نقش قدرت مرتبط و به یاد سپاری منبع در زمینه حافظه کاذب

We tested the impact of associative strength and retrieval heuristics in false source memory. We arranged 12-item associative lists in descending order of backward associative strength to a critical non-presented item and then split them into 6-item sub-lists at the median. High- and low-strength sub-lists were correlated with presentation source. Source claims for falsely remembered critical items were more often to the presentation source of high-strength items, which we have labeled the source-strength effect. With only two exceptions in Experiment 2, the source-strength effect was produced even when differences in source memorability led participants to infer generally that items lacking source evidence were presented in the low-strength source. Indeed, the influence of associative strength on source attributions was shown to be completely independent of this inference strategy in a third experiment. This study demonstrates that source details of concepts most highly related to critical items are retrieved with false memories.

After people encode a list of words related associatively to a non-presented theme item (henceforth known as the critical item), that item is falsely recalled and recognized at high rates (Deese, 1959, Read, 1996 and Roediger and McDermott, 1995). This procedure of presenting people with lists of items associated to a non-presented item, commonly called the DRM procedure (after Deese, 1959 and Roediger and McDermott, 1995), has prompted many studies examining its theoretical implications (for a review, see Roediger, McDermott, & Robinson, 1998). A most striking finding in this literature is the degree to which falsely remembered critical items are accompanied by vivid feelings of recollection (e.g., Brainerd et al., 2001 and Roediger and McDermott, 1995). People report remembering attributes of critical items such as the modality in which they were presented, associations thought of upon ostensibly encountering them in the study list, and where they were positioned in a study list, among other recollective attributes (e.g., Gallo et al., 2001, Hicks and Marsh, 1999, Mather et al., 1997, Neuschatz et al., 2001 and Norman and Schacter, 1997). In the current study, we explored the reasons why people come to believe that critical items were presented in a specific context by investigating the causes of an effect reported by Hicks and Hancock (2002). These researchers presented subsets of DRM lists to people in two different contexts, a male versus female speaker on a videotape. The male presented half of the items in each list theme, and the female presented the remaining items. Most important, the researchers explicitly manipulated the pairing of items to each context. In the control condition, they randomly assigned half of the items in a given list to the male or female context (cf. Hicks & Marsh, 1999). When critical items were falsely recalled or recognized, context attributions were no more likely to the male as compared to the female source. In the experimental conditions, however, Hicks and Hancock deliberately correlated presentation context to list items of either low or high associative strength with reference to the critical item. For example, items in a given DRM list that were higher in associative strength to the critical item were presented by the male, whereas the remaining list items were presented by the female. In these conditions, context attributions to critical items were more often to the context that presented list items of higher associative strength. We describe the finding that the critical item seems to take on the context characteristics of its strongest associates as the source-strength effect. The Source Monitoring Framework (SMF; Johnson et al., 1993 and Mitchell and Johnson, 2000) describes two general types of processes that influence how memories are attributed to certain contexts (or sources), and either type of process could potentially explain the source-strength effect. Some attribution processes are classified as memory processes. To translate remembered information into specific attributions, the rememberer sets criteria defining the types and amount of information that should be retrieved for memories from a certain source, evaluates the match between these criteria and the information retrieved for a given item, and decides that the item originated in the candidate source if the match is high. Inferential processes also impact source attributions. The SMF describes inferential processing as the use of general knowledge (as opposed to episodic details) to determine which of several contexts is the most plausible or logical source of a given item (e.g., Bayen et al., 2000 and Hicks and Cockman, 2003). Inferential processes allow people to make source attributions on the basis of educated guessing, as opposed to evidence from memory. From the perspective of the SMF, there are two types of mechanisms that may underlie the source-strength effect. The effect may arise because people retrieve source-specific details when critical items are probed. Such details may be retrieved directly from the critical item representation or from studied words that are most highly associated to the critical item. These alternatives are considered separately in General discussion, but at the core of each is the assumption that source details from the most highly associated list words are the most likely to be retrieved when critical items are considered on a memory test. Therefore, either mechanism represents a mnemonic locus of the effect that produces the illusion that critical items were presented in a particular context. Alternatively, associative strength may impact inferential processes, leading participants to guess that the critical item shares a source with its most highly associated words. We explored the latter possibility in the form of a specific inference strategy that may explain patterns of source attributions in the DRM paradigm. Specifically, when asked to indicate the source of a critical item, participants may retrieve related list words that they studied, determine the source of these words, and infer that the critical item shares the same source (Gallo et al., 2001 and Mather et al., 1997). This recalling-associates strategy is particularly likely when all of the list words related to a critical item are presented in a single source, and studies using this methodology have found that participants are as likely to attribute critical items to the source of their list items as they are to correctly specify the source of the list items themselves ( Gallo et al., 2001, Gallo and Roediger, 2003 and Mather et al., 1997). When words from a list are split between two sources (e.g., Hicks & Hancock, 2002), this inferential process could create the source-strength effect if one assumes that highly associated list words are the most likely to come to mind when participants try to retrieve words related to a critical item. The critical aspect of this alternative hypothesis