1. Introduction
The emergence of educational technologies provides unprecedented opportunities for language learning (Yang et al., Reference Yang, Wen and Song2023). In the technology-enhanced language learning (TELL) environment, learners are highly autonomous in managing and progressing their learning paces and personalizing learning resources; thus, effective participation in TELL requires learners to develop self-regulated learning (SRL) competence to set goals, manage time, and monitor progress (Yang et al., Reference Yang, Wen and Song2023).
SRL refers to learners’ ability to monitor, regulate, and control their cognition, behavior, and affect to achieve learning goals (Sharma et al., Reference Sharma, Nguyen and Hong2024). Extensive research has demonstrated the effectiveness of SRL in language learning, leading to growing scholarly interest in self-regulated language learning (SRLL), particularly in English as a foreign language (EFL) contexts (Zhang & Zou, Reference Zhang and Zou2024). Recent studies on technology-mediated SRLL indicate that TELL environments offer affordances that support SRLL strategy development, sustain learner engagement, and facilitate decisions about when and where to learn (Yang et al., Reference Yang, Wen and Song2023).
With the increasing emphasis on collaborative learning in TELL, SRLL is no longer confined to individual learners but is shared among peers (Sharma et al., Reference Sharma, Nguyen and Hong2024). Collaborative learning is valued for enabling learners to compensate for individual weaknesses, enhance critical thinking, and develop problem-solving skills (Qi & Derakhshan, Reference Qi and Derakhshan2025; Zhang et al., Reference Zhang, Zou, Cheng and Xie2022). In this case, the concept of technology-enhanced collaborative self-regulated language learning (T-CSRLL) refers to situations in which learners maintain individual responsibility for regulating their learning while engaging in collaborative activities that support interdependent regulation toward shared goals (Sharma et al., Reference Sharma, Nguyen and Hong2024). In T-CSRLL contexts, core SRLL strategies – such as planning, monitoring, control, and reflection – remain individually enacted but are shaped and enriched through interaction with peers (Zhang & Zou, Reference Zhang and Zou2024). This means that collaboration functions as a regulatory affordance that externalizes thinking, provides feedback, and supports strategy adaptation in response to evolving task demands, rather than replacing learners’ self-regulation. From this perspective, socially shared regulatory processes therefore mediate and strengthen individual self-regulation by enabling learners to negotiate, realign, and adapt strategies in collaborative contexts (Su et al., Reference Su, Li, Hu and Rosé2018). Previous studies investigating the learning effectiveness of T-CSRLL have focused on areas such as social regulation and modifications, collaboration strategies (Qi & Derakhshan, Reference Qi and Derakhshan2025), SRL skills and strategies, teacher support (Wang et al., Reference Wang, Zhou, Chen, Tong and Yang2024), learning motivation and autonomy, and language learning performances (Sharma et al., Reference Sharma, Nguyen and Hong2024).
However, relatively few studies have examined technology-mediated SRLL behaviors, particularly in collaborative settings, limiting understanding of how regulation differs between individual and collaborative learning processes. An empirical experiment in this domain is timely to address several gaps in the literature. First, although SRLL strategies in individual contexts are well documented, little is known about how these strategies are jointly constructed in TELL environments. This is important because collaboration redistributes regulatory control across learners, with regulatory decisions often emerging through interaction rather than individual monitoring. Second, despite evidence that TELL can enhance language learning outcomes, little is known about how SRLL in individual and collaborative contexts influences learners’ performance in TELL environments. This warrants further investigation because it depends not only on technological features but also on how learners regulate their learning. Third, prior research has largely focused on static behavioral patterns, with limited attention to the dynamic regulatory behaviors that emerge through peer interaction in T-CSRLL contexts. Addressing these gaps, the present study investigates whether learners’ SRLL strategy use, learning performance, and learning behaviors are influenced by peer collaboration in technology-enhanced environments.
