Listening is known as an under-investigated second language (L2) skill. Instructed EFL settings can often have quite limited access to English outside the classroom, resulting in many learners having difficulty improving their listening skills within the time and environment allowed. Thus, a number of researchers support including systematic L2 listening instruction in the form of strategy training to compensate for this lack of input (Vandergrift, Reference Vandergrift2004). Learner strategies are defined by Dörnyei and Skehan as “the learners’ active and creative contribution to enhance the effectiveness of his or her own learning through the application of individualized learning techniques” (2003, pp. 607–608). O'Malley and Chamot (Reference O'Malley and Chamot1990) categorize listening strategies into three groups: cognitive, metacognitive, and social/affective strategies. Cognitive strategies manipulate the information directly to improve learning, such as elaboration, inferencing, and deduction; metacognitive strategies are the higher-order executive skills such as evaluation, monitoring, and planning; lastly, social/affective strategies demand interaction and the affective control of the learner. Vandergrift et al. (Reference Vandergrift, Goh, Mareschal and Tafaghodtari2006), in a large-scale study with 115 L2 learners of English and 226 L2 learners of French, showed the potential role of strategic competence. Their results indicated that 13% of the variability in listening comprehension could be explained by strategic competence. Additionally, previous research has shown that strategy training could help learners improve their L2 listening comprehension when it is conducted for longer periods of time (Macaro, Reference Macaro2006).
Apart from instruction, individual differences such as working memory (WM) and language aptitude (LA) might play a role in L2 listening comprehension, since both are theoretically relevant factors to L2 listening skills. WM is the ability to attend to, store, and manipulate information simultaneously (Baddeley, Reference Baddeley2003) and it is related to many aspects of language comprehension and production (see Linck, et al., Reference Linck, Osthus, Koeth and Bunting2014 for a meta-analysis). Several studies have reported that L2 listening and WM are positively correlated (Brunfaut & Révész, Reference Brunfaut and Révész2015; Gu & Wang, Reference Gu. and Wang2007; Kormos & Sáfár, Reference Kormos and Sáfár2008). However, in a more in-depth and large-scale study, Andringa et al. (Reference Andringa, Olsthoorn, van Beuningen, Schoonen and Hulstijn2012) showed the opposite. They adopted an individual differences approach to explain listening comprehension with 121 native and 113 nonnative speakers of Dutch and explored the extent to which linguistic knowledge, processing speed, memory, and IQ could explain listening comprehension. The study reported that there were only marginal correlations between WM and listening comprehension, linguistic knowledge, and processing speed. In fact, WM could not explain unique variance in listening comprehension. Vandergrift and Baker (Reference Vandergrift and Baker2015, Reference Vandergrift and Baker2018) also reported that, although WM correlated significantly with L2 listening in the initial stages of analysis, the findings revealed lack of significant contribution of WM to listening comprehension in grade four and seven French immersion learners. On the other hand, a very recent study (Masrai, Reference Masrai2020) conducted with 130 B-2 level (i.e., upper intermediate according to the Common European Framework of Reference, 2001) students at a university reported that WM was a significant predictor of L2 listening comprehension measured through the IELTS. Masrai (Reference Masrai2020) reported that WM explained an additional 14% of the variance in L2 listening skill in addition to vocabulary knowledge. As can be seen, there is currently no consensus about WM's role in explaining L2 listening comprehension.
LA refers to “the specific talent for learning foreign languages that exhibits considerable variation between learners” (Dörnyei & Skehan, Reference Dörnyei, Skehan, Doughty and Long2003, p. 590), and it is also assumed to play a role in language processing, auditory processing, and retrieval processes (Skehan, Reference Skehan1998). Findings in relation to LA and L2 listening are even more limited and far from being conclusive. Li (Reference Li2016) conducted a meta-analysis and reported that the average correlation of LA and learning of L2 skills was in the range of .30. Specifically, a weak correlation between LA and listening skill (r = .12) was reported, and the weakest predictor for listening comprehension was the phonemic coding ability component. Some other studies also reported moderate (Nagata et al., Reference Nagata, Aline, Ellis and Ellis1999) to weak (Sáfár & Kormos, Reference Sáfár and Kormos2008) relationships between LA and L2 listening in EFL contexts or no relationship in an ESL context (Ranta, Reference Ranta and Robinson2002). How learning gains are measured may influence the differential correlations of LA and its components (Li, Reference Li, Wen, Skehan, Biedroń, Li and Sparks2019). As can be seen from the above review, previous research reported inconclusive findings about the relevance of WM and LA to L2 listening skill.
