توسعه حافظه کاذب در کودکان دو زبانه و بزرگسالان

The effects of within- versus between-languages (English–French) study and test on rates of bilingual children’s and adults’ true and false memories were examined. Children aged 6 through 12 and university-aged adults participated in a standard Deese–Roediger–McDermott false memory task using free recall and recognition. Recall results showed that: (1) both true and false memories increased with age, (2) true recall was higher in within- than between-languages conditions for all ages, and (3) there were fewer false memories in between-languages conditions than within-language conditions for the youngest children, no differences for the 8 and 12 years old, and by adulthood, there were more false memories in between-languages than within-language conditions. Recognition results showed that regardless of age, false recognition rates tended to be higher in between-languages than within-language conditions. These findings are discussed in the context of models of false memory development.

Past studies using the Deese–Roediger–McDermott (DRM) paradigm (Deese, 1959 and Roediger and McDermott, 1995) have demonstrated that young children may be less susceptible to false memories than adults (Brainerd et al., 2002, Howe, 2005, Howe, 2006, Howe, in press and Howe et al., 2004; but see Ghetti, Qin, & Goodman, 2002). Theoretical explanations such as fuzzy-trace theory (FTT) have attempted to account for these findings by suggesting that both verbatim and gist traces are encoded during list presentation (Brainerd & Reyna, 2005). These traces are qualitatively different from each other where the former is concerned with item-specific surface information and the latter with meaning-based information. It is this gist trace that is thought to be responsible for false recall in the DRM paradigm. Although young children are capable of extracting these types of traces, the ability to do so improves with age and cognitive development into later childhood and early adulthood (Brainerd & Reyna, 2005). An alternative, associative-activation model (e.g., Howe, 2005, Howe, 2006 and Howe, in press) argues that developmental trends in false memories occur not just because of correlated changes in children’s meaning extraction skills (i.e., growth in knowledge base) but also because of increased automaticity in the activation and accessibility of those concepts and associations in the child’s knowledge base. These increases in automaticity are brought about by additional exposure and proficiency using these concepts as well as the associations between related concepts. As exposure and proficiency increases, so too does the automaticity of activation of concepts and their associative links, making false memory production more and more adult-like, occurring without conscious effort or awareness (Howe, 2005). Although these models have been contrasted in other forums (Howe, 2005, Howe, 2006 and Howe, in press), a key test of these positions can be found by considering the development of bilingual memory. Bilingual children may be able to use meaning more effectively if they can access semantic representations through two different lexical forms. However, this may come at a cost as increased processing of meaning in young children may increase their levels of false recollection. This is exactly what has been found in studies with bilingual adults. Specifically, when adults study a list in one language (e.g., English or Spanish) and are asked to recall or recognize those items in the other language (i.e., English ⇒ Spanish or Spanish ⇒ English), true memory performance is reduced and false memory performance is enhanced relative to within-language study-test conditions (i.e., English ⇒ English or Spanish ⇒ Spanish). For example, Marmolejo, Diliberto-Macaluso, and Altarriba (2003) found that for Spanish–English bilinguals, there was lower true recall and higher false recall in between-languages study and test conditions than in within-language conditions. Similar findings were reported by Wakeford, Carlin, and Toglia (2005) with English–Spanish bilinguals using both recall and recognition measures. Using a 2(Acquisition: English, Spanish) × 2(Test: English, Spanish) design, Wakeford et al. (2005) found that veridical recall and recognition rates were lower in between-languages conditions than within-language conditions. Using a similar design, Sahlin, Harding, and Seamon (2005) found false recognition increased in between-languages as opposed to within-language conditions for English–Spanish bilingual adults. However, as additional study-test trials were provided, participants’ false recognition rates decreased as they came to rely more on language-specific lexical representations rather than on the conceptual representations used initially. Finally, Cabeza and Lennartson (2005), who used a paradigm similar to the previous ones but testing English–French bilingual adults, found that false recognition was robust in both within-language and between-languages conditions, but more correct recognition of “old” items occurred in the within-language than between-languages conditions. All of these findings with bilingual adults have been attributed to increased meaning processing in the between-languages study-test conditions than in the within-language conditions (e.g., Cabeza & Lennartson, 2005). Consistent with this is the finding that as the number of study-test trials increases, participants’ reliance on meaning decreases because stronger cue- or language-specific representations are available, and both of these events are associated with lowered false recollection (Sahlin et al., 2005). Taken together, these findings have been seen by some as support for FTT (Brainerd, Forrest, Karibian, & Reyna, 2006). The argument is that presentation of a list of words in one language (e.g., English) and requiring output (recall, recognition) in a different language (e.g., French, Spanish) makes it (a) less likely that bilingual adults will be able to access verbatim traces to support true memories, (b) less likely that bilingual adults will be able to access verbatim traces to suppress false memories, and (c) more likely that bilingual adults will access gist traces that support false memories. Because of the dual effect of decreases in verbatim memory (a and b), coupled with increases in gist processing (c), false memories are more likely in between-languages than within-language study-test conditions. However, these outcomes with adult participants do not provide necessary and sufficient support for FTT. This is because these results are also consistent with predictions from associative-activation models (e.g., Howe, 2005 and Hutchison and Balota, 2005). Like FTT, these latter models predict that increased meaning processing, along with decreased discriminability of the original stimulus, both circumstances