To this end, the study adopts WeChat, a widely used mobile application in Chinese-speaking communities, to design T-CSRLL activities for Chinese university EFL learners in a writing course. WeChat provides an interactive learning and communication environment (Jin, Reference Jin2018). Additionally, WeChat affordances, including chat logs, (a)synchronous interaction, group chats, and multimodal feedback (e.g., video, voice, message, image), enable learners to track task progress and support regulatory behaviors such as group negotiation, peer monitoring, and continual reflection based on shared feedback. Answers to the following questions are sought:
-
1. Does learners’ SRLL strategy use differ in the WeChat-based collaborative and individual SRL contexts?
-
2. Does learners’ writing performance differ in the WeChat-based collaborative and individual SRL contexts?
-
3. Do learners’ learning behaviors differ in the WeChat-based collaborative and individual SRL contexts?
Answering these questions can offer practical implications for designing mobile-based collaborative activities and inform educators on how to tailor learning supports in individual and collaborative learning environments that foster SRLL. Additionally, analyzing learners’ dynamic self-regulatory behavioral patterns in TELL can enrich the current understanding of SRLL from an individual model to a socially shared regulation process.
2. Literature review
2.1 Self-regulated language learning (SRLL)
SRL refers to learners’ use of strategies to plan, monitor, and control their cognition, affect, and behavior in order to achieve learning goals (Sharma et al., Reference Sharma, Nguyen and Hong2024; Zimmerman, Reference Zimmerman2002). In language learning, SRL has been shown to enhance learners’ attitudes, motivation, self-efficacy, and overall learning effectiveness. Consequently, increasing attention has been given to cultivating learners’ SRLL abilities to support active and lifelong language learning (Zhang & Zou, Reference Zhang and Zou2024).
The theoretical foundation of SRL is commonly grounded in Zimmerman’s (Reference Zimmerman2002) cyclical model, which includes three phases: forethought, performance, and self-regulation. These phases describe how learners set goals and plan learning, apply and monitor strategies during task execution, and evaluate outcomes to adjust subsequent learning behaviors. In language learning contexts, this model guides learners to plan language knowledge acquisition, monitor their learning performances, and reflect on learning outcomes while analyzing factors contributing to success or failure.
As SRL research has expanded, its application to language learning has led to varied classifications of SRLL strategies. This study adopts the inclusive framework proposed by Zhang and Zou (Reference Zhang and Zou2024), which synthesizes SRL research in language learning and categorizes SRLL strategies into metacognitive, cognitive, motivational, and behavioral dimensions. Metacognitive strategies enable learners to monitor errors, regulate learning processes, and manage task completion (Zimmerman & Moylan, Reference Zimmerman, Moylan, Hacker, Dunlosky and Graesser2009). Cognitive strategies refer to knowledge and skills that learners employ to complete learning tasks (Zhang & Zou, Reference Zhang and Zou2024; Zhang et al., Reference Zhang, Zou and Cheng2025). Motivational strategies help learners manage and control their emotions and efforts while learning, thereby sustaining engagement in learning. Behavioral strategies support learners’ construction and maintenance of conducive physical and social learning environments (Zhang & Zou, Reference Zhang and Zou2024; Zhang et al., Reference Zhang, Zou and Cheng2026).
A growing number of research studies have consistently demonstrated the positive effects of SRLL strategies on language development, particularly writing performance. Typically, Shen and Bai (Reference Shen and Bai2024) conducted a quasi-experimental study with Chinese university students, showing that SRLL strategy-based instructions significantly improved the students’ writing performance and strategy use, with sustained effects over time. Similarly, Teng and Zhang (Reference Teng and Zhang2020) reported that SRL-based writing instructions enhanced undergraduates’ strategic engagement as well as their linguistic and performance self-efficacy. Teng and Zhang (Reference Teng and Zhang2018) further showed that cognitive and metacognitive strategies mediated English writing performance, while motivational regulation functioned as an antecedent influencing the use of other SRL strategies. Teng (Reference Teng2020) demonstrated that integrating collaborative modeling of text structure into SRLL strategy improved both reading and writing skills among primary learners, while Teng et al. (Reference Teng, Wang and Zhang2022) identified planning, monitoring, and evaluating as core predictors of writing performance in young EFL learners using a validated self-regulated writing strategy questionnaire. Collectively, these findings suggest that SRLL strategies facilitate language learning through complementary mechanisms. Metacognitive strategies can help students identify errors and complete learning tasks on time; cognitive strategies prompt learners to internalize learning materials; motivational strategies can trigger and maintain learners’ interest and effort to learn; and behavioral strategies can enhance learners’ access to instructional support and opportunities for productive use of language (Zhang & Zou, Reference Zhang and Zou2024).
