In this study, we examined mechanisms that underlie free-recall performance in bilinguals’ first language (L1) and second language (L2) through the prism of serial-position effects. On free-recall tasks, a typical pattern of performance follows a U-shaped serial-position curve, where items from the beginning of the list (the primacy effect) and items from the end of the list (the recency effect) are recalled with higher accuracy than items from the middle of the list. The present study contrasted serial-position effects on the free-recall task in Korean-English bilinguals’ L1 vs. L2 and examined the relationship between an independent working memory (WM) measure and serial-position effects in bilinguals’ two languages. Results revealed stronger pre-recency (primacy and middle) effects in L1 than in L2, but similar recency effects in the two languages. A close association was observed between WM and recall performance in the pre-recency region in the L1 but not in the L2. Together, these findings suggest that linguistic knowledge constrains free-recall performance in bilinguals, but only in the pre-recency region. Keywords: bilingualism, free recall, primacy, recency Short-term memory (STM) enables us to encode and retain information for a short period of time. The structure and the function of verbal short-term memory (STM) has been the focus of active and productive research over the past century (e.g., Baddeley, 2009; Gathercole, Willis, Emslie, & Baddeley, 1992; Baddeley, Thompson, & Buchanan, 1975; Cowan, et al., 1992; Miller, 1956). However, the vast majority of this work has focused on monolingual speakers. Although a number of studies have examined the role of verbal memory in second language acquisition (Baddeley, Gathercole, & Papagno, 1998; Cheung, 1996; Kormos, & Sáfár, 2008), relatively little is known about the actual underlying mechanisms of verbal short-term memory function in bilingual individuals. For example, one common measure of STM is a free word-recall task where participants listen to lists of words and are asked to recall as many of the words as possible in any order (e.g., Cowan et al., 1992). A classic finding is that performance on the free-recall task is characterized by a unique U-shaped serial-position curve, where items from the beginning of the list (the primacy effect) and items from the end of the list (the recency effect) are recalled with higher accuracy than items from the middle of the list (e.g., Davelaar et al., 2005; Glanzer & Cunitz, 1966). However, no previous study has examined whether such classic U-shaped effects can be obtained in bilingual speakers, and whether these effects are rooted in the same fundamental mechanisms when examined in bilinguals’ native vs. second language. A traditional explanation of the U-shaped serial-position curve observed when performance on the free-recall task is plotted as a function of the location of the item on the list is the dual-component model of recall (e.g., Atkinson & Shiffrin, 1968; Glanzer, 1972; Glanzer & Cunitz, 1966). The dual-component model was proposed to suggest that two memory systems, short-term memory (STM) and long-term memory (LTM), were involved in performance on the free-recall task. The primacy effects on the free-recall task were interpreted to reflect the involvement of the LTM system in free-recall performance; that is, successful recall of the first few items on the list was attributed to the participants’ ability to rehearse these items and to transfer them into the LTM (e.g., Rundus, 1971). The recency effects on the free-recall task were interpreted to reflect the involvement of the STM in free-recall performance; that is, successful recall of the last few items on the list was attributed to the participants’ ability to store these items temporarily in the STM rather than in the LTM (e.g., Waugh & Norman, 1965). Support for the dual nature of the free-recall task came from classic experiments where variables known to influence the LTM vs. the STM were pitted against each other (Craik & Levy, 1970; Glanzer & Cunitz, 1966; Murdock, 1962; Raymond, 1969; Shallice, 1975). For instance, Glanzer and Cunitz (1966) demonstrated that a faster presentation rate reduced recall performance for items in the pre-recency position (primacy and middle regions) but did not affect recall of items in the recency position of the serial-position curve. Conversely, the increased delay between the end of the list and recall reduced recall performance for items in the recency position while it did not affect the recall of items in the pre-recency position. Many subsequent studies supported the dissociation between the STM and the LTM in modulating performance on free-recall tasks by showing that variables known to influence LTM function, such as list length (e.g., Murdock, 1962), word frequency (e.g., Raymond, 1969) and semantic similarity (e.g., Craik & Levy, 1970) affected recall performance of items in the primacy position while variables known to influence STM function, such as