Learning and Instruction 88 (2023) 101808
Available online 31 July 2023
0959-4752/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/).
Examining the relationship between English as a foreign language learners ’ cognitive abilities and L2 grit in predicting their writing performance
Jianhua Zhang
a,b, Lawrence Jun Zhang
b,*aSchool of Foreign Languages, Sichuan University of Arts and Sciences, China
bFaculty of Education and Social Work, University of Auckland, New Zealand
A R T I C L E I N F O Keywords:
L2 learners Cognitive abilities Grit
Writing performance Moderating effects
A B S T R A C T
Research aims: This study was designed to explore the interplay among English as a foreign language learners’ cognitive abilities and grit in predicting writing performance along different task complexity through path analysis and multiple group analysis.
Methodology: A total of 353 tertiary students from Western China were recruited to participate in this study. Their language aptitude and working memory capacity were evaluated by the LLAMA tests and the automated oper- ation span task, respectively. They were required to complete writing tasks with different complexities: argu- mentation (complexity+) and narration (complexity-).
Results and conclusion: The results revealed that (1) the effects of perseverance of effort and working memory on L2 writing performance were significant in the argumentative task, whereas the influences of perseverance of effort and consistency of interest on L2 writing performance were significant in the narrative one; (2) task complexity moderated the respective impacts of grammatical inferencing ability and working memory on L2 writing performance. Theoretical and pedagogical implications were discussed.
Author credit statement
Jianhua Zhang: Conceptualization, Methodology, Data collection, Formal analysis and interpretation, Drafting the manuscript, Writing and revising, Lawrence Jun Zhang: Conceptualization, Methodology, Formal analysis and interpretation, Writing and revising, finalizing for submission
1. Introduction
Language learning and the learning outcome are conceived to be complex, dynamic, nonlinear, and variable (Larsen-Freeman, 2007;
Lowie et al., 2010), where learners would encounter difficulties and challenges and thus need to persevere and remain interested (Chang &
Zhang, 2023). Grit, a personality trait, is defined as “perseverance and passion for long-term goals” (Duckworth et al., 2007, p. 1087), which involves working effortfully toward challenges and sustaining effort and interest over a long period of time when facing failure, adversity, and plateaus in progress (see also, Fathi et al., 2023). Scholars argued that grit should be conceptualized as domain-specific rather than domain-general in nature (e.g., Cormier et al., 2019; Duckworth &
Quinn, 2009; Schmidt et al., 2019; Teimouri et al., 2020). With regard to second or foreign language (henceforth referred to as L2) learning, L2 grit refers to “passion and perseverance for second language learning”
(Teimouri et al., 2020, p. 1). Studies have demonstrated that L2 grit and its components exert different effects on L2 achievement (e.g., Sudina et al., 2020; Teimouri et al., 2020).
Deliberate practice has been conceived to be an essential prerequisite for achieving success in mastering writing competence (Kellogg &
Whiteford, 2009; Panero, 2016), where learners need to devote them- selves to hard work over time and have sustained passion. Thus, being gritty is essential to mastering first language (L1) and L2 writing skills.
Furthermore, compared with L1 learners, L2 learners might possess less linguistic knowledge (e.g., vocabulary and grammar) and genre knowledge under their belt to utilize in their L2 writing. Acquiring L2 vocabulary and grammar would also require learners to invest enduring perseverance and passion over an extended period of time (Chang &
Zhang, 2021). It is evident that acquiring L2 writing skills is closely intertwined with mastering L2 linguistic and genre knowledge (Rij- laarsdam & Van den Bergh, 2006). Thus, grit is more vital to the pro- gression of L2 writing competence and skills in comparison to learning to write in an L1. However, no available studies have explicitly studied
* Corresponding author. Faculty of Education and Social Work, University of Auckland, 74 Epsom Avenue, Auckland, 1023, New Zealand.
E-mail address: [email protected] (L.J. Zhang).
Contents lists available at ScienceDirect
Learning and Instruction
journal homepage: www.elsevier.com/locate/learninstruc
https://doi.org/10.1016/j.learninstruc.2023.101808
Received 9 November 2022; Received in revised form 13 June 2023; Accepted 15 July 2023
the relationship between grit and L2 writing achievement.
When it comes to L2 writing, the role of cognitive abilities (i.e., language aptitude and working memory) has been established as the predictive factors of L2 writing achievements (e.g., Abu-Rabia, 2003;
Kormos & S´af´ar, 2008; Kormos & Trebits, 2012; S´af´ar & Kormos, 2008;
Lu, 2010; Michel et al., 2019; Mavrou, 2020; Sparks et al., 2009; R´ev´esz et al., 2017; Yang et al., 2019; Zalbidea, 2017), Language aptitude refers to a conglomerate of “basic abilities that are essential to facilitate foreign language learning” (Carroll & Sapon, 1959, p. 14), including four components: Phonetic coding ability, grammatical sensitivity, rote learning ability, and inductive learning ability. In contrast, working memory is defined as “a system of components that holds limited in- formation temporarily in a heightened state of availability for use in ongoing processing” (Adams et al., 2018, p. 341). Nevertheless, they have been conceived of as being innate and relatively stable, thus leaving little space for instructional inventions. In contrast, L2 students’
grit could be enhanced and improved through the instructional in- terventions (Shafiee Rad & Jafarpour, 2022), which could enhance their writing competence. Hence, examining the relationship between cognitive abilities and grit in the field of L2 writing would be significant in shedding insightful light on how to manipulate L2 student writers’
grit to improve their writing achievement.
Cognitive abilities and noncognitive factors (i.e., grit) are conceived of as complementary or substitutive in predicting learning achieve- ments, as evidenced in educational psychology research (Cred´e, 2018;
Cred´e et al., 2017; Gil-Hern´andez, 2021). It is also clear that the moderating effect of task complexity on that relationship was confirmed by Light and Nencka (2019). These findings might offer insights into how cognitive abilities and grit could interplay in predicting L2 writing achievements in writing tasks.
To bridge the gap in the available research, this study intended to scrutinize the interplay among L2 learners’ cognitive abilities (i.e., language aptitude and working memory) and grit in predicting their writing achievement along different task complexities through struc- tural equation modelling, where the relationship between cognitive abilities (i.e., language aptitude and working memory) and components of grit and L2 writing achievement and the moderating effect of task complexity on that relationship were examined.