The Present Study
To our knowledge, no other study has explored the contribution of individual difference variables (e.g., WM and LA) to the context of explicit strategy instruction on L2 listening. Therefore, this study aimed to investigate whether WM and LA explained any variance in listening comprehension of L2 learners beyond baseline listening ability and explicit strategy instruction. To this end, this study adopted a pretest/posttest non-randomized group design where intact groups were randomly assigned to the experimental condition (N = 19), which received explicit strategy instruction on L2 listening, and to the control condition (N = 17), which followed their L2 academic-oriented listening course syllabus.
Methods
Participants
The participants of this quasi-experimental study were thirty-six young adult learners of English in two intact classes at an English-medium university in Istanbul, Turkey. Participants were from different L1 backgrounds, and their level of English proficiency was A-2 (pre-intermediate) based on a Cambridge Placement Test (CPT) administered by the institution. Participants were at the preparatory school where they studied English for academic purposes before they started their departmental courses.
Instruments
Participants took an academic listening comprehension test, which included a seven-minute recording of a lecture entitled, Silent Languages, and eleven short answer questions that assessed selective listening ability that required processing the relevant, and disregarding the irrelevant, information while listening. The learners reported that they did not have any background knowledge about the topic. Since a parallel form of the test was not available, the same listening test was used both as the pre- and the posttest. To overcome the practice effect, the participants were not informed about the posttest, and the items and the answers were not shared with the participants after the first administration. It was assumed that a one-month interval between the pre- and the posttests was sufficient for them to forget about the items. The highest possible score on the test was eleven. Cronbach's alpha for the pre- and posttest items was .69 and .77, respectively. Two members of the authors scored the answers by following the criteria designed by the institution. The interrater reliability between the raters was 97% and the disagreements were resolved.
WM was assessed through computerized shortened complex span tasks (CSTs) that included one block of operation span task, two blocks of symmetry span task, and three blocks of rotation span tasks developed by Foster et al. (Reference Foster, Shipstead, Harrison, Hicks, Redick and Engle2015). These tasks required the participants to simultaneously store verbal (i.e., letters) or visual (i.e., the location of red squares or the direction and size of arrows) information while processing verbal (i.e., assessing accuracy of arithmetic operations) or visual (i.e., judging symmetry or rotating letters) information. The three tasks were administered to the participants in random order. A single WM measure was obtained based on the total number of correctly retrieved items (i.e., the number of correctly recalled letters for operation span, the number of correctly recalled squares for symmetry span, and the number of correctly recalled arrows for rotation span). The validity and reliability of these tasks as a measure of WM have been discussed in greater detail in Foster et al. (Reference Foster, Shipstead, Harrison, Hicks, Redick and Engle2015).
Lastly, participants took the complete set of LLAMA aptitude subtests (Meara, Reference Meara2005), which includes four subtasks that are thought to be relevant to processes in listening. LLAMA B taps memory through vocabulary learning; LLAMA D asks participants to recognize and differentiate sounds; LLAMA E requires participants to match sounds and symbols; and LLAMA F assesses the ability to identify and infer grammatical patterns. The LLAMA subtests include both verbal and visual stimuli except for LLAMA D, which is only verbal. The LLAMA subtests are also computerized and automatically scored.
Intervention
This study employed a twelve-hour-long explicit strategy instruction (SI) that was specifically designed for the purposes of the current study, following the Cognitive Academic Language Learning Approach method (CALLA; Chamot & O'Malley, Reference Chamot and O'Malley1994). Strategy training was direct, not incidental or accidental (O'Malley & Chamot, Reference O'Malley and Chamot1990; Oxford, Reference Oxford1990); in other words, the students were aware of their learning process. Direct instruction is meant to help students to become more autonomous and engaged in their learning process, leading to faster and more effective learning. As suggested by Plonsky (Reference Plonsky2011), SI was embedded in the curriculum and met four conditions suggested by Rubin et al. (Reference Rubin, Chamot, Harris and Anderson2007) to ensure its effectiveness: extensive practice, opportunities for metacognition, self-regulation for strategy use, and continuous assessment of effectiveness.
As the core of the intervention, the CALLA method integrates learning strategies and instruction in academic language and context with the assumption that learners will acquire academic language and context more effectively by using learning strategies. Following CALLA principles and its five steps (i.e., preparation, presentation, practice, evaluation, and expansion), the instructor first explicitly instructed and modeled the particular strategy and provided guidance on the use of that strategy. Then, students were given an individual or group task to practice the target strategy. After practice, the students discussed whether they could use the target strategy or not. Lastly, the instructor provided different tasks and activities for the transfer of strategies they learned. The students were also given a listening strategies handout, which included explicit information about listening strategies with examples. The intervention was in English and it lasted for a total of twelve hours over four weeks (i.e., three hours each week). On the other hand, the control group only followed the same weekly schedule of their L2 academic listening course. They participated in academic listening tasks designed for their regular classes.