phonological similarity (e.g., Shallice, 1975) affected recall performance of items in the recency position. More recently, the dual nature of the memory mechanisms supporting free-recall performance was instantiated in the context-activation model (e.g., Davelaar et al., 2005). This model suggests that two memory components are involved in free-recall performance: One is a changing context/episodic long-term memory system and the other is an activation-based short-term buffer. However, a number of studies challenged this STM-LTM dissociation model for explaining the U-shaped effects in free-recall data (e.g., Baddeley & Hitch, 1974; Glenberg, 1984; Pinto & Baddeley, 1991). For instance, Pinto and Baddeley (1991) observed recency effects in their free-recall data when recall was tested after a significant delay – a finding that is at odds with the idea that recency effects arise within the STM system. To solve these problematic findings associated with recency data, unitary-component approaches of recall performance have been formulated. The basic premise of the unitary approaches is that the probability of recalling an item is proportional to the ratio of the inter-presentation interval (the interval between items) to the retention interval (the interval between the item and its recall). Supporting this unitary component view, several alternative models have been suggested. For instance, the Temporal Context Model (TCM; Howard & Kahana, 1999) posits that words are associated with mathematically formulated temporal contexts, and words from the end of the list are recalled better because the context vector (in the just–presented item’s representation) is used as a retrieval cue, with the ensuing recall advantage for the associated nearby items. Alternatively, the Scale-Invariant Memory, Perception, and Learning model (SIMPLE; Brown, Neath, & Chater, 2007) posits that items that are temporarily distinctive are advantaged at recall, and therefore, the last few items are recalled better because there are fewer interfering neighbors in the end of a list. Yet, just like dual-component models, single-component models do not fully explain the U-shaped serial-position curve associated with free-recall performance (for instance, the ratio rule cannot explain primacy effects) and the debates between the proponents of single-component models (e.g., Glenberg et al., 1980; Neath & Crowder, 1990) and dual-component models (e.g., Atkinson & Shiffrin, 1968; Davelaar et al., 2005; Glanzer, 1972; Gillund & Shiffrin, 1984) of free-recall performance continue. The goal of the present study was not to differentiate between single- and dual-component models of free recall. Instead, we took the dual-component model as our starting theoretical point, because dual-component models explicitly posit influences from the LTM and the STM on free recall. This then makes them suitable to examining free recall performance in bilinguals, for whom the LTM and the STM influences on free recall may be more dissociable than in monolinguals. In cognitively-intact, adult speakers of a single language, the LTM and the STM systems function in tandem to support recall, such that is often difficult to dissociate performance patterns associated with LTM vs. STM. Examining free-recall performance in bilinguals may be useful not only because such examinations are woefully lacking in the literature, but also because LTM vs. STM influences on verbal recall may be more dissociable in bilingual speakers, especially when their performance in the native language is contrasted with their performance in the second language. There is reason to hypothesize that serial-position effects on a free-recall task may be different in bilingual speakers’ native vs. second language. Not surprisingly, a number of previous studies (e.g., Bialystok & Feng, 2009; Portocarrero, Burright, & Donovick, 2007) revealed that bilinguals tend to perform more successfully on language-based tasks in their native language (L1) vs. the second language (L2). Further, bilinguals also tend to perform better on verbal STM tasks when these are administered in their L1 vs. their L2 (e.g., Thorn & Gathercole, 1999). Such findings are interpreted as evidence for the link between long-term linguistic knowledge and STM function: That is, the more robust long-term linguistic knowledge associated with the L1 (vs. the L2) helps maintain the representations of L1 items (more so than of L2 items) presented on the STM task. In the present study, we capitalized on prior literature indicating a link between bilinguals’ LTM system and STM performance to examine the mechanisms that may underlie free recall in bilingual speakers. Specifically, we examined serial-position effects in L1 vs. L2 free-recall performance in late sequential bilingual speakers of Korean and English, who acquired their stronger native language, Korean, at birth, and who acquired their weaker second language, English, in their teens. The underlying hypothesis for this work was that if the LTM system is