2. Literature review
2.1. The effect of grit on L2 learning and L2 writing
Grit generally entails two subconstructs: perseverance of effort and consistency of interest (Duckworth et al., 2007). Studies on the role of grit in L2 learning have thrived in recent years and advanced in two lines of inquiry characterized by adopting different conceptualizations and measurements of grit: domain-general (e.g., Duckworth & Quinn, 2009;
Khajavy et al., 2021; Robins, 2019; Wei et al., 2019) and domain-specific (e.g., Sudina et al., 2020; Teimouri et al., 2020). The domain-general conceptualization of grit emphasizes its relative stability across time and situation (Von Culin et al., 2014), whereas the domain-specific conceptualization of grit focuses on the grit levels in various achieve- ment domains, such as sport (Cormier et al., 2021) and L2 learning (Sudina et al., 2020).
Prior to conceptualizing grit as a domain-specific construct, scholars initially adopted the domain-general conception of grit and scrutinized its role in L2 learning (e.g., Khajavy et al., 2021; Robins, 2019; Wei et al., 2019). Wei et al. (2019) explored the relationship between grit and English as a foreign language (EFL) proficiency among Chinese adoles- cents. The participants’ grit was evaluated by the Short Grit Scale designed by Duckworth and Quinn (2009), and their EFL proficiency was assessed by a standardized exam with a total score of 150. Corre- lational analyses revealed that grit was significantly positively corre- lated to EFL proficiency (r = .26, p < .001). Moderated mediation analysis revealed that (1) grit showed a direct positive effect on EFL
proficiency (β =.21, p <.001); (2) grit also showed a positive effect on EFL proficiency through enjoyment; (3) class environment had a moderating role in the relationship between grit and EFL proficiency.
Meanwhile, Robins (2019) probed the relationship between English as a second language (ESL) learners’ grit and their L2 achievement in an online context. L1-Spanish or L1-Portuguese ESL learners involved in an online ESL program were invited to participate in the study. Their grit was evaluated by the Grit Scale, while their L2 achievement was assessed by their grade point average (GPA). Correlational analyses revealed that grit and its components were significantly positively correlated to L2 achievement (rgrit =.158, p <.001; rCI =.116, p <.001;
rPE =.160, p <.001). In addition, Khajavy et al. (2021) probed the predictive effect of grit on tertiary EFL learners’ achievement. Their grit was evaluated by the Short Grit Scale, whereas their L2 achievement was measured by a composite test including reading comprehension, grammar, vocabulary, and pronunciation. They reported that none of the grit components significantly correlated with the participants’ EFL achievement (rPE = .06; rCI = .04). Structural equation modelling revealed that neither perseverance of effort nor consistency of interest could have a significant predictive effect on their EFL achievement (βPE
=.007, p =.925, Cohen’s f2 <.001; βCI =.070, p =.349, Cohen’s f2 <
.001).
Scholars have postulated that grit be domain-specific rather than domain-general (e.g., Cormier et al., 2019; Duckworth & Quinn, 2009;
Schmidt et al., 2019; Teimouri et al., 2020). Therefore, domain-specific conceptualizations of grit have been proposed, for instance, academic grit (Clark & Malecki, 2019) and teacher grit (Sudina et al., 2021). L2 grit is defined as “passion and perseverance for second language learning” (Teimouri et al., 2020, p. 1), similarly including two subcon- structs: perseverance of effort and consistency of interest. Accordingly, a corresponding domain-specific scale, the L2 Grit Scale, has been devel- oped and validated against L2 students (Teimouri et al., 2020), demonstrating high reliability (α =.80) and validity (i.e., significant correlation with GPA (r =.25, p ≤.001)). Plonsky and his colleagues examined the relationship between L2 grit and L2 achievements (e.g., Sudina et al., 2020; Teimouri et al., 2020). Teimouri et al. (2020) explored the relationship between L1-Persian EFL learners’ L2 grit and English achievements. These learners’ L2 grit was assessed by the L2 Grit Scale, while their English achievements were evaluated by grades in three English courses (grammar, laboratory, and speaking), GPA, and English language proficiency. They reported that the overall L2 grit and perseverance of effort had significant, positive, and modest correlations with all the language achievement measures (L2 grit, rgrammar =.26, p ≤ .001; rspeaking =.30, p ≤.001; rlaboratory =.24, p ≤.001; rGPA =.25, p ≤ .001; rproficiency =.31, p ≤.001; perseverance of effort, rgrammar =.24, p
≤.01; rspeaking =.37, p ≤.01; rlaboratory =.33, p ≤.001; rGPA =.27, p ≤ .001; rproficiency =.27, p ≤.001), whereas consistency of interest had nonsignificant and positive correlations with the achievement measures (rgrammar =.08; rspeaking =.01; rlaboratory =.02; rGPA =.05; rproficiency = .11). Sudina et al. (2020) further explored the relationship between L2 grit and language achievements in ESL and EFL populations with different mother tongues. They found that (1) perseverance of effort served as a significant positive predictor of self-rated proficiency in the EFL sample (β =.14, p =.05) but not in the ESL sample (β =.06, p = .54); (2) consistency of interest acted as a negative predictor in the ESL sample (β = − .45, p <.001) but as a nonsignificant predictor in the EFL sample (β = − .07, p = .35). Their findings suggested that language learning contexts might exert a moderating effect on the relationship between L2 grit and L2 achievement.
The progression of writing skills could be reliant on “the availability of adequate working memory resources and the capacity to allocate them appropriately to planning, sentence generation, and reviewing” (Kellogg & Whiteford, 2009, p. 256). Acquiring such cognitive control could be facilitated and accelerated by deliberate practices (Kellogg &
Whiteford, 2009; Monk, 2016; Panero, 2016). In light of the significant differences between L1 and L2 writing (Wei et al., 2021; Zhang, 2022),
L2 students would be required to devote more deliberate practices and thus be more persistent and passionate over time. As such, it is signifi- cant to explore the effect of L2 grit on L2 writing achievement.
2.2. The effect of language aptitude on L2 writing
Language aptitude is argued to exert effects on the translation and reviewing phases of the writing process through linguistic processing and available linguistic resource (Kormos, 2012), thus more or less determining L2 writing achievement. Experimental studies have been accumulated on the correlation between language aptitude and L2 writing (e.g., Kormos & Trebits, 2012; S´af´ar & Kormos, 2008; Sparks et al., 2009; Yang et al., 2019). Kormos and her colleagues investigated the relationship between language aptitude and L2 writing in the Hun- garian context. S´afar and Kormos (2008) probed the role of language ´ aptitude in predicting writing success in focus-on-form instruction.
