The Relationship Between Working Memory Capacity and L2 Oral Performance Under Task‐Based Careful Online Planning Conditionстатья из журнала
Аннотация: The study reported in this article aimed to investigate the way working memory capacity (WMC) interacts with careful online planning—a task-based implementation variable—to affect second language (L2) speech production. This issue is important to teachers, because it delves into one of the possible task-based implementation variables and thus could assist them in making empirically informed decisions in the classroom. It also bears significance for second language acquisition (SLA) researchers, in that it could help them test claims regarding the nature of interlanguages as well as the validity of speech production models that are essential in discussing the role and psycholinguistic functioning of planning in L2 performance and L2 acquisition (Ellis, 2005, 2009). In the area of SLA, one of the most widely accepted models of speech production is that of Levelt (1989). Succinctly put, according to this model "people produce speech first by conceptualizing the message, then by formulating its language representation (i.e., encoding it), and finally by articulating it" (Kormos, 2006, p. 7). In addition, the language produced is assumed to be monitored at three levels: conceptual, pre-articulatory, and post-articulatory (Levelt, 1989). It is inferred that the planning opportunity available to task performers is used to conduct monitoring at (one of) these levels. Planning could take place either prior to task performance (pre-task planning) or while performing a task (within-task planning; Ellis, 2005). According to Ellis, within-task planning could be further differentiated in terms of the extent to which the task performance is pressured or unpressured, depending on the amount of time available for task completion. Unpressured within-task planning is said to lead to careful online planning (COLP), which is defined as "the process by which speakers attend carefully to the formulation stage during speech planning and engage in pre-production and post-production monitoring of their speech acts" (Yuan & Ellis, 2003, p. 6). It is argued that providing learners with COLP would help them produce more accurate and complex language. Empirical support for this hypothesis comes from three studies. Yuan and Ellis (2003) compared the effects of pre-task and online planning on learners' complexity, accuracy, and fluency in oral production. Results of their study indicate that online planners produced both more accurate and complex language. They reasoned that when participants perform a task under time pressure, the working memory (WM) uses the limited time to access lexical information from long-term memory, but when they perform without any time pressure, they can access syntactic information, too. The other two studies, Ahmadian and Tavakoli (2011) and Ahmadian (2012), show that COLP assists both complexity and accuracy in participants' oral production. They also found that COLP led task performers to produce significantly less fluent language than did pressured online planners. Ahmadian and Tavakoli, and Ahmadian, interpreted their findings with reference to the limited capacity of the WM system. Working memory refers to the capacity-limited cognitive mechanism that enables us to temporarily maintain several pieces of information in mind while comprehending, thinking, speaking, and doing (Baddeley, 2003). If WM serves a significant function in our cognitive activities, we should expect important differences between high- and low-WMC individuals in terms of complex cognitive activities. Research findings lend support to this prediction. For example, Rosen and Engle (1997) found that only subjects with great WMC were able to perform verbal tasks fluently while monitoring their output. They also reported that subjects with low WMC made more errors. SLA researchers, too, acknowledge the role that differential WMC plays in the extent to which learners are successful in L2 acquisition and performance. Thus, one may hypothesize that individual differences vis-à-vis WMC would significantly correlate with the extent to which L2 learners benefit from the opportunity to plan their speech in favor of producing reasonably more complex, accurate, and fluent language. To date, however, only one study has examined this hypothesis. Guará-Tavares (2008) investigated the relationship among pre-task planning, WMC, and L2 speech performance. Results of her study indicate that WM correlated significantly with the measures of fluency in the planning group but not in the same task performed by the no-planning group. As for complexity, Guará-Tavares reported that WM was significantly correlated with complexity in her planning group but not in her no-planning group. In her study, WM was not related to accuracy. Nevertheless, it may be logical to posit that interindividual variation in WMC is likely to affect COLP more than pre-task planning, precisely because COLP is thought to involve planning what to say and how to say it while one is performing a task (Ellis, 2005); that is, planning speech while performing another cognitively demanding task. Based on this logic, one may predict that those with greater WM capacity should benefit from COLP more than those with lesser WM capacity—a prediction that guides the present study. The participants in this study were 40 English as a foreign language (EFL) learners (all male, 19–21 years old, with a mean age of 20.5) selected using random number tables from among 93 intermediate-level learners in a language center in Iran. All participants confirmed that they had not been to an English-speaking country and that they had virtually no opportunity to use English for communicative purposes outside the classroom context. They all signed written informed consent forms. To ensure that the participants were equal, they were tested in terms of proficiency level and online processing ability prior to the main study. The participants' language proficiency was tested by the grammar part of the Oxford Placement Test 2 (Allan, 1992). Their responses were scored on a scale of 100 points. Participants had a range of scores between 55 and 60 (M = 57.24; SD = 1.02). The reliability