involved in generating pre-recency effects on the free-recall task (in line with dual-component models), then pre-recency effects should be more robust in bilinguals’ L1 – i.e., in the language that is stronger and more proficient – than in bilinguals’ L2. In order to even more precisely delineate the mechanisms that may underlie serial-position effects in bilinguals’ free-recall performance, we also examined the associations between the pre-recency effect and the recency effect on the one hand, and an independent measure of working memory (WM) on the other hand, in bilinguals’ L1 and L2. WM, unlike STM, involves the ability to store information for a brief period of time in the service of a cognitively-demanding task. A number of studies have documented a link between performance on free-recall tasks and performance on WM tasks (e.g., Unsworth, Brewer, & Spillers, 2011; Unsworth & Engle, 2007). For instance, Unsworth, Brewer, and Spillers (2011) found that individuals with high WM capacity utilized better strategic encoding processes (e.g., rehearsal, grouping, etc.) and more effective retrieval plans than individuals with low WM capacity on a cued recall task. In addition, Guida et al. (2013) found an interaction between serial-position effects on a free-recall task (pre-recency and recency) and WM span (high WM group and low WM group). Specifically, the difference in recall performance between the high WM group and the low WM group was present for the items in the pre-recency region (where presumably WM capacity could modulate LTM processes), but was reduced and non-significant for the items in the recency region (where presumably LTM processes could not modulate recall performance). In the present study, we relied on the association between WM capacity and LTM influences on pre-recency data established by previous studies to further test the possibility that LTM differently contributes to free-recall performance in the L1 vs. the L2. We hypothesized that the relationship between WM and recall performance in the pre-recency region would be stronger in the L1 than the L2 because bilinguals’ LTM should be more stable in the L1 than the L2. In summary, delineating the mechanisms of U-shaped effects in free-recall performance in terms of LTM vs. STM involvement has proven difficult due to the integrated nature of the two memory systems in monolingual speakers. Examining serial position effects in bilingual speakers’ native vs. second language may yield a dissociation between the two memory systems. The goal of the present study was to examine free-recall performance in bilinguals’ L1 vs. L2 as a function of serial-position effects. Twenty Korean-English bilinguals (Mean Age = 29.7, SD = 4.9; Mean Years of Ed = 19.8, SD = 3.7; Mean Non-Verbal IQ = 115.8, SD = 11.1) were recruited from the University of Wisconsin-Madison. All participants spoke Korean as their native language and acquired English as their second language in early teen years, with a mean acquisition age of 11.3 years (SD = 2.8). Detailed data were obtained regarding the participants’ length of residence in Korea and the United States, the contexts in which they acquired Korean and English, and the contexts in which they were exposed to Korean and English at the time of the study levels (as reported on the Language Experience and Proficiency Questionnaire, LEAP-Q; Marian, Blumenfeld, & Kaushanskaya, 2007). Participants also completed standardized tests of English vocabulary knowledge (Peabody Picture Vocabulary Test – IV Mean = 88.5, SD = 10.4; Expressive Vocabulary Test Mean = 97.1, SD = 13.2). Examination of self-reported data indicated that our participants were late, sequential bilingual speakers, whose self-reported language skills in Korean were more robust than their self-reported language skills in English. However, participants’ knowledge of English was quite robust, as indicated both by their self-reports and by their performance on the standardized measures of English vocabulary knowledge. Specifically, self-reported speaking proficiency levels in English ranged between adequate and good, and scores on the vocabulary measures placed the participants firmly in the average performance range. See Table 1 for the participants’ background characteristics. Demographic Characteristics of Bilinguals
