توجه به پردازش جان کلام جهانی اثر سن در حافظه کاذب را از بین می برد

Counterintuitive age increases have been reported for the Deese–Roediger–McDermott (DRM) false memory illusion. The current theoretical explanation of this effect assumes that it is due to age increases in spontaneous interconnection of DRM list words’ meanings. To test this explanation, 11-year-olds and young adults studied DRM lists under conditions that (a) encouraged them to form such meaning-based connections or (b) discouraged them from doing so. In line with the explanation, the usual developmental increase in false memory disappeared in the first condition but was preserved in the second condition. Also in line with the explanation, conjoint recognition analyses revealed that encouraging participants to form meaning connections increased their reliance on gist-based similarity judgments.

The reintroduction of Deese (1959) list-learning paradigm by Roediger and McDermott, 1995 and Read, 1996 into the experimental literature provided researchers with a powerful laboratory-based method for using memory errors to understand cognitive processes. In the Deese–Roediger–McDermott (DRM) paradigm, participants study lists of words, such as door, glass, and shade, all of which are related to a critical nonpresented word, such as window. On later memory tests, participants falsely recognize and recall the presentation of nonpresented critical lures (e.g., window) at the same rate as they correctly recognize and recall actually presented list items (e.g., shade). Subsequently, researchers have used the DRM paradigm to better understand semantic memory in adults (e.g., Benjamin, 2001 and Roediger and McDermott, 1995) and the development of semantic memory throughout childhood (e.g., Brainerd et al., 2002 and Holliday and Weekes, 2006) and into the later stages of life (e.g., Dehon and Bredart, 2004 and Thomas and Sommers, 2005). Interestingly, the extant data demonstrate that young children (5–9 years of age) are relatively immune to the high rates of false recall and recognition of critical lures committed by adults (e.g., Brainerd et al., 2002, Dewhurst and Robinson, 2004 and Lampinen et al., 2006). In contrast, older children (11 years of age or older) falsely recognize and recall significantly more critical lures than do younger children (e.g., Brainerd, Forrest, Karibian, & Reyna, 2006). Age increases in the recognition form of the illusion are not artifacts of the high levels of response bias that younger children exhibit because those increases have been measured with bias-corrected statistics such as A′ and d′ (e.g., Lampinen et al., 2006 and Weekes et al., in press). Age increases in the recall form of the illusion are not an artifact of young children simply exhibiting poorer recall generally because they are quite capable of recalling studied items. In fact, of all the words recalled, younger children recall a higher proportion of targets than do older children (e.g., Brainerd et al., 2002). Finally, the empirical testing of the invariance of the DRM memory illusion across childhood was not pursued by accident; rather, increases in the DRM illusion across childhood were predicted on theoretical grounds ( Reyna and Brainerd, 1991, Reyna and Brainerd, 1995 and Reyna and Kiernan, 1994) before the confirming studies were conducted. An account of semantic false memories in children and adults that predicts developmental differences of false recall and recognition rates in the DRM paradigm is provided by fuzzy-trace theory (Brainerd and Reyna, 1998 and Reyna and Brainerd, 1995). According to fuzzy-trace theory, individuals encode verbatim and gist representations in parallel, and these representations vary in terms of their precision and content. For example, verbatim traces are composed of an item’s surface features. On recognition memory tests, retrieval of verbatim traces tends to lead to the acceptance of targets through a process known as an identity judgment and the rejection of semantically related lures through a process known as recollection rejection. In contrast, gist traces represent the overall meaning of an event. Retrieval of gist traces can result in the correct acceptance of targets but can also result in the false recognition of semantically related lures. Acceptance of items based on gist typically occurs through a similarity judgment. Brainerd and colleagues (Brainerd et al., 2003 and Brainerd et al., 2001) have also demonstrated gist to lead to subjectively compelling false memories via a process referred to as “phantom recollection” (illusory conscious recall of false detail). In addition to the dissociation between verbatim and gist traces, Reyna and Kiernan (1994) drew a distinction between gist representations for single events and gist representations that integrate meaning across multiple events. Subsequently, Neuschatz, Lampinen, Preston, Hawkins, and Toglia (2002) extended this distinction to what they referred to as “local” and “global” gist. They defined local gist as the meaning of an event when considered in isolation. For example, when studying a list of category exemplars, such as dog, cat, goldfish, turtle, cow, horse, pig, zebra, monkey, and lion, a child will likely extract the meaning of each item even if there is a failure to note how the items relate to one another (e.g., Brainerd & Reyna, 2007). Thus, local gist refers to the extraction of the meaningful features of an item without regard to the other items on the list, consistent with how information is encoded in global memory models such as MINERVA2 ( Hintzman, 1986). In addition, individuals can encode gist representations based on the meaningful relations among items through a process that Neuschatz and colleagues (2002) referred to as global gist extraction. Consider an instance in which an individual is presented with a series of numbers such as 1, 7, 13, 21, 105, 159, and so forth (for additional examples, see Reyna & Brainerd, 1991). While processing these items, it is likely that the meaning of each individual item will be processed, but an individual also might extract the global gist that the list contains a sequence of monotonically increasing odd numbers. In regard to the DRM paradigm, global gist is an emergent property of the relations among items when they are considered as a whole. A central characteristic of the DRM paradigm is that the critical lure is an emergent property of a list when it is considered as a whole. DRM list items relate to one another in that they all are semantic associates of the critical nonpresented item. It is proposed by some researchers that the global nature of DRM lists encourages false recall and recognition of critical lures (e.g., Lampinen et al., 2006). For instance, some researchers have compared the rates at which critical lures are falsely recalled and recognized when DRM lists are presented in a blocked format as opposed to a random format (Brainerd et al., 2003, McDermott, 1996 and Toglia et al., 1999). In the blocked format, all of the words related to a single theme are presented sequentially in the same list, with a different list of words being presented for each of the themes. In the random format, the same words are presented, but their presentation is randomly distributed across all of the lists presented at study. Young adults typically demonstrate greater true and false recall in the blocked condition than in the random condition (Brainerd et al., 2003 and Toglia et al., 1999). However, such blocking does not significantly affect the rate at which 6- and 9-year-olds falsely recognize critical lures (Lampinen et al., 2006). Such results suggest that structural manipulations that encourage individuals to process how DRM list items relate to one another influence the rate at which young adults, but not children, process and remember DRM lists. The failure of children to demonstrate the typical pattern of results in the DRM paradigm could result from a relative deficit in their ability to make meaning-based connections across the items presented on a DRM list. In terms of fuzzy-trace theory, children are less able than adults to extract global forms of gist. Indeed, developmental data from other memory paradigms suggest that young children do not automatically connect meaning across multiple events (for reviews, see Bjorklund, 1987 and Bjorklund and Muir, 1988). For example, after studying categorical lists of words, adults cluster their recall responses by categories, suggesting that they connect the meaning of a categorical list across the list’s exemplars. In contrast, children do not spontaneously cluster their output by categories when recalling the same types of categorical lists until 11 or 12 years of age, suggesting that until this point they do not spontaneously connect the meaning of a categorical list across the list’s exemplars (Bjorklund, 1987 and Schneider and Pressley, 1989). In addition, developmental studies of proactive interference demonstrate that children are less proficient than adults at connecting meaning across multiple events. When studying and recalling a series of lists whose members belong to the same semantic category, adults’ recall performance declines on each successive list (Underwood, 1957). However, children do not exhibit similar evidence of meaning-based proactive interference until roughly 11 or 12 years of age (Bjorklund, 1978). These findings suggest that adults are better than children at processing meaning-based connections between multiple events. Such a processing difference could account for the age differences observed in the DRM paradigm. Yet it could be argued that age differences observed in the DRM paradigm result from children failing to make any kind of connection between multiple events rather than a selective deficit in processing meaning-based connections between multiple events. In this respect, a review of the relevant literature demonstrates that children younger than 11 years of age are selectively deficient in processing meaning across items but not in processing surface-level features such as phonology (e.g., Bach and Underwood, 1970 and Hasher and Clifton, 1974). The tendency of children to selectively attend to surface-level features as opposed to meaning-based features of events affects the rate at which they falsely recall critical lures in the DRM paradigm. For example, Brainerd and Reyna (2007) presented 6-, 10-, and 14-year-olds with categorical lists composed of either semantic or phonological associates and later tested their memories with a recognition memory test containing nonpresented associates taken from the studied lists. The standard age effect was observed in those children who had studied categorical lists of semantically related words. The oldest children falsely recognized the nonpresented exemplars from the semantic lists at far greater rates than did the younger children. However, for those children who studied categorical lists of phonologically related words, the youngest children falsely accepted more critical nonpresented exemplars than did the older children. These results provide clear evidence that children from the youngest age group made connections between the phonological associates but failed to make connections between items presented on the semantically related lists of words (for related results, see Dewhurst and Robinson, 2004 and Holliday and Weekes, 2006). Moreover, the semantic processing deficit observed in young children appears to be limited to making meaning-based connections across multiple events as opposed to a more general deficit in extracting meaning from isolated events. In the same set of experiments, Brainerd and Reyna (2007) reported that children from all age groups sampled falsely accepted critical lures at equivalent rates when these critical lures were taken from categorical lists of semantic associates for which only a single associate from each list had been presented at study. As such, it is not simply the case that children are unable to process the meaning of single events, and it is not the case that children cannot make connections across multiple events. Rather, children exhibit a selective deficit in drawing meaning-based connections across multiple events. As demonstrated by a review of the extant data, a considerable amount of research has been conducted using the DRM paradigm to investigate the development of semantic memory in 5- to 9-year-olds. However, considerably less research has investigated DRM performance in older children. When children older than 9 years of age are included in cross-sectional designs, they typically serve as the capstone age group. An inference that could be drawn from the extant data is that the semantic memories of older children function as well as those of young adults. Indeed, a major inference drawn from past research has been that children older than 9 years of age automatically form meaning-based connections across multiple events. However, an often underemphasized observation is that an ability to form meaning-based connections across multiple list items appears to be still developing even in older children. Although children older than 9 years of age provide data demonstrating that they are capable of extracting global forms of gist, the data also suggest that older children are less efficient at doing so than are older adolescents and adults (Brainerd et al., 2006). For instance, 