coefficient (Cronbach's alpha) was 0.91. To assess online processing ability, following Yuan and Ellis (2003), all participants were required to take the listening subtest of the TOEFL (Test 1 from Reading for TOEFL Work Book published by Educational Testing Service). Hale (1989) argues that participants' scores on the TOEFL listening section may be indicative of their online processing ability. Scores obtained from this test ranged between 39 and 42 (M = 40.00; SD = 0.92). Therefore, as these two tests revealed, the participants were almost equal in terms of language proficiency and online processing ability. This study used a correlational design and was conducted in a laboratory equipped with computers and audio recording devices. The study was carried out in two sessions with a 1-week interval in between. During the first session, participants' verbal working memory capacity was assessed using a version of listening span task (adapted from Mackey, Philp, Egi, & Fujii, 2002). In the second session, participants performed an oral narrative task under the condition specified for task performance. WM span tasks are assumed to afford the closest approximation of WMC because they induce individuals to simultaneously process and maintain information (Daneman & Carpenter, 1980). In the present study, 36 Farsi sentences were taken from a high school textbook (note that Osaka and Osaka, 1992, found that WMC was language independent). The lengths of the sentences ranged from 9 to 13 words. By rearranging some content words half of the sentences were designed to be syntactically possible but semantically implausible. During the session, each participant was presented with prerecorded sets of sentences read at normal rate. A set comprised three, four, or five sentences, and each sentence span level consisted of three sentences with 2-second intervals in between. The participants were asked to react to each sentence by selecting either an ACCEPTABLE or NOT ACCEPTABLE button (written in Farsi) on the computer screen. At the end of each set a chime sounded and the participants were required to write down the final word of each sentence within that set. The recalled words were neither highly abstract nor semantically related. For each exact repetition of a word, 1 point was awarded. The possible score could range between 0 and 36. Building on Ahmadian and Tavakoli (2011) and Ahmadian (2012), in the present study COLP was operationalized experimentally in a two-pronged way: (a) by providing participants with ample time for task completion and (b) by requiring all participants to start task performance straight away. This latter attempt is usually taken to avoid participants' engaging in pre-task planning. All participants watched a 14-minute silent video of a cartoon and then were asked to narrate the story of the video in English (i.e., their L2) under COLP conditions. The logic in using a silent video of a cartoon was to preclude learners from taking advantage of the immediate exposure to authentic language. Further, the monologic nature of the task was assumed to induce learners to produce stretches of language that were not contaminated by interactional variables (Yuan & Ellis, 2003). In this study, three guiding principles informed the choice of the measures with which to assess the CAF triad: (1) complexity, accuracy, and fluency are themselves multidimensional and multifaceted constructs, and each represents an array of subconstructs, hence the requirement of using multiple measures for assessing each construct (Housen & Kuiken, 2009); (2) some measures used to assess the subconstructs of the CAF triad may tap the same facet of a construct and thus may cause what Norris and Ortega (2009) call redundancy in measurement, and therefore it is imperative to utilize complementary but distinct measures for assessing each principal construct; and (3) to enhance the comparability of results, it is advisable to use the same measures as used in previous studies (e.g., Ellis & Barkhuizen, 2005). Based on these criteria and following Yuan and Ellis (2003), Ahmadian and Tavakoli (2011), and Ahmadian (2012), the following complementary but distinct measures were chosen and used to assess the CAF triad: The transcribed narrations were segmented, coded, and scored based on the measures chosen for assessing the CAF triad. To ensure that the segmentation and scoring of the transcripts were conducted reliably, 50% of the data were segmented, coded, and scored by an independent expert. Intercoder/inter-rater reliability coefficient magnitudes were above .89 for all measures (with a mean of .91). The scores were then entered into SPSS version 16.0 and were checked in terms of normality of distribution via skewness and kurtosis indices. The Pearson correlation coefficient was used to answer the research questions. The overall time (seconds) used by each participant for task completion was calculated to see whether all participants have taken almost equal time for narrating the story (M = 429.20 seconds). It was decided prior to the study to exclude the outliers—defined as one or more than one SD(s) above or below the mean—from the analysis. The rationale behind this decision was to ensure that COLP had been operationalized homogeneously in the sample. Therefore, the time consumed by each participant was converted to a z-score. Four cases were excluded from the analyses by virtue of being more than one SD above or below the mean (z-scores for the excluded cases were −4.29, −1.29, 1.64, and 1.24). Bearing in mind that COLP has been operationalized homogeneously in the sample, I now discuss each of the research questions in turn. Is there a relationship between WMC and oral production of complex language under COLP condition? To answer this question, two complementary but distinct measures were used: syntactic complexity and syntactic variety. As displayed in Table 2, there is no statistically significant association between WMC and these two measures of complexity (see Table 1). This dissociation between WMC and the complexity of oral performance, which is in a way at odds with Guará-Tavares's (2008) findings regarding pre-task planning, may be attributable to the fact that complexity pertains to learners' tendency to take risks and use the cutting