To measure participants’ WM capacity, complex NWR tasks in both English and Korean were constructed, where participants were asked to repeat nonwords, while also completing a secondary animacy judgment task. For the English WM-NWR task, 48 English nonwords were selected from Gupta et al. (2004) corpus. Of these, 16 were 2-syllable nonwords, 16 were 4-syllable nonwords, and 16 were 6-syllable nonwords. For the Korean WM-NWR task, 48 Korean nonwords were created, following phonologically plausible syllables in Korean (Lee, 2006). For a secondary judgment task, 48 nouns were selected in each language, with half of the nouns being animate and half of the nouns being inanimate. The nonword stimuli followed language-specific phonotactics, such that English nonwords followed English phonotactics while Korean nonwords followed Korean phonotactics. The nonwords across two languages were matched on acoustic duration. One hundred and thirty five English nouns were selected. The nouns were 1, 2, and 3 syllables in length, grouped into lists of 10, 15, and 20 words. There were three 10-word lists, three 15-word lists, and three 20-word lists for each syllable-length. Across lists and syllables, nouns were matched on lexical frequency (Brysbaert & New, 2009) and on concreteness (MRC Psycholinguistic Database). One hundred and thirty five Korean nouns were selected and grouped based on the same criteria as the English nouns. Across lists and syllables, Korean nouns were matched on lexical frequency (the National Institute of the Korean Language’s 2003 Modern Korean Usage Frequency Survey) and on concreteness (based on their English translations). See Appendixes 1 and 2 for English and Korean words lists. The English nouns were recorded by a female native speaker of English while the Korean nouns were recorded by a female native speaker of Korean. Task parameters were matched for the English and the Korean word-recall tasks, and speed of presentation was controlled across list lengths within each syllable length, with the average speed of 1.1 words/sec. All participants completed the free-recall tasks and the WM-NWR tasks in both languages during different sessions, and the procedure for each task was identical for both languages. Participants completed the Korean tasks first and completed the English tasks a week later. This order of sessions was fixed because the English tasks were more taxing for the participants than the Korean tasks. In order to avoid attrition, it was decided to maintain the order across participants, with Korean always being the language of the first session. The instructions were always administered in Korean (across both sessions) to ensure understanding. For the WM-NWR task, the 2-syllable nonwords were presented first, followed by 4-syllable nonwords, and 6-syllable nonwords. The order of nonwords at each syllable length was randomized for each participant. Each participant heard a nonword first, followed by the noun (animate/ inanimate), and was asked to judge the animacy of each noun by pressing “/” for an animate noun, and “z” for an inanimate noun. The participant then repeated the nonword as accurately as possible after a cue. The time between the presentation of the nonword and the cue to repeat it was set to 4000 msec. Each participant’s productions were recorded and coded off-line. For the free-recall task, each participant heard 1-syllable words in 10-word, 15-word, and 20-word lists; followed by 2-syllable words in 10-word, 15-word, and 20-word lists; etc. The order of words in each list was randomized for each participant. After listening to each list of words, participants were cued to recall as many words as possible regardless of their order. Each participant’s productions were recorded and coded off-line. All participants were administered two English vocabulary measures, the Peabody Picture Vocabulary Test (PPVT-III, Dunn & Dunn, 1997) and the Expressive Vocabulary Test (EVT, William, 1997), as well as the non-verbal IQ measure, the Visual Matrixes subtest of the Kaufman Brief Intelligence Test (KBIT-2, Kaufman, & Kaufman, 2004). Finally, all participants were asked to complete the LEAP-Q (Marian, Blumenfeld, & Kaushanskaya, 2007) to elicit data regarding their language background. Coding was done by a native speaker of English for English tasks and a native speaker of Korean for Korean tasks. For the WM-NWR task, proportion correct score was obtained for each nonword by calculating the proportion of correctly recalled phonemes out of total number of phonemes per nonword. Phonemes correct rather than syllables or nonwords correct was used because the nonword repetition task was quite difficult, and more global measures of accuracy may have underestimated participants’ levels of performance. A similar approach to coding nonword repetition data was used by other studies (e.g., Dollaghan & Campbell, 1998; Moore, Tompkins, & Dollaghan, 2010). Data were collapsed across syllable lengths to obtain a measure of verbal working memory capacity in each language. For the free-recall task, all productions were transcribed word-for-word and coded for correctness. Omissions, semantic associates, and duplications were scored as incorrect. Initially, each list was divided into three regions: Primacy, middle, and recency. The visual representation of bilinguals’ free-recall performance in the L1 vs. the L2 can be found in Figure 1. Performance in each language clearly followed the traditional U-shaped serial-position curve. However, in order to boost power and to reduce the number of follow-up comparisons, primacy and middle region data were collapsed to yield pre-recency region data. This approach is in line with prior studies where pre-recency and recency effects (rather than primacy and recency effects) were contrasted (Guida et al., 2013). Each list was thus divided into the pre-recency region and the recency region. Specifically, for 10-word lists, the first 7 words were denoted as the pre-recency region, and the last 3 words were denoted as the recency region. For 15-word lists, the first 10 words were denoted as the pre-recency region and the last 5 words were denoted as the recency region. Finally, for 20-word lists, the first 13 words were denoted as the pre-recency region and the last 7 words were denoted as the recency region. The proportion correct scores were obtained for each region in each list. Data were collapsed across lists and syllable lengths to boost power. Three types of analyses were conducted. First, to compare WM capacity between bilinguals’ two languages, a paired-samples t-test comparing bilinguals’ performance in English vs. Korean on the WM-NWR task was conducted. Second, to examine whether there was an effect of language (L1/L2), position (pre-recency/recency), and list length (10/15/20-word list) on free-recall performance, a 2×2×3 repeated-measures ANOVA was conducted. Lastly, correlation analyses were conducted to test the relationship between WM capacity and free-recall performance in each position, within each language. A comparison between bilinguals’ performance on the WM-NWR task in English vs. Korean revealed a significant difference between bilinguals’ two languages, t (19) = 9.32, p < 0.001); participants performed better on this WM-NWR task in their L1 (M = 0.89, SD = 0.04) than their L2 (M = 0.76, SD = 0.08). Similarly, for the secondary animacy judgment task in English vs. Korean, participants performed better in their L1 (M = 0.96, SD = 0.04) than in their L2 (M = 0.84, SD = 0.08), t (19) = 6.94, p < 0.001). The repeated-measures ANOVA yielded a main effect of position, F (1, 19) = 65.01, p < 0.001, ηp2 = 0.77, a main effect of list length, F (1, 19) = 145.03, p < 0.001, ηp2 = 0.88, a 2-way interaction between language and position, F (1, 19) = 11.74, p < 0.001, ηp2 = 0.38, and a 3-way interaction between language, position, and list, F (2, 38) = 10.69, p < 0.001, ηp2 = 0.36. Participants recalled more words in the recency regions (M = 0.56, SE = 0.02) than in the pre-recency region (M = 0.28, SE = 0.02), p < 0.001, and more words in 10-word list (M = 0.53, SE = 0.01) than 15-word lists (M = 0.40, SE = 0.01), p < 0.001 and in 15-word lists than in 20-word lists (M = 0.33, SE = 0.01), p < 0.001. To identify the locus of the interaction between position and language, follow-up comparisons were conducted. Pair-wise comparisons between positions (collapsed across list lengths) within each language revealed that there were significant differences in recall performance between the pre-recency region and the recency region both in the L1, t (19) = − 5.36, p < 0.001 and in the L2, t (19) = − 8.70, p < 0.001. Pair-wise comparisons between each language for each position revealed that bilinguals recalled significantly more words in the pre-recency region (t (19) = 4.58, p < 0.001) when they used their L1 (M = 0.33, SD = 0.09) than when they used their L2 (M = 0.24, SD = 0.10). However, there was no difference between L1 (M = 0.53, SD = 0.11) and L2 recall (M = 0.58, SD = 0.12) for the items in the recency region (t (19) = − 1.37, p = 0.19). See Figure 2 for the visual representation of these data. Follow-up comparisons for the 3-way interaction examined differences between position effects (pre-recency vs. recency) at each list length for each language, as well as differences between languages (L1 vs. L2) at each position for each list length. Paired sample t-tests between pre-recency and recency data at each list length for each language revealed that in the L1, there was no significant difference in recall performance between pre-recency and recency regions for 10-word lists, t (19) = − 1.23, p = 0.23. However, bilinguals recalled significantly more words in the recency region than in pre-recency region for 15-word lists, t (19) = − 4.77, p < 0.001, and 20-word lists, t (19) = − 9.59, p < 0.001. In the L2, bilinguals recalled significantly more words in the recency region than in pre-recency region across all list lengths, including 10-word lists, t (19) = − 6.46, p < 0.001, 15-word lists, t (19) = − 7.47, p < 0.001, and 20-word lists, t (19) = − 6.42, p < 0.001. Paired sample t-tests between L1 and L2 at each position for each list length (see Figure 3) revealed that for 10-word lists, bilinguals recalled more words in the L1 than in the L2 in pre-recency region, t (19) = 7.41, p < 0.001, but not the recency region, t (19) = − 1.82, p = 0.09. Similarly, for 15-word lists, bilinguals recalled more words in the L1 than in the L2 in pre-recency region, t (19) = 2.07, p = 0.05, but not the recency region, t (19) = − 1.70, p = 0.11. However, for 20-word lists, bilinguals recalled more words in the L1 than in the L2 in the recency region, t (19) = 2.43, p < 0.05, but there was no difference in