Hungarian-English bilingual students at a secondary school were assigned to write in three different genres, and their writing perfor- mance was assessed in terms of content and accuracy. Four components of their language aptitude were measured by the standard Hungarian language aptitude test: phonetic coding ability, inductive learning ability, grammatical sensitivity, and rote learning ability. The aptitude test was administered twice to students: at the start and the end of one academic year. They found that students’ writing performance showed only a significant relationship with inductive learning ability examined at the start of the year (r =.44, p <.01) and nonsignificant, weak re- lationships with other abilities measured at the beginning (rphonetic coding ability = − .06; rgrammatical sensitivity = − .04; rrote learning ability =.21) and end of the year (rphonetic coding ability = .05; rinductive learning ability = .02;
rgrammatical sensitivity =.02; rrote learning ability =.02). Kormos and Trebits (2012) elaborated on the role of language aptitude in L2 writing by focusing on the correlations between components of language aptitude and different aspects of text quality in the narrative tasks.
Hungarian-speaking English students were recruited to complete two writing tasks: the cartoon description task requiring students to describe a comic strip involving six pictures and the picture description task requiring them to tell a story according to six unrelated pictures. Their writings were assessed in terms of fluency, lexical variety, syntactic complexity (i.e., the diversity and sophistication of syntactic structure), and accuracy. Their language aptitude was measured by the same test as in S´af´ar and Kormos (2008). Results demonstrated that grammatical sensitivity significantly correlated with only one measure of syntactic complexity in the written cartoon description task (r =.46, p <.01).
Meanwhile, Sparks and his colleagues examined the predictive role of language aptitude in L2 writing. Sparks et al. (2009) explored the effect of language aptitude on L2 proficiency and writing. The partici- pants involved were secondary school students who had learned Span- ish, French, or German for two years. Their language aptitude was examined by the Modern Language Aptitude Test (MLAT), and their L2 proficiency by the self-developed tests covering reading, listening, speaking, and writing. Additionally, they were also required to write a response letter in their L2. It was found that language aptitude showed a predictive effect on L2 writing performance (β =.58, p <.001) but a better one on general L2 proficiency (β =.60, p <.001). Sparks et al.
(2011) furthered their early study by focusing on aptitude components.
They found that (1) components of language aptitude showed significant correlations with scores of the L2 writing test (rnumber learning =.45, p <
.01; rphonetic script =.45, p <.01; rspelling clues =.49, p <.01; rwords in sentences =.50, p <.01; rpaired-associates =.34, p <.05); (2) three factors significantly predicted L2 writing performance: language analysis (R2 = .12), phonology/orthography (R2 =.10), and self-perceptions of lan- guage skills (R2 =.09).
Recently, Yang et al. (2019) explored the relationship between lan- guage aptitude and its components and the overall quality of Chinese tertiary EFL learners’ narrative writing. The participants were required to write a narrative essay based on a sequence of pictures. Their
language aptitude was evaluated by the LLAMA test modules developed by Meara and her colleague (Meara, 2005; Meara & Rogers, 2019). It was revealed that phonetic coding ability had a moderate relationship with the overall quality (r =.34, p <.05), while grammatical inferencing ability correlated slightly with the quality (r = .09). However, after reviewing 66 studies relevant to language aptitude in L2 learning through a meta-analysis, Li (2016) concluded that aptitude and its components exerted nonsignificant predictive effects on L2 writing (daptitude =.34; dphonetic coding =.26; dlanguage analytic ability =.20; drote memory =.12) except for number learning (d =.42) and spelling clues (d
=.42).
The accumulated studies presented a mixed picture of the relation- ship between language aptitude and L2 writing achievement because of different measures of language aptitude adopted by scholars and writing tasks of various genres. Therefore, to clarify the relationship between language aptitude and L2 writing achievement, it is necessary to include writing tasks with different genres in the research design.
2.3. The role of working memory in L2 writing
Working memory is involved in all phases of the writing process:
planning, translating, programming, executing, reading, and editing (Kellogg, 1996). After reviewing a series of studies designed to test Kellogg’s (1996) model of working memory in writing, Kellogg et al.
(2013) summarised that “the empirical evidence convincingly supported the supremacy of the central executive in writing, the important role of the phonological loop in the linguistic process of sentence generation, and the visual-spatial sketchpad in the thinking or problem-solving act of planning” (p. 178). It is clear that research has established that working memory plays a vital role in L1 writing performance and pro- cesses (Kellogg et al., 2013). Accordingly, research into working mem- ory has gained increasing popularity in L2 writing. Scholars focused on different components and functions of working memory, for instance, phonological memory (Kormos & S´af´ar, 2008), working memory ca- pacity (e.g., Lu, 2010; Mavrou, 2020; R´ev´esz et al., 2017), storage and processing function (Zalbidea, 2017), task-switching (Zalbidea, 2017).
The available studies on the role of working memory capacity in L2 writing were narrowly reviewed here.
Lu (2010) explored the effect of working memory capacity on Chi- nese EFL learners’ writing performance in timed argumentative writing.
The participants’ working memory capacity was measured by the operation span (OSPAN) task. They were asked to write an essay on the prompt adopted from the TOEFL Test of Written English (TWE) writing prompts. Their essays were assessed in terms of content and organiza- tion, and language use. He found no correlation between working memory capacity and the overall quality of the participants’ argumen- tative essays.
Recently, R´ev´esz et al. (2017) explored the role of working memory capacity in L2 academic writing. The participants’ working memory was measured by the automated OSPAN task and their writing achievement by IELTS academic writing task. Several dimensions of the participants’
writing performance were evaluated: overall quality, lexical diversity, syntactic complexity, discourse complexity, and accuracy. Spearman correlation analyses indicated that working memory capacity nonsig- nificantly correlated with dimensions of L2 students’ writing achieve- ments (overall quality (− .01 <r <.35), lexical complexity (− .15 <r <
.35), syntactic complexity (-.17 <r <0.30), discourse complexity (− .18
< r < .30), and accuracy (r = − .19)). Meanwhile, Mavrou (2020) investigated the relationship between working memory capacity and L2 students’ writing achievement in a narrative task. Dimensions of the participants’ writing performance were examined: complexity, accu- racy, and fluency. Linear regression analyses revealed the predictive effects of working memory capacity on syntactic complexity (i.e., sub- ordination density, complexity achieved via subordination, (B =.088, p
=.012)) and linguistic accuracy (i.e., the percentage of words that are correct in terms of grammar and lexical choices) (B =.516, p =.002).
The studies reviewed above might suggest the potential moderating effects of genre on the relationship between working memory capacity and L2 students’ writing achievement. Accordingly, it is imperative to go further to examine whether genre might moderate the relationship between working memory and L2 writing achievement, thus offering insights into the role of working memory in L2 writing.