edge of their grammatical knowledge; thus, viewed from this perspective, this aspect of performance has little (if anything) to do with WMC. However, regardless of the way we account for this dissociation, this finding assumes both pedagogical and theoretical relevance. Because the complexity of language is conceived of as "the scope of expanding or restructured second language knowledge" (Wolfe-Quintero, Inagaki, & Kim, 1998, p. 4), and if the production of complex language under COLP conditions is not associated with WMC, and if interindividual variations in terms of WMC do not interfere with the positive effects of COLP, then learners with different WMC can benefit from this implementation variable—perhaps, if used over an extended period of time, in favor of L2 acquisition. So this may speak to the viability of COLP as a task-based implementation variable for language development and for virtually all learners with different WMC. Is there a relationship between WMC and oral production of accurate language under COLP condition? As illustrated in Table 2, the correlation coefficient for both measures of accuracy is statistically significant and positive (p < .01). This indicates that WMC and the accurate production of language under COLP conditions are positively correlated, which is in contrast to Guará-Tavares's (2008) results. But how can we account for this finding? Based on Levelt's (1989) model, the grammatical encoding of the message takes place at the formulation stage, where the prelinguistic message needs to be matched with the appropriate linguistic structure extracted from learners' repertoire of explicit knowledge. Because working memory is responsible for, among other things, maintaining important information (e.g., the prelinguistic message) while doing some other cognitively demanding tasks (in this case, planning speech that involves searching for the appropriate linguistic structure), we may reason that those with higher WMC may be able to encode more accurate language. Furthermore, it was noted that language is monitored at three levels. Here again, it may be logical to suggest that participants with higher WMC are more prone to success in monitoring operations during the COLP opportunity and, as a result, produce more accurate language. Is there a relationship between WMC and oral production of fluent language under COLP condition? Results presented in Table 2 show that the correlations between WMC and fluency measures (Rates A and B) were statistically significant and positive. This finding, which agrees with Guará-Tavares's (2008) results, is also consistent with psycholinguistic and neurolinguistic evidence, which indicates that "fluent verbal comprehension and production in the L1 [first language] (and by extension in the L2) is clearly linked to capacity in STM [short-term memory]" (Dewaele, 2002, p. 227). From a psycholinguistic perspective, in native speech most processes (e.g., lexico-grammatical searches, syntacticization) proceed automatically and in parallel (except for conceptualization and monitoring), hence the smooth and relatively fluent language produced by native speakers (see Skehan, 2009). With second or foreign language learners, however, most operations require controlled attention and run serially, which exerts heavy processing loads on the limited capacity of working memory. In other words, as Temple (2000, p. 296) argues, in L2 speech working memory may simply get "overloaded" by (a) attempts to compensate for "the reduced knowledge base of lexis, syntax, morphology, and phonology"; (b) "greater monitoring and repairing"; and (c) allocating resources to inhibit the L1. These operations require even more cognitive control when learners are engaged in COLP. It is no surprise, thus, that higher WMC strongly and positively correlates with fluent language on the part of the participants who perform narrative tasks under COLP conditions. Overall, the findings of this study accord with those of Rosen and Engle (1997). But comparing the results of the present study with those of Guará-Tavares (2008), who investigated the interaction between pre-task planning and WMC, reveals that planning, in general, and WMC interact in complex and intricate ways to affect L2 oral performance. This issue is certainly worth further exploration with larger data sets, using different task types and planning conditions, and, more important, considering the interaction of other individual difference factors with WMC. This is in fact in line with Ellis's (2009) proposal that, in order to move toward a full theoretical account of the role of planning, we need to take into account how, among other sets of variables, individual differences influence the way learners plan their speech and how these variables interact with planning conditions to affect L2 speech and, in turn, L2 acquisition. Task-based language teaching has been criticized on the grounds that performing a task by virtue of its meaning-centered and outcome-oriented nature may at best lead to the production of impoverished and pidginized language, which is of very little value for L2 acquisition (Seedhouse, 1999). However, as results of the host of planning studies indicate, requiring learners to perform tasks under certain conditions can induce them to produce complex and accurate language (for a review, see Ellis, 2009, as well as the recent planning studies cited in this article). Results of this line of inquiry could, therefore, contribute to the enrichment of task-based language pedagogy in English language teaching contexts, as it could afford teachers with procedural options that have been thoroughly investigated from different perspectives and in various contexts. Mohammad Javad Ahmadian received his PhD in applied linguistics at the University of Isfahan, Iran. His major research efforts and output have been in the area of task-based planning. He is interested in cognitive approaches to SLA, task-based language learning, and L2 speech production processes.
Год издания: 2012
Авторы: Mohammad Javad Ahmadian
Издательство: Wiley
Источник: TESOL Quarterly
Ключевые слова: EFL/ESL Teaching and Learning, Second Language Learning and Teaching, Educational and Psychological Assessments
Открытый доступ: closed
Том: 46
Выпуск: 1
Страницы: 165–175