نقش واکنش مرتبط در مطالعه و یادآوری معنایی هدایت شده در آزمون در حافظه کاذب: فرضیه پاتریک و دییز

False recall is found for semantically related words that are not presented on both categorized and associatively structured study lists. Four experiments provide evidence that the associative list method produces false memories because of semantic processes involved in studying list words (the Kirkpatrick hypothesis), but that false memories produced by categorized lists occur because of the use of semantic knowledge at test (the Deese hypothesis). In a free association task, words from associative lists, but not categorized lists, tended to evoke critical words as responses, indicating that our categorized list words have low associative strength to critical nonpresented items. Studying those associative lists, but not the categorized ones, produced indirect priming effects in stem completion. Critical nonpresented words from categorized lists showed a priming effect only when participants were instructed at test to try to complete stems with studied list words (i.e., stem cued-recall). The results highlight important differences between categorized and associative list methods, and indicate that false memories can be caused by semantic processes that occur at the time of a memory test.

People often misremember events, sometimes incorrectly claiming to recall or recognize nonpresented words that are semantically related to studied items. For example, false recall and recognition of specifically targeted nonpresented words have been demonstrated frequently when participants study and recall lists constructed from words that are all associated with a single nonpresented critical word (e.g., Deese, 1959; Read, 1996; Roediger III & McDermott, 1995). Such lists, sometimes called DRM lists (named for popularly cited works by Deese, and by Roediger & McDermott), are often designed such that list words have high associative strength to critical nonpresented words, and will be referred to herein as associatively structured lists, or simply associative lists. A similar false recall effect occurs for categorized lists of words that omit highly typical or dominant category members (e.g., Smith et al., 1998, Smith et al., 2000 and Smith et al., 2001). That is, when highly dominant category members (e.g., chair for the category furniture, orange for fruit) are left off of presented lists of common category members, the common nonpresented words are often falsely recalled. These false memories have been termed semantic confusion errors by Smith et al. (2001); such intrusions and false alarms are made because critical items are semantically related to the materials that were actually studied. In the present study we focus on the question of when, in the course of learning and remembering, semantically influenced false memories occur. Are such semantic confusion errors caused by processes that occur as a function of studying word lists, processes that occur when materials are tested and remembered, or both? In an early description of false memory effects, Kirkpatrick (1894) discussed the effects of semantically related material on false memories: There were some incidental illustrations of false recognition. About a week previously in experimenting upon mental imagery I had pronounced to the normal students ten common words. Many of these were recalled and placed with the memory list. Again, it appears that when `spool,' `thimble,' `knife,' were pronounced, many students at once thought of `thread,' `needle,' `fork,' which are so frequently associated with them. The result was that many gave those words as belonging to the list (Kirkpatrick, 1894, p. 608). Kirkpatrick considered the effect of semantic confusion on false memory to be a study phenomenon, attributing the errors to the time when the list of words was first pronounced. Thus, the Kirkpatrick hypothesis states that false memories occur because of processes that take place when words are studied. Some 65 years later, Deese (1959) found similar false memory effects using associatively structured lists, constructed such that each word on a list was closely associated with a single nonpresented linking word. For example, a list might contain the words “dream,” “pillow,” “nap,” and “bed,” in addition to other words that are strongly associated with the critical nonpresented linking word, “sleep.” In contrast to Kirkpatrick's account, Deese described the effects of associated words as occurring at the time of the memory test, stating that, “in the process of recollection, words and concepts associated with remembered items will be added” (p. 21). The Deese hypothesis indicates that recall or recognition of some of the list words at test activates associations to the critical nonpresented word, thereby leading to false memory of the nonpresented words.1 In the present study, we extend the Deese hypothesis to include effects at test of any type of semantic knowledge, including not only associations, but category knowledge and conceptual information, as well. The preponderance of false memory research with associatively structured lists supports the Kirkpatrick hypothesis. Experiments that manipulate variables at study often find effects of those study factors on false recall and recognition. For example, some studies have reported effects of the level of processing of associative list words at study on subsequent false memory (e.g., Rhodes & Anastasi, 2000; Thapar & McDermott, 2001; Toglia, Neuschatz, & Goodwin, 1999). Likewise, whether associative lists are presented in a blocked vs. random order at study affects false memory (e.g., Mather, Henkel, & Johnson, 1997; McDermott, 1996). Indirect priming effects found by McDermott (1997) and McKone and Murphy (2000) are also more consistent with Kirkpatrick's (1894) hypothesis, supporting the notion that associative responses during study cause false memories. McDermott (1997) and McKone and Murphy (2000) found that nonpresented words from associatively structured lists were indirectly primed, as measured by a stem