2.4. Cognitive abilities, grit and their links
The relationship between cognitive abilities and grit appealed to scholars because of the innate nature of cognitive abilities and the malleability of grit. After a meta-analytic synthesis of 87 studies related to grit, Cred´e et al. (2017) conjectured that the relationship between grit and performance might be moderated by the tasks adopted and abilities and metacognition. Specifically, grit might serve as a better predictor of performance in difficult tasks than in easy ones because success in challenging tasks requires sustained effort and deliberate practice.
Additionally, Cred´e (2018) made further assumptions: “it is also possible that noncognitve attributes such as grit are particularly important for individuals at the lower end of the ability spectrum” (p. 4).
Based on Cred´e’s conjectures, Light and Nencka (2019) proposed the skill substitution hypothesis: the synergistic effect of grit for high-ability students and the compensatory effect of grit for low-ability students. To be specific, grit might enhance high-ability individuals’ ability advan- tage. In contrast, low-ability individuals might substitute or compensate for their ability shortcomings by devoting more effort to achieving the same success in certain task performance. These effects were assumed to be moderated by task complexity. Light and Nencka (2019) found that high-ability individuals would benefit more from the synergistic inter- action between grit and ability when performing complex tasks, whereas low-ability individuals must substitute their grit for their ability disad- vantages in achieving success in less complex tasks. These findings might provide insights into the interplay among grit, cognitive abilities, and learning: Grit would be vital for achieving success in less complex tasks, while both grit and cognitive abilities would be essential for achieving the same success in complex tasks.
When it comes to L2 learning and teaching, task complexity has been framed within the Limited Attentional Capacity Model (Skehan, 1998, 2003) and the Cognition Hypothesis (Robinson, 2001, 2003, 2005).
Compared with Skehan’s model, Robinson’s Cognition Hypothesis might be more finetuned in that it offers more specific guidance on how to operationalize the dimensions of task complexity (Kuiken & Vedder, 2007). In his Cognition Hypothesis, Robinson included three compo- nents: task complexity, task condition, and task difficulty. Task complexity is narrowly concerned with two cognitive task features:
resource-directing and resource-dispersing. According to Robinson (2005), resource-directing dimension involves such variables as ±few elements, ± here/now, ±perspective-taking, ± spatial reasoning, ± causal reasoning, and ± intentional reasoning, while resource-dispersing dimension entails ±planning, ±single task, ±prior knowledge, ±task structure, ±few steps, and ±independency of steps.
After a meta-analytic review of the available studies, Johnson (2017) found that increasing the resource-directing and resource-dispersing features of a task would exert significant impacts on different aspects of L2 writing performance. A single writing task involves both genre and task complexity. Scholars has established the relationship between genre and task complexity (e.g., Ong & Zhang, 2010, 2013; Rahimi &
Zhang, 2022; Yoon & Polio, 2017). Argumentative writing tasks have been argued to be more cognitively complex than narrative ones because the former involve logical causal reasoning while the latter involve in- formation transmission, especially for the purposes of entertaining the audience (Ong & Zhang, 2013; Rahimi & Zhang, 2022; Yoon & Polio, 2017; Zhang, 2021). For example, framed with the Cognition Hypoth- esis, Yoon and Polio (2017) found that argumentative tasks are cogni- tively more complex than narrative tasks in the L2 context.
2.5. Research questions
To narrow down the lacuna in the literature, this study is designed to explore the interplay among L2 learners’ cognitive abilities (i.e., lan- guage aptitude and working memory) and grit in predicting their writing performance in genres (i.e., tasks with different complexity) through structural equation modelling in order to show how the relationships would play out. In this study, we focused on components of grit (i.e., perseverance of effort and consistency of interest) rather than overall grit because its factorial structure was not well supported (Cred´e et al., 2017). Specifically, this study aimed to address the following questions.
RQ1. What roles do cognitive abilities and components of L2 grit play in predicting EFL students’ writing achievement in a narrative task?
RQ2. What roles do cognitive abilities and components of L2 grit play in predicting EFL students’ writing achievement in an argumentative task?
RQ3. Does genre/task complexity moderate the effects of cognitive abilities and components of L2 grit on EFL students’ writing achievement?
Based on the aforementioned insights, we formed the following hypotheses.
Hypothesis 1. Components of L2 grit might have more predictive ef- fects than cognitive abilities in predicting EFL learners’ writing achievement in a less complex writing task.
Hypothesis 2. Components of L2 grit and cognitive abilities might have significant predictive effects in predicting EFL learners’ writing achievement in a complex writing task.
Hypothesis 3. Task complexity might moderate the effects of cogni- tive abilities and components of L2 grit on EFL students’ writing achievement.
3. Methods 3.1. Participants
A total of 389 EFL students from two medium-ranking universities in Western China were recruited to participate voluntarily in this study through convenience sampling after signing formal consent. Of the participants, 59.13% (n =230) were females, and 40.76% (n =159) were males. Their ages ranged from 18 to 22, with an average of 20.4.
They enrolled in programs such as Electronic Engineering (n = 62, 15.94%), Computer Science (n = 67, 17.22%), Education (n = 68, 17.48%), Business (n =103, 26.48%), Administration (n =89, 22.88%).
They were in different grades, of which 38.56% (n =150) were fresh- men, 33.42% (n =130) sophomores, and 28.02% (n =109) juniors.
3.2. Instruments 3.2.1. L2 Grit Scale
The L2 Grit Scale (L2GS) was directly modified from Teimouri et al.’s (2020) instrument to measure respondents’ personality traits and behavioural features in L2 contexts. Specifically, this scale was designed to gauge perseverance and passion for second language learning, mainly English as a foreign language. It was a domain-specific self-report in- strument, with a 5-point Likert scale ranging from 1 (not like me at all) to 5 (very much like me). Teimouri et al. (2020) validated the scale with L1-Persian EFL students with different English proficiencies in different age groups and reported high validity and reliability. Given that the L2GS was developed and validated in the L1-Persian context, the diction of the selected items was checked and modified to be suitable for the L1-Chinese EFL context and was used after it was translated into Chinese and piloted. In the same fashion, the accuracy and equivalence of the
translation were verified and backed up by means of translating and back-translating. The scale was piloted before being put into use. Two items (items 2 and 3 in the original scale) were removed because they showed factor loading lower than the recommended threshold of .5 (Comrey & Lee, 2016) against the targeted L2 students during the pilot study. The Cronbach coefficient for the subscale for measuring perse- verance of effort is .788 (p <.001), and that for consistency of interest is .733 (p <.001). The correlation between the total scores for the subscale of perseverance of effort and that of consistency of interest is .227 (p <
.01).
