, in the free-recall task, recall of a list item tends to be followed by recall of an item studied in a nearby list position, even though the instructions permit subjects to recall items in any order they wish (Kahana, 1996). Many models of recognition and recall capture this distinction between item and associative information (Gillund & Shiffrin, 1984; Hintzman, 1988; Humphreys, Pike, Bain, & Tehan, 1989; Murdock, 1982; Shiffrin & Steyvers, 1997). In these models, item information reflects a weighted summed similarity between the probe item and the contents of memory. Associative information is typically stored independently by means of a conjunctive process that binds the information making up the individual items. Associations in these models not only are formed between simultaneously processed items, but also can span several items that co-occur in a working memory buffer (e.g., Kahana, 1996; Sirotin, Kimball, & Kahana, in press). Although many models assume a strict boundary between item information, driving recognition, and associative information, driving recall, it has long been recognized that recognition and recall are complex tasks that reflect the interaction of multiple memory systems, processes, or operations (see Kahana, Rizzuto, & Schneider, in press, for a review). In the case of recognition memory, a major line of research assumes that two distinct item-specific processes drive recognition judgments: a fast acting familiarity process and a slower recollective process (Arndt & Reder, 2002; Atkinson & Juola, 1974; Mandler, 1980; Yonelinas, 2001). Familiarity in these models coarsely maps onto the summed-similarity notion described in the preceding paragraph. Recollection is a recall-like process that involves the recovery of specific source information about the remembered item and is accompanied by a conscious experience of having seen or heard the target item (Tulving, 1985). What exactly is the nature of the recollective process that underlies some recognition judgments? We consider here the possibility that recollective experience reflects, in part, the shared information between items studied in temporal proximity rather than simply the detailed features of an individual item’s occurrence. In recall tasks, such associative features may be seen in the strong tendency for recall of an item to evoke memories of its neighbors (Kahana, 1996). To the extent that recognition of an item as ‘‘old’’ involves a recall-like recollection process, one might observe temporally defined retrieval effects in item recognition, despite the fact that the task does not explicitly test associative information. This could point to the existence of common associative mechanisms in associative recall and item recognition. Thus, our main goal in this article is to examine the effects of temporal co-occurrence on recognition performance. Light and Schurr (1973) offered some evidence for the possible role of temporal associative mechanisms in an item Address correspondence to Michael J. Kahana, Department of Psychology, University of Pennsylvania, Philadelphia, PA 19104, e-mail: [email protected], or to Marc W. Howard, Department of Psychology, Syracuse University, Syracuse, NY 13244, e-mail: marc@ memory.syr.edu. PSYCHOLOGICAL SCIENCE 898 Volume 16—Number 11 Copyright r 2005 American Psychological Society recognition task. For one group of subjects, old test probes were presented in the same order as they were studied; for a second group, the order of the old test probes was randomized. Subjects in the same-order group performed better on the recognition test than subjects in the random-order group. In a related study, Ratcliff and McKoon (1978) had subjects make recognition judgments on nouns read as part of a story. They found that recognition memory for an old item was enhanced when the previous old-item probe came from within the same linguistic unit (i.e., from the same proposition or sentence rather than a different proposition or sentence). With these linguistically organized materials, Ratcliff and McKoon did not find an effect of within-proposition interword distance. Their effect, rather, was carried by the co-occurrence of items within a given proposition or sentence. To assess the effect of temporal co-occurrence on recognition performance, we conducted a recognition memory experiment using pictures as stimuli. The recognition test was a random sequence of test probes that included the old items from the list intermingled with an equal number of new items that served as lures. Subjects pressed one of six keys in response to each probe, rating their confidence that they had seen it before from 1 (sure new) to 6 (sure old). The subjects’ ability to discriminate the old pictures from the new pictures had to be a consequence of their memory for the studied items. A recognition test might include the subsequence of test probes (. . . O23, N, O12, O7, N, N, O39, . . . ), where N denotes a new item and Ox denotes an old item from position x in the study list. We define the lag, r, between two successive old items (. . . Oi, Oj . . . ) as the distance, j i, between the items on their initial presentation (see Fig. 1). Suppose that recognition of a test item, Oi, brings forth the mental state that prevailed when Oi was first encoded. Suppose further that this retrieved mental state contributes to the retrieval environment that determines subsequent recognition judgments. Then, Fig. 1. Illustration of the item recognition task and the calculation of lag. After studying a series of pictures, subjects judged each picture in a test series as ‘‘old’’ (previously studied) or ‘‘new’’ (not previously studied). We define lag as the distance in the study list between successive ‘‘old’’ test items. Volume 16—Number 11 899 G. Schwartz et al.