completion test. In those experiments few or no relevant associates to the critical nonpresented words were presented on the stem completion test, so associative responses leading to the critical word at test were unlikely to occur. Thus, the stem completion findings in those experiments can be attributable to processes that occur during study. Further evidence of such associative responses at study causing false memories has been reviewed and provided by McEvoy, Nelson, and Komatsu (1999). Whether such associative responses are cases of unconscious spreading activation, implicit associative responses (IARs, e.g., Bousfield, Whitmarsh, & Danick, 1958; Kimble, 1968; Underwood, 1965) that are covertly but consciously experienced, or strengthening of pre-existing associations (Zeelenberg, Shiffrin, & Raaijmakers, 1999) is not directly addressed in the present study. What is addressed is whether semantic knowledge causes false memories solely because of study processes, or whether effects can also occur because of processes that take place at the time when a test occurs. Little evidence from studies with associative list materials supports the Deese hypothesis, which attributes false memories to processes initiated when memory testing occurs, and some results directly contradict the Deese hypothesis. For example, Roediger III, McDermott, and Marsh (2000), using associative list materials, tested critical nonpresented words that were preceded by varying numbers of associated list words. The Deese hypothesis predicts that false recognition and recall should be increasingly likely as more list words are encountered at test. Roediger et al. found, however, that the test position of critical nonpresented words did not affect false memories on a word stem cued-recall test, or on a yes/no recognition test. These results directly contradict the Deese hypothesis that false memories are test-induced. The likelihood that people would have associative responses to critical nonpresented words seems quite high for associatively structured lists because those list words were selected on the basis of high backward association strength. Backward association strength refers to the probability that a presented list item will evoke the critical nonpresented word (e.g., dream evokes sleep; Robinson & Roediger III, 1997). Associative responses that activate the critical nonpresented item, therefore, are more likely for lists with high backward association strengths. The categorized lists used by Smith et al. (1998), on the other hand, were not chosen to selectively evoke associative responses of critical nonpresented category members. That is, although critical nonpresented words are conceptually related to presented words in the categorized list method, the categorized lists may have low backward association strengths in relation to critical nonpresented items. Despite the low backwards associative strengths of some categorized lists, such lists might evoke false memories for reasons other than associative responses that occur at study. The category structure of categorized lists may guide memory at the time of testing, more in line with the Deese hypothesis. Such a hypothesis is consistent with the findings of Smith et al. (2000), who found that the category structure of categorized lists strongly influenced false recall of critical nonpresented category members. In that study, nonpresented category members that were highest in output dominance and typicality were more likely than nondominant atypical nonpresented category members to be falsely recalled. Therefore, although studying such list words might not activate critical nonpresented words (the Kirkpatrick hypothesis) via backwards associations, recalling such lists should nonetheless elicit critical items (the Deese hypothesis) via the category structure. In the present study we first selected a set of categorized and associative lists, and compared the associative strengths of the two types of lists. In Experiment 1 participants gave free associations to words from a set of associative and categorized lists. It was predicted that free association to words from associatively structured lists would be far more likely than members of categorized lists to elicit critical nonpresented words as responses. In Experiments 2 and 3, it was predicted that indirect priming of nonpresented categorized words would not occur, because such effects appear to be due to associative processes at study. It was also hypothesized that category knowledge used to guide recollection would cause false recall of critical nonpresented items from categorized lists. Experiment 4 tested a different prediction of the Deese hypothesis with categorized list materials. In Experiment 4, instructions to complete stems were either incidental, as in Experiments 2 and 3 (i.e., participants were told, “fill in each stem with any word that comes to mind”), or intentional (i.e., “use studied words to fill in stems whenever possible”). Intentional instructions on the stem completion test of Experiment 4 should highlight the category structures of the categorized lists; therefore such instructions were predicted to bring about an indirect priming effect in stem completion.2 That is, intentional instructions change the task into a stem cued-recall test that should show the same retrieval-guided false recall effect seen in the other recall tests. Because the instructions were manipulated at the time of the test it was predicted that an effect would occur with categorized list materials. Experiment 1 Two sets of lists were selected: associative lists drawn from McDermott's (1997) materials, and categorized lists taken from Smith et al.'s (2001) materials. In Experiment 1, participants were asked to give free associations to both types of list words, with the 10 words in each list presented in a single block. Critical items from the lists, which were not presented on the lists, were expected to occur as associative responses quite often for associatively structured lists, but far less often for categorized lists.