3.2.2. The LLAMA tests
The LLAMA tests were utilized to measure the participant’s language aptitude (i.e., vocabulary learning ability and grammatical inferencing ability). They were designed as a computer-based aptitude test battery (Meara, 2005) and developed as an online free-accessible one (Meara &
Rogers, 2019). The LLAMA tests were employed in this research in that they are robust (Rogers et al., 2017) and available at the following address: http://www.lognostics.co.uk/tools/LLAMA_3/index.htm.
Based on the classic MLAT, the battery was developed as a free suite of language-neutral tests (Rogers et al., 2017), which is user-friendly and requires approximately 25 min for completion. It is comprised of four subtests: vocabulary acquisition, sound recognition, sound-symbol as- sociations, and grammatical inferencing. LLAMA_B evaluates students’ vocabulary learning skills, specifically their ability to establish con- nections between unfamiliar names and unfamiliar objects. LLAMA_D, a test of sound recognition that is an addition to MLAT, assesses students’ ability to recognize sound patterns in spoken language. LLAMA_E, a test of sound-symbol associations, examines students’ skills to establish correspondences between sounds and spellings. LLAMA_F, a test of grammatical inferencing, evaluates students’ ability to pick up the grammatical rules of an unfamiliar language. Each sub-test is individu- ally and automatically scored during the administration process. The picture stimuli and verbal materials employed in these tests were adapted from a British-Columbian indigenous language and a Central American language. Question items are randomly selected and pre- sented on the web pages every time so that potential cheating caused by the memorization of these items would be avoided.
The LLAMA tests were initially piloted and then put into use.
Volunteer participants were invited and informed of the LLAMA test website. Copies of the Chinese manual for the LLAMA tests were distributed to the participants, who were asked to read the manual carefully and comprehend it before taking part in it. Participants can have access to the website through their smartphones or computers.
After completing the tests, they were asked to send the results to the authors by email. Furthermore, we requested the participants’ scores from the manager of the LLAMA test website to cross-validate their re- ported scores. The scores of LLAMA_B and LLAMA_ F range between 0 and 20. As stated above, each sub-test module is individually and automatically scored and therefore, individual scores were given for two targeted aspects of participants’ language aptitude. The LLAMA tests have been reported to have high reliability, α =.78 (see e.g., Granena, 2013). Pitifully, the reliability of the LLAMA tests for the sample involved in this study cannot be calculated because of the inaccessablity of the scores of the items included in these tests.
3.2.3. The automated operation span task
An OSPAN task was utilized to evaluate the participants’ working memory capacity in English. The OSPAN task adopted in this research was the Automated OSPAN developed by Unsworth et al. (2005), which was constructed on the basis of Engle and his colleagues (e.g., Kane &
Engle, 2003; Kane et al., 2001; Turner & Engle, 1989). In the OSPAN test, participants are asked to perform two tasks: completing several composite mathematical operations, including addition, subtraction, multiplication and division (e.g., (10–2)*7 = ?) and memorizing the words or letters followed immediately. The OSPAN task was selected
because (1) it would avoid the confounding effect of language profi- ciency on working memory capacity measured by the reading span task (e.g., Service et al., 2002; Van den Noort et al., 2006); and (2) it has established good correlations with other measurements of working memory, with a high Cronbach coefficient, α = .78, as reported in Unsworth et al. (2005).
The Automated OSPAN was accessible from the website: http://
www.millisecond.com, where participants can take in and finish the test through Inquisit Web that the authors distributed to them. It included two trials: a practice trial and a real trial. The practice trial was divided into three sessions: a simple letter span, math operation, and both. During the practice trial, participants were asked to complete three practice sessions so that they would familiarise themselves with the test.
Then, participants were required to complete three sets whose set size ranged from 3 to 7. For instance, in a real trial set with set size 3, par- ticipants were asked to compute an arithmetic operation, decide whether the provided answer is right, memorize the letter appearing on the screen, and click the memorized letter on the screen in the right sequence. There is a time limit for participants to complete the math problem. If they could not finish within the time limit and the program might move on automatically, their trial would be counted as a failure.
After completing the test, the participants were provided with five scores: OSPAN score, total number correct, math errors, speed errors, and accuracy errors. The OSPAN score indicates the total of correctly recalled sets, which was utilized to indicate working memory capacity in this study. Unfortunately, this online Automated OSPAN just offered a single general score rather than the scores of the items included, which did not allow us to calculate the reliability coefficients for the sample involved in this study.
3.2.4. English writing test
Participants involved in this study were asked to write two English compositions of at least 150 words according to the given prompts: one was argumentation, and the other was narration. The theme of the argumentative writing task was concerned with the utility of computers and the Internet for Chinese university students’ learning. The topic for the argumentative writing was chosen from the old item pool of the College English Test, Band 4 (CET 4), which showed high validity and reliability. In contrast, the topic for the narrative writing was designed as culturally inoffensive and closely related to participants’ daily life (see Appendix A for details). With reference to the literature and con- sultancy with experts in task complexity, the argumentative task was coded as complexity+and the narrative task as complexity-.
Participants’ writing performances were assessed by Jacobs et al.’s (1981) ESL Composition Profile, one of the established analytic scoring rubrics in L2 writing studies, which has been widely used to evaluate the writing proficiency levels of L2 students around the world due to its simple use (Huang & Zhang, 2020; Teng & Zhang, 2016; Xu et al., 2022).
The profile was designed to assess five dimensions of L2 writing per- formance and also assigned different weights to them: content (30%), organization (20%), language (25%), vocabulary (20%), and mechanics (5%). These dimensions were rated at four levels from high to low:
excellent to very good, good to average, fair to poor, and very poor. Two experienced CET4 writing raters were invited to score the collected es- says independently, according to Jacobs et al.’s scoring rubrics.
The inter-rater reliability for the two writing tasks are as follows: In argumentative writing, the inter-rater reliability is .897 (p <.001) for content, .831 (p <.001) for organization, .804 (p <.001) for vocabulary, .912 (p <.001) for language use, .584 (p <.001) for mechanics, and .939 (p <.001) for the total score; in narrative writing, the inter-rater reli- ability is .891 (p <.001) for content, .795 (p <.001) for organization, .772 (p < .001) for vocabulary, .921 (p <.001) for language use, .541 (p
<.001) for mechanics, and .885 (p <.001) for total scores. The Cron- bach coefficient for scores of argumentative writing is .921, and that for scores of narrative writing is .787. The Pearson correlation coefficient for students’ total scores in the two writing tasks is .192 (p < .01).