recall performance between the L1 and the L2 in pre-recency region, t (19) = 0.25, p = 0.80. In the L1, there was a significant relationship between WM capacity and free recall in the pre-recency region (r = 0.53, p < 0.05), such that higher WM capacity was associated with higher recall of words in the pre-recency region. However, there was no relationship between WM capacity and free recall in the recency region (r = −0.18, p = 0.46). In the L2, there was no relationship between WM capacity and free-recall performance either in the pre-recency region (r = −0.06, p = 0.81) or the recency region (r = 0.37, p = 0.11). The purpose of the current study was to examine whether (1) bilinguals’ encoding and retrieval patterns on the free-recall task would be shaped by different levels of linguistic knowledge, as instantiated by the differences between bilinguals’ L1 and L2, and (2) whether recall performance from the different regions of the lists would be differentially related to WM capacity. We found that some aspects of free-recall performance in bilinguals are similar across the L1 and the L2. Bilinguals’ performance on a free-recall task followed the typical U-shaped serial-position curve both in the L1 and the L2. Furthermore, bilinguals showed better performance when recalling items from the recency regions than from pre-recency regions both in the L1 and the L2, suggesting that in general, bilinguals in this study adopted a global recency recall strategy. Finally, bilinguals demonstrated list-length effects in both the L1 and the L2, such that free recall of shorter lists was more successful than of longer lists (consistent with Baddeley et al., 1975; LaPointe & Engle, 1990; Lovatt, Avons, & Masterson, 2000). However, we also observed distinct patterns of free-recall performance in the L1 and the L2, suggesting an effect of language experience on the mechanisms that underlie free-recall performance in bilinguals. Bilinguals’ performance on the free-recall task was characterized by a stronger pre-recency effect in the L1 than in the L2. Bilinguals in our study were late, sequential bilinguals with stronger language skills (as indexed by broad self-ratings of proficiency speaking, understanding, and reading) in their L1 (Korean) than in their L2 (English). According to the dual-component model of U-shaped free-recall performance, pre-recency effects are considered to be rooted in the LTM system (e.g., Murdock, 1962; Raymond, 1969), and stronger pre-recency effects in the L1 than in the L2 observed in the present study are consistent with this view. That is, it appears that bilinguals were better able to take advantage of their linguistic knowledge (LTM) to scaffold free recall in their L1. Since bilinguals tested here were more proficient in their L1 than their L2, it is likely that they were able to rehearse the first items on the list more effectively when the task was conducted in the L1, their more proficient language, than in the L2, their less proficient language. This finding is also in line with the more recent context-activation model of free recall (e.g., Davelaar et al., 2005) that construes primacy effects to be the result of an interaction between a dynamic activation buffer that stores phonological and semantic information and selective updating supported by the attentional control system. Therefore, bilinguals’ stronger pre-recency effects in the L1 may be interpreted to suggest that bilinguals can selectively update and activate items from the beginning of the list in a buffer system more effectively in the L1 than in the L2 because of the more robust semantic and phonological knowledge associated with the native language. In contrast to the language effects in the pre-recency region, bilinguals in the present study recalled items from the recency regions equally well in the L1 and the L2. That is, bilinguals were able to maintain and retrieve items from the ends of the lists in their weaker language as efficiently as they did in their native language. However, this finding is qualified by the fact that we observed a three-way interaction among language, list length, and position effects. This interaction appears to be driven by two patterns of results. First, the position effect (i.e., better recall in the recency than the pre-recency region) was observed across all list lengths in the L2 but only for the longer lists (15- and 20-word lists) in the L1. Previous studies have also reported list-length effects in serial position curves, such that primacy effects are dominant for shorter lists (i.e., 3–4 word lists) while recency effects are dominant for longer lists (e.g., Ward, Tan, & Grenfell-Essam, 2010). The broad recency effects observed in our data across list lengths is likely due to the fact that in our study, shorter lists (10-word lists) may in fact exceed the length of lists that yields dominant primacy effects. The finding that the recency effect was not significant for 10-word lists in the L1 indicates that 10-word lists operated as “shorter