Procedures
The intact classes were randomly assigned to experimental and control groups. Before the intervention, both groups were administered the L2 listening comprehension test as a pretest, the CST, and the LLAMA. After the intervention, both groups took the L2 listening comprehension test as the posttest. Due to the time restrictions, a delayed posttest could not be administered (see Table 1 for the data collection procedure).
Table 1 Data collection procedure
Before the analyses were conducted, univariate normality was examined for all the variables. Based on skewness values and histograms, no severe violations from normality were observed. Independent samples t-tests indicated that the experimental and control group did not differ significantly on the WM and LLAMA measures.
Results
Listening strategy intervention as a unique feature of the current study was found to be effective since independent samples t-test analyses revealed that the groups did not differ in terms of their pretest scores (M = 4.11, SD = 2.45 for the control group; M = 4.29, SD = 2.83 for experimental group) but the intervention group had a significantly higher posttest score than the control group (M = 5.32, SD = 2.65 for the control group; M = 7.76, SD = 2.66 for the experimental group, p < .01, d = .92, 95% CI: [0.26–1.57]).
In order to examine the interrelationships between WM, LA, and listening scores, Pearson product-moment correlations were obtained. Table 2 shows that pretest listening scores are somewhat related to WM and LLAMA B. On the other hand, listening posttest scores significantly related to only LLAMA D.
Table 2 Correlations among the measures (N = 36)
Note. *p < .01, **p < .001
A hierarchical regression analysis was conducted in order to determine whether WM and aptitude explained any variance beyond baseline listening and strategy instruction. Regarding the assumptions of multiple regression, multicollinearity was checked through tolerance and VIF statistics. The average VIF was 1.14 and all the tolerance values were well above 0.2. Thus, it was concluded that there was no collinearity in the data (Field, Reference Field2000). Additionally, an examination of residual plots did not show severe normality violations.
Baseline listening scores entered the analysis in the first step while listening strategy instruction as a categorical variable entered into the analysis in the second step. The WM and LA scores entered the analysis in the last step. Although all three models were significant (see Table 3), the WM and LA scores that entered the analysis in the last step did not explain a significant amount of variance in listening comprehension scores. On the other hand, the baseline listening scores explained 52% of the variance in the postlistening scores while listening strategy instruction explained an additional 16% of variance.
Table 3 Results from the hierarchical regression analysis
Note. *p < .01, **p < .001
Discussion and Conclusion
This study investigated whether strategy-based instruction could boost L2 listening comprehension and whether WM and LA explained any variance in listening comprehension of L2 learners beyond baseline listening ability and explicit strategy instruction. Our findings revealed that strategy instruction is effective and learners’ baseline listening ability is a major determinant in L2 listening comprehension while WM and LA did not make a significant contribution to L2 listening comprehension.
Listening strategy intervention is a unique feature of the current study, and it explained about 16% of the variance in listening comprehension in addition to baseline listening ability. Although very few in number, studies on strategy instruction in L2 listening have shown that strategy instruction is rarely effective with L2 listening due to the nature of listening, which limits the teachability of strategies (see Plonsky, Reference Plonsky2011 for a review). However, the present study has demonstrated that young adult learners seem to benefit from explicit strategy instruction when it is designed with the basic principles of previous research on strategy instruction. We assume that the embeddedness of the intervention and ample opportunities for practice and reflection have been prominent in the success of the intervention.