According to the benchmark proposed by Cohen (1988), students’ per- formances in two writing tasks had a weak connection to each other, thus demonstrating that students had distinct performance in two gen- res, as revealed in Yoon and Polio (2017).
3.3. Procedure
Due to the impact of the COVID-pandemic and the resulting lock- down, we had to collect the data entirely online. This study included two phases, as demonstrated in Fig. 1.
In phase one, before responding to the L2 Grit Scale, the participants were requested to sign the consent form. They completed the L2 Grit Scale posted on an online questionnaire platform named Wenjuanxing (literally translated to mean “Questionnaire Star”, www.wemjuanxing.
com), which shared similar functions with Qualtrics. As mentioned in the above section, all the items of the scale were translated into Chinese when presented to the participants, which could guarantee they could fully understand the items and avoid potential misunderstandings.
The scale was intended to elicit L2 students’ authentic context-based information about perseverance of effort and consistency of interest for L2 learning. They were told that their responses might not be scored as right or wrong according to specific scoring rubrics before giving the responses. If they could offer authentic responses, they might be greatly appreciated, and their response would be highly valued. They were also informed that their responses would not bring out any impact on the assessment of their course grades at all. Before distributing the survey links to the participants, we reviewed and clarified the instructions. We asked the teachers to confirm that the instructions were clear to them and then distribute the scale to their students by sharing the links via WeChat or QR codes. Any doubts and comments from the participants were recorded and addressed during and after responding to the scale.
On average, participants spent approximately 2–3 min finishing the questions in the scale.
After responding to the scale, the participants were required to finish LLAMA_B and LLAMA_F online within 40 min because vocabulary learning ability and grammatical inferencing ability that were tested by LLAMA_B and LLAMA_F were essential to acquiring L2 vocabulary and grammar and thus vital to acquiring L2 writing competence. Further- more, to lessen their cognitive burden, they were asked to complete the
Automated OSPAN task through Inquisit Web within 40 min the next day.
In phase two, the participants were asked to complete the writing tasks mentioned in the above section in two rounds. They completed the argumentative and narrative writing tasks in separate lessons with a break of 10 min and followed the same task order: the argumentative task in the first round and the narrative one in the second round. They were required to complete each of the writing tasks within 40 min on an online writing platform named Pigai, where they were not allowed to use spelling and grammar check. As a result, a total of 778 English essays were collected to assess the participants’ writing performance.
During the data collection, we were available online to answer the participants’ questions concerning the scale, the cognitive tests, and writing tests, to offer instructions on how to follow the procedures, and to solve the potential technical issues to keep the data collection running smoothly.
3.4. Data analysis
Prior to our analysis, the data went through screening and cleaning first, following the procedures proposed by D¨ornyei and Taguchi (2010).
In the data cleaning, 36 of the responses indicating that participants who lacked effort or intentionally misbehaved (e.g., offering the same and/or self-contradictory responses to the items of the questionnaire) were deleted from the database. The checking and cross-examination of missing data were done with the help of SPSS 25. After data screening and cleaning, 36 participants were removed from participation and thus, 353 participants were retained for the final analysis.
The correlations among the variables involved in this study were calculated by employing Pearson correlation analysis before any infer- ential statistical analysis was performed on the collected data. To address the first two research questions, path analyses were conducted to examine the effect of cognitive abilities and components of L2 grit on EFL learners’ writing achievements in narrative and argumentative tasks. To address the third research question, multiple group analyses were conducted to investigate the moderating role of task complexity in the effects of cognitive abilities and components of L2 grit on EFL learner’ writing achievement where task complexity was treated as a categorical variable (i.e., complexity +vs complexity-). The path ana- lyses and multiple group analyses were done with the help of Mplus 8.3, where two models were constructed for each research questions: Model 1 was constructed without correlating components of language aptitude
Fig. 1. Flowchart of data collection.
Fig. 2. The hypothesized path model.
Note. WMC =working memory capacity, VLA =vocabulary learning ability, GIA = grammatical inferencing ability, PE = perseverance of effort, CI = consistency of interest, QUA =quality, TC =task complexity.
and L2 grit as demonstrated in Fig. 2, while Model 2 was constructed with correlating components of language aptitude and L2 grit. The model fitnesses in the path analyses were estimated by maximum like- lihood estimation with robust standard errors, while those in multiple group analyses were by the maximum likelihood estimation procedure.
The models were evaluated with reference to the combination of goodness-of-fit indices recommended by Kline (2016): the χ2 test sta- tistic with its level of significance, the comparative fit index (CFI), the Tucker-Lewis index (TLI), the root mean square error of approximation (RMSEA), and the standardized root mean square residual (SRMR). The threshold values for CFI, TLI, RMSEA, and SRMR are .95, .95, .05, and .05, respectively. The moderating effect of task complexity was assessed by employing the bias-corrected bootstrapping procedure (Pituch et al., 2006).
4. Results
4.1. Descriptive statistics
Descriptive statics concerning language aptitude, working memory capacity, responses to the L2 Grit Scale, and human judgment of the argumentative and the narrative writing performance are reported in Table 1.
The statistics in Table 1 demonstrate that the skewness of all the involved variables ranged from − .412 to .678, which is within the range of the critical value of ±3.0, whereas their kurtosis varied from − 1.408 to .329, which falls in the range of the cut-off value of ±8. The results suggest that the data was normally distributed. Additionally, it is noticed that scores for L2 students’ grammatical inferencing ability were rela- tively lower than those for their vocabulary learning ability and the expected average, suggesting that they had relatively stronger vocabu- lary learning ability but relatively weaker grammatical inferencing ability. Besides, Analysis of Variance revealed that L2 students’ scores in argumentative writing differed significantly from those in narrative writing (F (1, 704) =41.488, p <.001).
The correlational matrix in Table 1 shows that (1) working memory capacity correlated significantly with vocabulary learning ability (r = .256, p <.01), grammatical inferencing ability (r =.212, p <.01), and quality of argumentative writing (r =.215, p <.01); (2) vocabulary learning ability correlated significantly with grammatical inferencing ability (r =.437, p <.01) and quality of argumentative writing (r = .123, p <.05); (3) grammatical inferencing ability correlated signifi- cantly with quality of argumentative writing (r =.142, p <.01); (4) perseverance of effort correlated significantly with consistency of in- terest (r =.227, p <.01) and quality of narrative writing (r =.185, p <
.01); and (5) consistency of interest correlated significantly with quality of narrative writing (r =.197, p <.01).