lists” in the bilinguals’ native language and as “longer lists” in the bilinguals’ second language. Second, the native-language recall advantage was observed in the pre-recency region for shorter lists (10- and 15-word lists) but in the recency region for the longest list (20-word list). This therefore accounts for the broad finding of L1 advantages in the pre-recency but not the recency region, since the recency effect in the L1 for 20-word lists was washed out by non-significant recency effects in the L2 for 10- and 15-word lists. List-length effects in recall tasks may reflect involvement of different processes, such that performance on longer lists shares a great deal of variance with performance on complex span task, while performance on shorter lists does not (Unsworth & Engle, 2006). That is, recall of longer lists (but not shorter lists) likely involves reliance on focus-of-attention mechanisms (Cowan, 2001). We observed an L1 recall advantage for shorter list lengths in the pre-recency region likely because the robust L1 knowledge in the LTM enabled participants to rehearse the first few items on these shorter lists more efficiently in the L1. Conversely, we observed an L1 recall advantage for longer lists in the recency region likely because longer lists demanded recruitment of additional attentional resources, and these resources were utilized more efficiently in the L1 than in the L2. An alternative (or an additional) possibility is that redintegration processes acted upon recall performance for the longer lists but not the shorter lists, with more effective redintegration occurring in the L1. Redintegration is posited to underlie the ability to reconstruct incomplete language representation from the memory trace when retrieving verbal materials from the STM (e.g., Hulme et al., 1997; Schweickert, 1993).When the temporal capacity of the STM exceeds the limit for longer lists, the items stored in the LTM are replaced by newly entered items from the STM. It may be that the redintegration process is more active for longer lists in the L1 than in the L2 because the LTM is more robust and stable in the L1. The reconstruction process in the L1 may therefore be more efficient than in L2, resulting in L1 recall advantages for longer list. While it is difficult to pinpoint the exact mechanisms that underlie the interactions among language, list-length, and serial-position effects in our data, it is clear that serial recall in bilinguals is a highly dynamic system, characterized by complex relationships between linguistic knowledge and basic parameters of free recall. Our findings both converge and diverge with the previous data regarding LTM influences on serial recall, likely because previous studies have not dissociated pre-recency and recency effects (Thorn & Gathercole, 1999; Thorn, Gathercole, & Frankish, 2002). The finding that there were clear L1 recall advantages in the pre-recency region converge with previous studies of L1 advantages, and reiterate the involvement of the LTM in recall performance, with stronger L1 facilitating recall. The finding that there were no L1 advantages in the recency region diverge from previous studies of L1 advantages, and indicate that other cognitive processes (i.e., the focus of attention) can overtake L1 influences on STM, especially for longer list lengths (Unsworth & Engle, 2006). By considering the STM system in a more nuanced way, and by zeroing in on serial-position effects in free-recall performance, here we observe that pre-recency effects in free recall (which are likely influenced by the LTM system more than recency effects) are constrained by linguistic knowledge. Correlation analyses used to examine the relationship between recall performance and WM capacity revealed that cognitive processes that underlie bilingual recall may differ across L1 and L2, especially for sequential bilinguals whose L2 was acquired later in life. The positive correlation obtained between WM capacity and pre-recency effects is consistent with previous studies suggesting that high WM capacity is linked to better strategic memory retrieval (e.g., Unsworth, Brewer, and Spillers, 2011), especially of the items from the pre-recency regions (e.g., Guida et al., 2013). However, finding such a relationship only in the L1 but not in the L2 indicates that the ability to draw upon the WM system to support free recall is less viable in the context of a relatively weak linguistic knowledge base associated with the L2. One caveat to this interpretation is that the current study cannot dissociate the effects of L1 vs. L2 on free recall from language-specific effects. That is, it may be that these findings characterize free recall in Korean vs. English rather than free recall in the native vs. the second language. In an effort to inform this issue, we analyzed pilot data collected for this study from 19 monolingual speakers of English. Because the goal of the study was not to compare monolingual and bilingual free-recall performance, this group of monolingual