2.2 Technology-enhanced collaborative SRLL (T-CSRLL)
Educational technologies play an important role in language learning (Yang et al., Reference Yang, Wen and Song2023). In technology-enhanced contexts, learners are expected to be self-regulated learners possessing strategic thinking, metacognition, time and resource management skills, and high motivation to effectively engage with digitally mediated learning (Wang et al., Reference Wang, Zhou, Chen, Tong and Yang2024). Accordingly, increasingly more researchers and educators have conducted studies on technology-enhanced SRLL, which encourages learners to engage in self-initiated use of technologies to create optimal and self-fulfilling language learning experiences. Previous studies have integrated technologies into SRLL instruction based on Zimmerman’s (Reference Zimmerman2002) cyclical model, embedding digital tools to support learners’ language development and their use of metacognitive, cognitive, and resource management strategies (Tran & Ma, Reference Tran and Ma2025; Yang et al., Reference Yang, Wen and Song2023). Empirical evidence indicates that technology-enhanced SRLL instructions can lead to learners’ improvements in vocabulary efficiency (Chen et al., Reference Chen, Wang and Chen2014; Chen & Lee, Reference Chen and Lee2018), writing scores and writing efficacy, and the better use of SRL strategies compared with when they did not receive technology-enhanced SRLL instructions (Chen & Lee, Reference Chen and Lee2018; García Botero et al., Reference García Botero, Botero Restrepo, Zhu and Questier2021; Tran & Ma, Reference Tran and Ma2025).
While early SRLL research primarily emphasized individual cognitive processes, recent studies have increasingly acknowledged the social dimension of self-regulation in collaborative learning environments (Law et al., Reference Law, Ge and Eseryel2016). Although collaboration in TELL has been shown to enhance learners’ language knowledge, collaboration skills, reasoning, and problem-solving abilities (Su & Zou, Reference Su and Zou2022; Zhang et al., Reference Zhang, Zou, Cheng and Xie2022), only a few empirical studies have investigated the effects of collaborative SRLL strategies on language development, particularly in technology-enhanced environments. For instance, Ma and Chiu (Reference Ma and Chiu2024) developed a self-regulated, collaborative, and personalized vocabulary learning approach in mobile environments, which improved learners’ productive vocabulary knowledge and fostered SRLL through community sharing. Similarly, Su et al. (Reference Su, Li, Hu and Rosé2018) investigated the occurrence of self-regulation and social regulation during wiki-based collaborative reading activities in EFL contexts, finding that learners actively participated in collaborative activities, and the degree of social regulation was impacted by learner proficiency; that is, high achievers tended to apply a more continuous and smooth regulatory pattern while lower achievers were likely to be lost in a single repeated pattern.
Despite evidence supporting the benefits of technology-enhanced SRLL, most existing studies have focused on individual regulation, with relatively few examining how collaboration influences language development and regulatory processes in T-CSRLL contexts (e.g., Ma & Chiu, Reference Ma and Chiu2024). This constitutes a critical gap because collaborative SRLL involves multiple members to regulate their collective learning process, which is more facilitative to learning because of joint negotiation, realignment, or adaptation compared to individual self-regulation (Su et al., Reference Su, Li, Hu and Rosé2018). Moreover, educational technologies such as mobile applications (e.g., WeChat) offer new chances for extending collaborative language learning beyond the classroom (Tran & Ma, Reference Tran and Ma2025). However, even though understanding how learners regulate their behaviors over time is crucial for revealing how learners manage and adapt to individual and collaborative SRLL activities supported by technologies, research examining learners’ dynamic sequences of learning behaviors in T-CSRLL settings remains scarce. Addressing these gaps, the present study investigates EFL learners’ SRLL strategy use, writing performance, and learning behavior patterns within a WeChat-based collaborative self-regulated writing program.