4.2. Results of path analyses
Table 2 shows that Model 2 fitted better with the collected data than Model 1 for an argumentative task (χ2 =48.098 <166.923, CFI =.986
>.801, TLI =.981 >.720, RMSEA =.038 <.100, AIC =8406.336 <
15,198.763, BIC =8406.336 <15,914.295) and for a narrative task (χ2
=50.926 <170.524, CFI =.982 >.796, TLI =.976 >.713, RMSEA = .029 <.101, AIC = 8098.615 < 15,563.995, BIC = 8210.742 <
15,718.654). The final path models for argumentative and narrative tasks and their standardized path coefficients are shown in Fig. 3 and Table 3, respectively.
Table 3 reveals that in the argumentative task, the effects of perse- verance of effort and working memory capacity on the overall quality of writing performance were statistically significant (βPE =.108, p =.046, βWMC =.185, p <.001), whereas the impacts of vocabulary learning ability, grammatical inferencing ability, and consistency of interest on that quality were not significant (βVLA =.015, p =.527, βGIA =.088, p = .135, βCI = − .038, p =.407). In contrast, in the narrative task, the in- fluences of perseverance of effort and consistency of interest on the overall quality of writing performance were significant (βPE =.123, p = .046, βCI =.141, p =.019), whereas the impacts of vocabulary learning ability, grammatical inferencing ability, and working memory on that quality were not significant (βVLA =.061, p =.293, βGIA = − .061, p = .134, βWMC =.056, p =.268).
4.3. Results of multiple group analyses
The statistics in Table 4 show that Model 2 fitted better with the collected data than Model 1 (χ2 =195.719 <440.443, CFI =.951 >
.808, TLI =.944 >. 772, RMSEA =.047 <. 095, AIC =17,986.537 <
32,912.736, BIC =18,223.637 <33,250.148).
The statistics in Table 5 demonstrate that task complexity did play a moderating role in the respective influences of grammatical inferencing ability and working memory capacity on the overall quality of writing performance (tGIA = − 2.12, p =.034; tWMC = − 2.206, p =.027) rather than in the influences of vocabulary learning ability, perseverance of effort, and consistency of interest on that quality (tVLA =.147, p =.883;
tPE = − .203, p =.839; tCI =1.911, p =.056). The moderating effects of task complexity were well supported by the 95% bias-corrected confi- dence intervals: The confidence interval for the moderating effect of task complexity on the relationship between grammatical inferencing ability and writing performance and on the relationship between working memory capacity and writing performance went from − .436 to − .006 and from − .109 to − .008, which excluded the cut-off value of zero.
Table 1
Descriptive statics of involved variables and correlation matrix.
Mean SE Skewness Kurtosis Min. Max 1 2 3 4 5 6
1.WMC 52.235 17.235 − .412 −.816 10 75
2.VLA 10.249 6.577 .156 −1.408 1 20 .256**
3.GIA 6.731 4.605 .678 −.477 1 19 .212** .437**
4.PE 9.224 2.5132 .145 −.337 3 15 .022 −.02 −.031
5.CI 12.87 2.662 .051 −.593 7 20 .024 .017 −.075 .227**
6.QUA (argumentation) 76.251 6.868 − .188 −.491 56 95 .215** .123* .142** .091 −.007
7. QUA (narration) 73.303 5.173 .154 −.14 61 95 .06 .037 −.068 .185** .179** .192**
PE1 3.215 .984 − .196 −.875 1 5
PE2 3.292 .848 − .146 −.111 1 5
PE2 3.198 .846 .034 −.089 1 5
PE4 3.161 .888 − .077 −.843 1 5
CI1 3.258 .994 − .133 −.539 1 5
CI2 2.816 1.01 .358 −.708 1 5
CI3 3.15 1.027 − .145 −.718 1 5
Note. WMC =working memory capacity, VLA =vocabulary learning ability, GIA =grammatical inferencing ability, PE =perseverance of effort, CI =consistency of interest, QUA =quality, **p <.01, *p <.05.
5. Discussion
This study was designed to scrutinize the interplay among learners’ language aptitude and working memory, grit, and writing performance in tasks with different complexities.
The first two research questions are concerned with the predictive effects of cognitive abilities and components of L2 grit on EFL learners’
writing performance in an argumentative and a narrative task. Results of these path analyses demonstrated that the effects of perseverance of effort and working memory capacity on EFL learners’ writing perfor- mance were significant in the argumentative task, whereas the in- fluences of perseverance of effort and consistency of interest on EFL learners’ writing performance were significant in the narrative one. To start with, the finding of the positive effects of perseverence of effort on writing performance in the L2 learner group is consistent with Sudina et al. (2020) and Teimouri et al. (2020), where perserverance of effort was found to be positively correlated with L2 learners’ language achievement. The significant positive impact of consistency of interest found in the less complex narrative task rather than in the complex argumentative task suggests that with task complexity increasing, the accountability of consistency of interest on writing performance might weaken. Besides, this study may provide partial evidence for the sub- stitutive or complementary relationship between cognitive abilities and grit assumed in the skill substitution hypothesis. The above-mentioned results of path analyses might imply the substitutive relationship be- tween working memory capacity and components of grit for less com- plex writing tasks and the synergistic relationship between working memory capacity and grit for complex writing tasks. Specifically, when achieving the success in less complex writing tasks, both factors of learners’ L2 grit (i.e., perseverance of effort and consistency of interest) are more critical than their working memory capacity. In contrast, when achieving the same success in more complex tasks, learners’ persever- ance of effort and working memory capacity might be equally impor- tant. The substitutive relationship and the synergistic relationship Table 2
Model fit indices for the path models.
Genre Model χ2 df CFI TLI RMSEA SRMR AIC BIC
ARG 1 166.923 37 .801 .720 .100 .096 15,759.636 15,914.295
2 48.098 39 .986 .981 .026 .038 8294.209 8406.336
NAR 1 170.524 37 .796 .713 .101 .096 15,563.995 15,718.654
2 50.926 39 .982 .976 .029 .039 8098.615 8210.742
Note. ARG =argumentation, NAR =narration.
Fig. 3. The final specified path models for two writing tasks.
Notes. PE =perseverance of effort, CI =consistency of interest, WMC =working memory capacity, VLA =vocabulary learning ability, GIA =grammatical infer- encing ability. Fig. 3 (a) for argumentative task, Fig. 3 (b) for narrative task.
Table 3
Standardized path coefficients of path models of cognitive abilities and com- ponents of L2 grit on quality.