that pertain to between-languages conditions but not within-language conditions, will lead to fewer true memories and more false memories. These theories do, however, make different predictions when we consider the development of false memories in bilingual children. To see how these theories can be distinguished, first consider what FTT predicts about bilingual children’s false memories. If FTT is correct, relative increases and decreases in children’s true and false memories depend on the extent to which between-languages study-test conditions prompt bilingual children to increase gist processing at the expense of verbatim memory. Specifically, according to FTT, the development of children’s false memories is related to their increased ability to extract gist. To the extent that between-languages memory tasks increase gist processing and decrease reliance on verbatim memory, FTT predicts that bilingual children should produce more false memories in between-languages study-test conditions than within-language ones. Of course, whether this task increases children’s gist processing depends on the development and organization of bilinguals’ mental lexicon. According to Silverberg and Samuel (2004), the age of acquisition of an additional language can play an important role in the organization of the mental lexicon such that only early learners (acquisition prior to the age of 7) possess a shared conceptual/semantic store for both languages (L1 and L2) whereas those of proficient late learners remain separate, making conceptual mediation in the second learned language (L2) impossible without translating it into the first learned language (L1). Although their conceptual stores remain unconnected, proficient late learners possess a shared lexical store for both languages where similar lexical forms compete with each other independent of language (Silverberg & Samuel, 2004). Thus regardless of high proficiency in L2, the organization of the late learners’ bilingual memory can never resemble that of an early bilingual. Other models of bilingual development assume either that there is a single conceptual store or if separate stores do exist, they are overlapping. In Kroll and Stewart’s (1994) revised hierarchical model, beginning L2 learners’ access conceptual representations in their nondominant language by initially translating to L1. However, conceptual mediation in L2 is feasible for highly proficient learners but reliance on L2-L1 lexical links to access conceptual meaning remain nonetheless (French & Jacquet, 2004). Kroll and Stewart (1994) demonstrated that even highly fluent bilinguals possess asymmetrical strength between L1–L2 lexical and conceptual links, where conceptual links are stronger for L1 than for L2 and L2–L1 lexical links are stronger than L1–L2 links. Thus, in this model both strength and direction are important factors in bilinguals’ memory performance, where L2 learners’ performance may be affected by the direction of the translation or the strength of the association between L2 and the conceptual store. Although the revised hierarchical model has been supported in bilingual research (McElree et al., 2000 and Talamas et al., 1999), others have reported that even novice L2 learners are capable of conceptual mediation in L2 suggesting that semantic information is stored and accessed even by late, low proficiency L2 learners (Altarriba & Mathis, 1997). These findings may suggest an alternate path in the development and organization of bilingual memory, one in which both lexical and conceptual stores, as well as the connections between them, continue to grow as proficiency and exposure increase. Overall, then, the literature on bilingual development supports the idea that even young second-language learners are capable of increased gist processing given a between-languages task. Thus, to the extent that between-languages memory tasks reduce reliance on verbatim traces and increase gist processing, FTT predicts increases in children’s false memory rates in between-languages than within-language study-test conditions. Such increases may be more likely in older than younger children given older children’s better gist processing more generally. However, to the extent that younger children have greater difficulty spontaneously extracting gist than older children, between-languages study-test conditions might just boost younger children’s gist processing too, leading to larger false memory “gains” for younger than older children. For FTT, then, the predictions are that for false recall: (a) older children and adults will evince more false memories than younger children regardless of language condition and (b) all participants should show higher rates of false memories in between-languages than within-language conditions. For true recall, because within-language conditions preserve verbatim information, true recall rates should be higher in these conditions than the between-languages conditions for all participants regardless of age. As mentioned earlier, Howe, 2005, Howe, 2006 and Howe, in press has pointed out that increases in children’s false memory with age may not be due simply to increases in processing meaning (gist) but rather to increases in the automaticity with which children access or activate associations in their knowledge base, associations that mediate false remembering (for a similar model of adults false recollection, see Hutchison & Balota, 2005). Automaticity, particularly in children’s recall, may play a role in the rate of between-languages false memories where highly proficient bilinguals may be more skilled in accessing concepts, as well as the associative and categorical relations among them, in both languages (Kotz & Elston-Guttler, 2004). Dufour and Kroll (1995) demonstrated that more fluent bilinguals are equally rapid when categorizing words within and between languages as compared to less fluent bilinguals who demonstrated different categorization speeds depending on the target’s language. This has also been confirmed by McElree et al. (2000) who examined the speed of processing in balanced and unbalanced bilinguals and determined that access to conceptual information in both languages was highly dependent on exposure and proficiency. This increase in speed of translation may affect rates of true and false memories in the DRM task, where distinctive information may be lost at the cost of L2 automaticity, resulting in more false memories in conditions that require accessing the concept through two distinct forms (dual processing) for bilinguals equally proficient in both languages. Although FTT and the associative-activation accounts frequently make similar predictions, they do so for different reasons. For example, both models agree that true recall should be better in within-language than between-languages conditions and that both true and false recall rates should increase