is that participants do not directly retrieve source details associated with a critical item. Instead, they infer that critical items share a source with the most highly associated list items following a failed attempt to remember source details directly associated to the critical item. To avoid confusion, we will present a specific example illustrating the differences between the memory-based and inference-based explanations of the source-strength effect. Assume that a person studied words associated to the critical item “sleep,” by hearing the strong associates over headphones (e.g., “nap” and “bed”) and seeing the weak associates on a computer screen (e.g., “coma” and “sedative”). According to the memory-based account, when the person searches for source evidence associated with “sleep,” she should find evidence consistent with the auditory source. That is, she should experience an illusory memory of hearing the word “sleep” in the study phase created by the strong associative relationship of the heard items to “sleep.” According to the inference-based account, when the person searches for source evidence associated with “sleep,” she should find no evidence in favor of the word being either seen or heard in the study phase. She may nevertheless be convinced that “sleep” was studied because it is highly familiar and decides to render a source decision. To choose between sources, she would retrieve words associated with “sleep,” and words like “nap” and “bed” should be the most likely to come to mind. If she remembers that “nap” and “bed” were heard, then she should guess that “sleep” was heard as well. These two scenarios exemplify the critical difference between the two theoretical alternatives under investigation. In the first scenario, source evidence from associates is genuinely mistaken for evidence encoded for the critical item itself. In the second scenario, source evidence from associates is used to inform a guessing strategy. The purpose of our study is to determine if the source-strength effect reflects memory or inferential processes. To achieve this goal, we tested situations in which a powerful inference strategy other than the recalling-associates strategy would guide source attributions. The alternative inference process that we investigated is the memorability heuristic, a more general version of the it-had-to-be-you effect ( Johnson et al., 1981 and Marsh et al., 1997). When participants are presented with items from two sources that differ in their memory strength or memorability (such as self-generated versus other-generated items), they are much more likely to erroneously attribute items to the low-memorability source (other-generated) than to the high memorability source (self-generated). People expect to have vivid memories for items encountered in a high memorability source, and they assume that items were presented in the low memorability source whenever mnemonic evidence is sparse. To evaluate how the source-strength effect interacts with the memorability heuristic, we compared sources with large differences in memorability (e.g., heard versus generated words) to sources of roughly equal memorability (e.g., heard versus seen words) both between and within our experiments. Evaluating sources that differ in memorability will help to determine if the source-strength effect reflects source evidence that is directly retrieved when critical items are tested. Under the hypothesis that source details are not associated with critical items, participants should have little or no mnemonic evidence on which to base their source decisions for critical items they falsely believe were presented. When sources differ in memorability, participants should assume that the item originated from the low-memorability source because vivid source information should be available if the item was from the high memorability source. When the sources are of roughly equal memorability, participants will have to resort to a different guessing strategy, perhaps recalling associated studied words and assuming that the test word shares the same source. Thus, if the source-strength effect is based on inferential processes, the effect may be observed only when source alternatives do not differ in memorability. In other words, any memorability difference between sources should encourage an alternative inference strategy that preempts the recalling-associates strategy driving the source-strength effect. By contrast, if critical items take on the source characteristics of their high-strength list words, the source-strength effect should be observed even when sources differ in memorability. For example, even if participants require vivid source information to claim that an item was studied as a picture, they should be more likely to find such vivid details for critical items whose high associates were studied as pictures than for critical items whose low associates were studied as pictures. Participants will be guided by source evidence for critical items when such evidence is available and apply the memorability heuristic for critical items when such evidence is lacking. Experiment 1 We used visual and auditory sources in Experiment 1. Past work in source memory has demonstrated that the memorability of seen and heard sources is roughly similar (Hicks and Marsh, 2001 and Marsh and Hicks, 1998). Each DRM theme was split into two sub-lists by separating the strongest six associates and the weakest six associates (cf. Hicks & Hancock, 2002). Each sub-list was presented either visually or aurally. As such, the source of a given sub-list was associated either with strong associates or with weak associates. In addition, items in a given sub-list were presented (a) in blocked fashion adjacent to the other sub-list from the same DRM theme—henceforth called the blocked format, (b) randomly interspersed with items from the counterpart sub-list, but all items from a given theme were presented before items from a different theme were presented—henceforth called the mixed format, or (c) presented quasi-randomly such that two items from the same sub-list could not be presented until at least one item from each of the remaining DRM themes was presented—henceforth called the spaced format. A schematic of these formats is presented in Fig. 1. The blocked format matches the procedure used by Hicks and Hancock (2002); we included the other encoding formats to provide within-experiment replications of the source-strength effect and to ensure that the effect is not an artifact of the particular procedure used in the original demonstration.

The results of the current study establish that the source-strength false memory effect is driven by memory processes rather than by inferential processes. Moreover, we have shown that presenting high-strength associates can create the illusion that critical items were encountered in highly memorable sources. As such, the source-strength methodology promises to play a prominent role in solving the mystery of how people come to experience vivid memories of events that did not occur.