11- and 12-year-olds demonstrate evidence of spontaneously connecting meaning across multiple events by clustering their free recall by category (Bjorklund & Hock, 1982), but these children are still not as efficient at doing so as are older children and young adults (Bjorklund & Jacobs, 1985). Moreover, the vast majority of the extant DRM data has focused on the age at which children appear to behave in a manner suggesting that they have extracted a global gist representation. In contrast, the current research sought to fill a void in the extant literature by selectively attending to the ability of what typically are the oldest participants included in developmental DRM studies and contrasting their performance with that of young adults. Specifically, we were interested in how structurally manipulating the manner in which DRM lists are presented affects the ability of 11-year-olds to extract a global gist representation in comparison with young adults. To do so, we used a manipulation that was subtler than a block versus random presentation manipulation that has been used previously to investigate differences in local and global processing in younger and older adults (Thomas & Sommers, 2005). For the current study, participants viewed DRM list items alongside an accompanying context word, making the DRM lists more akin to traditional paired associative learning tasks (e.g., Cook, Marsh, & Hicks, 2006). For example, each word included on the DRM list for which window is the critical item was presented with an associate at study (e.g., break–shatter, shade–drapes, ledge–balcony, house–home, open–close). The inclusion of associates was intended to provide a context that would influence the gist extracted for each of the list items. The nature of the associates presented with DRM list items was manipulated in such a fashion as to influence the meanings of the DRM list items. In some instances, the associates biased the meanings of list items toward the gist of the list’s critical lure (e.g., break–glass). In other instances, the associates biased the meanings of list items away from the gist of the list’s critical lure (e.g., drinking–glass). In this fashion, we investigated the extent to which such an encoding manipulation influenced false alarm rates of critical lures on a paired associate recognition memory test. For example, would participants indicate having seen window–shade presented at study more so when the study context had biased the meaning of shade toward that of window (e.g., shade–drapes) than when the study context had biased the meaning of shade away from that of window (e.g., shade–tree). Note that this is a subtle manipulation. The meanings of the DRM list items were not changed drastically. However, even modest shifts in the meaning extracted for each list item could facilitate 11-year-olds in extracting a global form of gist, thereby increasing the rate at which they falsely accept critical lures. Such an observed increase would demonstrate that these participants are developmentally lagging behind young adults in their ability to extract global forms of gist. In addition, to validate that 11-year-olds developmentally lag behind adults in the use of global forms of gist, the experiment was designed to allow process measures of memory to be estimated using conjoint recognition. Conjoint recognition is a theoretically motivated multinomial model of memory that provides estimates of memory processes outlined in fuzzy-trace theory. Previous research has demonstrated that conjoint recognition is an effective methodology for providing process measures for children as young as 5 years of age (Brainerd, Holliday, & Reyna, 2004). For conjoint recognition, participants make memory judgments under one of three different instruction sets: verbatim, gist, or verbatim + gist. Under verbatim instructions, individuals are asked to accept only those items that exactly matched items that were presented at study. Under gist instructions, participants are asked to reject items that were presented at study and to accept items that match the gist of what was presented. Under verbatim + gist instructions, participants are asked to accept items that were presented at study in addition to any items that are related to studied items. The observed data can be fit to a multinomial model to provide estimates of verbatim- and gist-based memory judgments. The conjoint recognition model is ideally suited to the current experiment because manipulating the context in which DRM list items are encoded is predicted to affect the ability of children to extract a global form of gist. To the extent that such a manipulation is successful, children are expected to accept more critical lures by making gist-based similarity judgments. Such a pattern of findings would not be predicted in adults because adults are highly proficient at encoding global forms of gist. As such, estimates of similarity judgments will likely be equivalent between the context–toward and context–away conditions for young adults. In contrast to critical lures, related lures are unstudied list items sampled from DRM lists studied previously. Critical lures are inherently related to all list items. However, a related lure, which is just an unstudied DRM list item, is related only to the critical lure and is not inherently related to all of the other list items. Thus, a related lure does not necessarily serve as a good match to the global gist of its corresponding DRM list. As such, a manipulation of study context intended to enhance participants’ ability to process global gist should not affect the rate at which they falsely accept related lures using gist-based similarity judgments. Thus, the study manipulation should not affect similarity judgments for related lures in either children or adults because previous research has demonstrated that both age groups are equivalent in their abilities to process the local forms of gist that are used to make similarity judgments for related lures (e.g., Brainerd & Reyna, 2007). As such, a very specific pattern of process differences is predicted in the current experiment, and the use of a process model such as conjoint recognition will allow any such process differences to be identified. To recap, children are predicted to show higher rates of global gist-based similarity judgments for critical lures in the context–toward condition than in the context–away condition. In contrast, young adults are predicted to demonstrate equivalent rates of global gist-based similarity judgments for critical lures across study conditions. Moreover, the study manipulation is not predicted to affect similarity judgments for related lures in either children or adults.