with age. FTT makes these predictions because of changes in the relative involvement of gist and verbatim memories and the associative-activation account predicts this because of increases in the automaticity of access to memory traces and their interitem links. However, these accounts do diverge when making predictions about the development of false memories in bilingual processing tasks. Unlike FTT, the associative-activation account requires not only that the conceptual structures associated with both languages be present in a participant’s knowledge base, but also that the processes supporting the activation of these concepts and their interitem associations be automatic. As Howe (2005) showed, despite the fact that concepts used in DRM tasks are present in young children’s knowledge bases, their access to them (as well as interitem links) is not as automatic as that found in adults. Because higher rates of false memories are associated with increased automaticity of activation (e.g., Gallo and Seamon, 2004, Raaijmakers and Zeelenberg, 2004 and Roediger et al., 2002), even though bilingual concepts exist in children’s knowledge base their mere existence does not guarantee increases in false memories. It is only when access is automatic that increases in false memories will occur. Indeed, as already seen, this automaticity in access to bilingual memory comes about with additional exposure and proficiency in both languages (McElree et al., 2000). Because this automaticity is more likely in adults than young children, the associative-activation account, in contrast to FTT, predicts more false memories in between-languages than within-language conditions for adults but not young children. That is, while FTT predicts higher false recall for between-languages than within-language conditions regardless of age, the associative-activation account predicts a cross-over interaction with age. In the current study, we tested these different predictions by examining true and false memories in bilingual individuals varying in age (and, hence, bilingual exposure and proficiency) using a bilingual DRM task. As noted, both FTT and the associative-activation accounts predict true recall should be better in the within-language than the between-languages conditions and that true and false recall rates should increase with age. Where they differ is with false recall. According to FTT, bilingual individuals should have more false memories in the between-languages conditions (when translation is required) than in the within-language conditions. FTT also predicts increases in children’s false recollection rates when gist processing is made easier and these increases should be more evident with children whose gist processing is particularly compromised (i.e., the youngest children). Alternatively, associative-activation models predict that increases in false recollection occur depending on the relative automaticity of the associative connections underlying false memories and given that automaticity is correlated with exposure and proficiency, between-languages false memories should exceed within-language false memories more frequently in older, more proficient bilinguals. That is, age-related differences in automaticity may mean that only older, adult participants should exhibit higher rates of false memories than younger participants even though conceptual-level representations for both languages overlap very early in bilingual development. In addition to recall, we also examined bilingual participants’ recognition performance. These recognition tests provide a validity check inasmuch as they assess (a) whether children and adults have access to the concepts used in the DRM task in both languages, (b) that this access is similar across languages both within age groups as well as across age groups, and (c) whether children and adults have access to the originally presented items in the language in which they were originally presented. Of course, because recognition performance depends more on matching the probe with the same representation in memory, all theories would predict that within-language true recognition rates should exceed between-languages rates as the item was experienced twice in the same language in the former but not the latter condition. More importantly, both FTT and the associative-activation accounts have different predictions when it comes to recognition. According to FTT, because recognition tests provide explicit retrieval cues for verbatim traces that are not found in recall tests, they are more likely to tap verbatim traces of targets than free recall tasks where cues, if any, are generated internally and emphasize gist processing (see Brainerd & Reyna, 2005). What this means is that true recognition should be higher and false recognition lower in within-language conditions. For between-languages conditions, because participants have experienced concepts in both languages (one at study and the other at test), recognition cues, regardless of language, should tap at least one of the verbatim traces. This could mean either that true recognition should again exceed false recognition performance or that true and false recognition should be approximately equal depending on the extent to which verbatim processing has been compromised given the greater likelihood of gist processing in between-language conditions. Finally, true recognition should be higher and false recognition rates lower in the within-language conditions than in the between-languages conditions because the same verbatim trace has been presented twice in the former but not the latter conditions. This item type (true vesus false) × condition (within-language vesus between-languages) interaction is similar to the findings mentioned earlier in which additional within-language repetition for English–Spanish bilinguals decreased false recognition rates (Sahlin et al., 2005). In contrast, for the associative-activation model, although the cues for memory performance are more explicit in recognition tests, performance should resemble that found at recall. That is, if bilingual children and adults are accessing overlapping conceptual representations (Altarriba & Mathis, 1997) in between-languages conditions but language cues are no longer diagnostic (i.e., they do not provide distinctive information that can be used to discriminate presented from unpresented items in memory), then there should be more confusion as to what was and what was not presented in the between-languages than within-language conditions. What this means is that false recognition should be higher and true recognition lower in between-languages conditions than within-language conditions. Moreover, although true and false recognition rates may not differ in within-language conditions, false recognition rates should be higher than true recognition rates in between-languages conditions.