Paths β SE Z p
Argumentation
VLA→QUA .036 .057 .632 .527
GIA→QUA .088 .061 1.493 .135
WMC→QUA .185 .05 3.709 <.001
PE→QUA .108 .056 1.941 .046
CI→QUA −.038 .061 -. 829 .407
Narration
VLA→QUA .061 .058 1.051 .293
GIA→QUA −.061 .06 − 1.499 .134
WMC→QUA .056 .05 1.108 .268
PE→QUA .123 .062 1.999 .046
CI→QUA .141 .06 2.338 .019
Note. VLA =vocabulary learning ability, GIA =grammatical inferencing ability, working memory capacity, PE =perseverance of effort, CI =consistency of in- terest, QUA =quality.
Table 4
Model fit indices for the path models.
Model χ2 df CFI TLI RMSEA SRMR AIC BIC
1 440.443 106 .808 .772 .095 .092 32,912.736 33,250.148
2 195.719 110 .951 .944 .047 .048 17,986.537 18,223.637
revealed in this study might contribute to our understanding of in- teractions between cognitive and noncognitive individual differences (i.
e., cognitive factors and motivational variables) in L2 writing.
The third research question is concerned with the moderating effect of task complexity on the relationship between cognitive abilities and components of L2 grit and EFL learners’ writing performance. Results of multiple group analyses revealed that task complexity moderated the respective impacts of grammatical inferencing ability and working memory on EFL learners’ writing performance. To begin with, the moderating role of task complexity in the relationship between gram- matical inferencing abilities and EFL writing performance could be interpreted as a function of the communicative functions of different writing tasks assigned (i.e., argumentative and narrative tasks). It has been acknowledged that the communicative functions assumed to be achieved by different genres would require different language use (Biber
& Conrad, 2009; Yoon, 2021; Zhang, 2021, 2022). Yoon (2021) revealed
that L2 students have an awareness of the different language use required by argumentative and narrative tasks and that they show higher nominal complexity in the argumentative task than in the narrative one. Producing high nominal complexity would be highly dependent on L2 students’ grammatical inferencing ability. Therefore, grammatical inferencing ability might play a more significant role in argumentative writing than in narrative one. Furthermore, the findings that task complexity moderated the effect of working memory capacity on EFL learners’ writing performance might be partially compatible with Zalbidea (2017), who revealed a significant correlation between working memory capacity and linguistic features in more complex tasks rather than those in less complex ones. The relevance of working memory capacity to more complex writing tasks would be interpreted as a function of its limited capacity and L2 students’ allocation of atten- tional resources. As Robinson (2011) argued, increasing cognitive complexity of a task would exert impacts on L2 production process and input processing (i.e., allocating attentional and memory resources to input and retaining that input). Working memory is assumed to be involved in all the phases of the writing process (Kellogg, 1996). Logical causal reasoning required in the argumentative task would demand L2 students to allocate more attentional resources in certain phases and thus increase their cognitive load, which might influence L2 writing performance. In addition, this study offered more evidence to support the significant effect of working memory capacity on writing perfor- mance in argumentative writing, contrary to Lu (2010), who found that working memory did not exert such influence on the quality of L2 learners’ argumentative writing. This inconsistency might be caused by different writing prompts and different foci of the scoring rubrics adopted. In our study, the prompt for the argumentative writing task was adopted from the old item pool of CET 4, and the participants’ performance was evaluated by using Jacob et al.’s (1981) ESL Compo- sition Profile, including content, organization, language, vocabulary, and mechanics with different weights, as shown in the method section.
In contrast, in Lu (2010), the writing prompt of the argumentative task was selected from the TWE writing tasks, and L2 learners’ writing per- formance was assessed in terms of content, organization, and language use.
6. Conclusion
This study explored the interplay among L2 learners’ cognitive abilities and grit in predicting foreign language writing performance along different task complexities through path analysis and multiple group analysis. Results revealed that (1) the effects of perseverance of effort and working memory capacity on L2 writing performance were significant in the argumentative task, whereas the influence of perse- verane of effort and consistency of interest on L2 writing performance were significant in the narrative one; (2) task complexity moderated the respective impacts of grammatical inferencing ability and working memory on L2 writing performance.
The findings from the study might have implications for the teaching of L2 writing. The positive predictive effect of perseverance of effort on L2 learners’ writing performance in both argumentative and narrative tasks suggests that teachers could utilize different interventions to maintain or strengthen L2 learners’ perseverance to improve their L2 writing quality. In the same vein, the positive predictive effect of con- sistency of interest on L2 learners’ writing performance in narrative tasks suggests that teachers might employ different techniques to keep learners’ interest in English learning constant and thus improve their writing performance in narrative tasks. This is because having a positive mindset is significant to foreign language learning success (Li et al., 2023; Zhang et al., 2022). Similarly, the positive predictive effect of working memory capacity on L2 learners’ writing performance in complex tasks rather than in less complex tasks might also suggest that teachers could take advantage of students’ working memory and manipulate the complexity of writing tasks to better predict students’ writing achievements.
Admittedly, there may be some limitations because of the constraints of experimental methods and conditions as well as available resources.
We only focused on two components of language aptitude (i.e., vocab- ulary learning ability and grammatical inferencing ability) and working memory capacity in an L2 writing context because of the limited Internet resources that might make it difficult for the participants to access these online tests, especially during the time of the COVID-19 pandemic.
Therefore, it is expected that the follow-up studies would entail more components of language aptitude and different components of working memory. Furthermore, the participants involved in this study were almost intermediate L2 learners, and thus the findings of this study could not be generalized to the cohorts of advanced and low-level L2 learners.
Accordingly, the results of path analyses and multiple group analyses should be verified with these groups of learners uncovered in this study.
In addition, this study just focused on writing tasks in two genres:
Table 5
Moderating effect of task complexity.
Paths Genre Complexity b Difference BC 95% CI
B t p L U
VLA→QUA Argumentation Complexity+ .038 .011 .147 .883 −.13 .155
Narration Complexity- .049
GIA→QUA Argumentation Complexity+ .131 − .233 − 2.12 .034 −.435 − .006
Narration Complexity- − .101
WMC→QUA Argumentation Complexity+ .074 − .057 − 2.206 .027 −.109 − .008
Narration Complexity- .017
PE→QUA Argumentation Complexity+ 1.436 − .182 − .203 .839 −1.925 1.647
Narration Complexity- 1.255
CI→QUA Argumentation Complexity+ − .416 1.529 1.911 .056 −.062 3.804
Narration Complexity- 1.113
Note. VLA =vocabulary learning ability, GIA =grammatical inferencing ability, working memory capacity, PE =perseverance of effort, CI =consistency of interest, QUA =quality, A =argumentation, N =narration.