One interesting finding of the current study is the lack of a role for WM and LA in L2 listening skills. Although a number of studies have shown a significant relationship between WM and L2 listening comprehension (e.g., Brunfaut & Révész, Reference Brunfaut and Révész2015; Gu & Wang, Reference Gu. and Wang2007; Kormos & Sáfár, Reference Kormos and Sáfár2008; Masrai, Reference Masrai2020), our findings corroborate those studies that reported a lack of a significant relationship (Andringa et al., Reference Andringa, Olsthoorn, van Beuningen, Schoonen and Hulstijn2012; Vandergrift & Baker Reference Vandergrift and Baker2015, Reference Vandergrift and Baker2018). Almost all of these studies demonstrated that linguistic knowledge was the strongest predictor of L2 listening comprehension, while the contribution of WM beyond linguistic knowledge was inconclusive.Footnote 1 Although linguistic knowledge was operationalized differently in these studies, such as vocabulary, grammatical accuracy, and segmentation accuracy (Andringa et. al, Reference Andringa, Olsthoorn, van Beuningen, Schoonen and Hulstijn2012), L1 and L2 vocabulary (Vandergrift & Baker, Reference Vandergrift and Baker2018), aural and written vocabulary knowledge (Masrai, Reference Masrai2020), and baseline listening ability as in the present study, its role in L2 listening comprehension was consistent across the studies. On the other hand, different measures of WM in these studies seem to yield inconsistent results regarding the role of WM in L2 listening comprehension. The studies that reported significant results have used verbal measures such as a listening-span task in the L2 (Gu & Wang, Reference Gu. and Wang2007; Masrai, Reference Masrai2020) or backward digit-span task (Brunfaut & Révész, Reference Brunfaut and Révész2015; Kormos & Sáfár, Reference Kormos and Sáfár2008), while those that did not report a significant relationship used verbal WM measures in the L1 such as the Nonword List Recall that taps the phonological loop (Vandergrift & Baker, Reference Vandergrift and Baker2018), nonword recognition task as well as backward digit span task (Andringa et. al, Reference Andringa, Olsthoorn, van Beuningen, Schoonen and Hulstijn2012), or the rotation and symmetry visual span tasks as well as operation span task in the present study. As such, the inconsistent findings about the role of WM in L2 listening comprehension might be attributable to different measurements of WM or different measures of L2 listening comprehension. Previous research used a variety of L2 listening tasks with multiple-choice answers (Vandergrift & Baker, Reference Vandergrift and Baker2015, Reference Vandergrift and Baker2018), questions with short or open-ended alternatives (Andringa et al., Reference Andringa, Olsthoorn, van Beuningen, Schoonen and Hulstijn2012; Masrai, Reference Masrai2020). Similarly, the type and length of L2 listening recordings differ—academic listening such as IELTS (Masrai, Reference Masrai2020) or general listening skills such as real-life speech (Vandergrift & Baker, Reference Vandergrift and Baker2015; Reference Vandergrift and Baker2018) or short monologues (Andringa et al., Reference Andringa, Olsthoorn, van Beuningen, Schoonen and Hulstijn2012).
Just like WM, LA did not significantly explain any variance in L2 listening comprehension either. It is possible that listening comprehension demanded by the task in the current study might not have tapped LA abilities. The selective listening task included in the current study aims at identifying direct meaning relations within or across sentences and processing takes place at the local level. Participants are expected to employ bottom-up processes to comprehend information and top-down processes to predict incoming information. Previous research has reported some evidence for the relationship between L2 listening and aptitude. However, there seems to be a special role for the listening task for the relationship to emerge. In other words, if the task does not include any target grammar structure or vocabulary, but taps only listening comprehension skills, LA may not be predictive of success. This explains why Nagata et al. (Reference Nagata, Aline, Ellis and Ellis1999), using a listening task focusing on L2 target vocabulary, found a relationship between listening and aptitude, while no such relationship was observed by Sáfár and Kormos (Reference Sáfár and Kormos2008), Ranta (Reference Ranta and Robinson2002), and the present study. The latter studies aimed to explore listening comprehension with tasks that did not focus on any specific linguistic target item but aimed at measuring top-down listening comprehension skills. Therefore, the findings of the current study are more compatible with Sáfár and Kormos (Reference Sáfár and Kormos2008) and Ranta (Reference Ranta and Robinson2002) as the listening task tapped both bottom-up and top-down listening processes rather than specific vocabulary or grammatical structure.
We speculate that explicitness in the instruction helped the participants during the listening comprehension task regardless of their WM and LA. Relevant to this explanation, one might argue that the finding could be related to the issue of “the threshold” (Sawyer & Ranta, Reference Sawyer, Ranta and Robinson2001, p. 342). In other words, WM and LA were not vital because the learners were not challenged enough. As previous research has shown, WM plays a greater role with complex tasks such as inferential comprehension instead of literal comprehension (Alptekin & Erçetin, Reference Alptekin and Erçetin2011). It is possible that the selective listening task in the present study was not challenging enough. Instruction might have provided enough scaffolding so the participants did not need to rely on any other resources. In conclusion, the present study proposes listening strategy instruction as a strong predictor for L2 listening comprehension with no significant role for individual learner factors, namely WM and LA.
This was a small-scale study that explored the effect of listening strategy instruction in relation to individual difference variables of LA and WM. As expected, the study has some limitations. The small sample size is a major limitation that causes a lack of power in the statistical analyses.Footnote 2 Future research should analyze potential interactions between ID variables and strategy training with a larger sample size by adopting an aptitude-treatment-interaction design, as suggested by DeKeyser (Reference DeKeyser2012). How learner profiles such as high-low WM or high-low LA benefit from instruction and how different types of tasks affect different learner profiles can be further research areas. Finally, the choice of listening task could be a limitation for the current study, as it may have failed to trigger the use of IDs. Future research should use a variety of more challenging listening tasks, preferably standardized tests with established validity.