participants did not match the bilingual group in demographic characteristics (Mean Age = 24.73; Mean Years of Education = 15.89; Mean Non-Verbal IQ = 105.87). However, we conducted correlation analyses between the monolinguals’ WM data and free-recall data in order to examine whether the pattern would be similar to the pattern observed for bilinguals’ L1 (Korean) or L2 (English). The logic was that if the pattern of results observed for the bilingual participants is reflective of native- vs. second-language dynamics, the monolingual data would resemble bilingual L1 data. Conversely, if the pattern of results observed for bilingual participants is reflective of Korean vs. English differences, the monolingual data would resemble bilingual L2 data. When correlation analyses were conducted between monolinguals’ performance on the WM-NWR task and the free-recall task, we found a significant relationship between WM capacity and free recall in the pre-recency region (r = 0.5, p < 0.05), but not in the recency region (r = − 0.23, p = 0.34). These findings mirror precisely the pattern of results observed for the Korean data in bilinguals. This indicates that our findings for Korean-English bilinguals reflect native-language vs. second-language dynamics characterizing free recall, rather than the language-specific factors associated with Korean vs. English. The present findings have implications for models of free recall, whose interpretation and validation have proven difficult because of the integrated nature of memory systems in cognitively-intact monolingual speakers. By investigating bilinguals’ performance on the free-recall task, we can begin to disentangle the (possibly) distinctive contributions of the LTM and the STM to free recall because bilinguals, especially bilinguals like those studied here, have distinct levels of knowledge (i.e., LTM) associated with their native language vs. their second language. Although single-component models (e.g., Baddelely, 1986; Neath & Crowder, 1990) can readily account for comparable recency effects in L1 and L2 free-recall performance, they cannot easily account for the distinctive patterns of pre-recency effects in bilinguals’ two languages observed here. The goal for the future studies would be to examine whether the pattern of results for L1 vs. L2 observed here is specific to late, sequential bilinguals, with more proficient levels of native-language vs. second-language knowledge, or whether it would generalize to other bilingual populations, including those who acquired their two languages simultaneous, and those with balanced levels of L1/L2 proficiency. Further, future studies should incorporate objective measures of linguistic knowledge in bilinguals’ two languages, and include tests of specific linguistic domains (e.g., phonology, vocabulary, syntax) in order to delineate which aspects of linguistic knowledge constrain STM performance. In the current study, the administration of standardized vocabulary measures in English served to establish fairly strong levels of English vocabulary skills in our bilingual participants. However, it would have been useful to also gain an understanding of how participants’ English vocabulary skills compare with their Korean vocabulary skills to corroborate the self-reported data. Finally, future work should fully counterbalance the order and language of the tasks. In the current study, the English and the Korean sessions occurred on different days, a week apart, thus the consistent order of testing (Korean first) was unlikely to influence the results since order effects typically guarded against by counterbalancing (i.e., fatigue) would be unlikely. In conclusion, the findings of the present study revealed distinctive patterns of recall in bilinguals’ two languages that are generally congruent with the dual-component model of free recall (e.g., Craik & Levy, 1970; Davelaar et al., 2005; Murdock, 1962; Shallice, 1975). Stronger pre-recency effects in the L1 than the L2 suggest that the LTM (presumably the locus of linguistic knowledge) plays a crucial role in the ability to adopt active encoding and retrieval strategies when recalling items from the beginning of the lists. Further, bilinguals’ WM capacity is closely related to recalling items in the pre-recency region in the L1 but not the L2, suggesting that dynamic interactions between the LTM and the STM that characterize free-recall performance in bilinguals are crucially dependent on bilinguals’ knowledge of their two languages. This research was supported by NIDCD Grants R03 DC010465 and R01 DC011750 to Margarita Kaushanskaya. The authors are grateful to Stephanie Van Hecke, Jenna Osowski, Julie Winer, and Marissa Stern for help with data collection and data coding. English Words List (1-Syllable, 2-Syllable and 3-Syllable Words)
Korean Words List (1-Syllable, 2-Syllable and 3-Syllable Words)
|
In the context of serial position, the effect is when one can recall items at the end of the list.
Pos Terkait
Copyright © 2024 ketiadaan Inc.