2.3 Learning behaviors in T-CSRLL
Most research on technology-enhanced SRLL has relied on questionnaires, interviews, and self-report measures to assess learners’ self-regulatory abilities across technology-supported learning contexts (Yang et al., Reference Yang, Wen and Song2023). However, such approaches often neglect learners’ dynamic processes of learning behaviors as they unfold during task completion (Su et al., Reference Su, Li, Hu and Rosé2018). Since different behavior patterns may have different influences on learning performances, researchers are encouraged to adopt a sequential mining approach to capture learners’ actual learning processes in technology-mediated environments and to inform instructional design more precisely (Hwang et al., Reference Hwang, Hsu, Lai and Hsueh2017). Although limited in number, existing studies employing sequential analyses provide valuable insights into learners’ dynamic learning behaviors. Hwang et al. (Reference Hwang, Hsu, Lai and Hsueh2017) designed a problem-based English listening game, and this gaming approach benefited learners’ learning achievement and motivation. Moreover, learners with higher levels of English anxiety performed more complex learning and gaming behaviors and achieved better learning outcomes than those with lower levels of anxiety. In a subsequent study, Chen and Lee (Reference Chen and Lee2018) found that the application-driven game greatly enhanced learners’ flow experience and learning self-regulation, and learning regulation behaviors enabled them to proceed at individual paces and make decisions about planning and reviewing vocabulary. Focusing on the impact of technologies on learner behaviors, Yang et al. (Reference Yang, Chang, Hwang and Zou2020) conducted a lag sequential analysis to analyze behavioral patterns of learners using and not using a cognitive complexity-based competition game (CLBG), finding that the CLBG affected learners’ vocabulary performances by upgrading or downgrading their gaming levels to ensure individuals learned with tasks suitable for their proficiency levels. Similarly, Hwang et al. (Reference Hwang, Chien and Li2021) showed that learners using a multidimensional repertory grid game engaged in more diversified and higher-order behaviors, such as reviewing materials and monitoring learning progress, than those who did not use the game.
Despite these contributions, prior studies have rarely examined changes in behavioral patterns during collaborative SRLL tasks. As behavioral sequences and transitions may differ substantially between individual and collaborative learning, further research in T-CSRLL contexts is needed to illuminate how learners regulate their learning during collaboration and to inform the design of adaptive instructional support. Therefore, this study addresses the abovementioned research gaps by exploring EFL learners’ dynamic behavioral patterns in T-CSRLL and identifying the behavior differences when they participate in collaborative and individual technology-enhanced SRLL.
3. WeChat-based collaborative self-regulated English writing program
3.1 Design principles
In this study, WeChat was adopted as the mobile learning platform to design a collaborative self-regulated English writing program for Chinese EFL learners, primarily due to its accessibility, timely communication features, and pedagogical affordances. As an integral part of daily life in China, WeChat enables users to communicate through text, voice, and group chats; seek assistance; share learning resources; submit assignments; and access online learning materials via official accounts, educational videos, and audio resources (Jin, Reference Jin2018).
As shown in Figure 1, the WeChat-based self-regulated English writing instruction program was designed based on the cyclical phase model of SRL (Zimmerman & Moylan, Reference Zimmerman, Moylan, Hacker, Dunlosky and Graesser2009), first principles of instruction (FPI) theory (Merrill, Reference Merrill2012), and self-regulated strategy development (SRSD) theory (Harris & Graham, Reference Harris, Graham, Fidalgo, Harris and Braaksma2017).
The design framework.
The writing process (pre-writing, while-writing, and post-writing) was aligned with SRL theory and implemented across five lessons, each completed within a 90-minute weekly session. To ensure consistent learning schedules, all activities were completed during class time, and the platform was closed after class. Within each session, learners could freely access learning materials without restrictions on time or frequency. A recommended learning sequence (pre-lesson video, pre-task survey, content learning, and quiz completion) was provided, while learners retained autonomy to adjust the order based on individual needs.
Lessons 1 and 2 focused on the pre-writing stage (forethought phase), comprising activities such as deconstructing sample essays, searching topic-related information, planning content, and learners motivating themselves via expected rewards. These activities operationalized metacognitive (e.g., goal setting), cognitive (e.g., reviewing), behavioral (e.g., resource management), and motivational strategies.
Lesson 3 targeted the while-writing stage (performance phase), during which learners drafted their essays while applying strategies related to self-control and self-observation. Tasks included rehearsing language resources, drafting logically structured texts, monitoring progress with writing checklists, managing time and learning environments, regulating emotions, and engaging in peer interaction and contribution making in the collaborative condition.
Lesson 4 addressed the post-writing stage (self-reflection phase), guiding learners to evaluate and revise their drafts. Learners assessed their writing using structured checklists and criteria, refined their drafts, and compared their drafts with peers’ writings. These activities supported self-judgment and self-reaction through strategies such as self-evaluation and feedback handling.
Lesson 5 served as a booster session, allowing learners to review and consolidate all previously introduced SRLL strategies. This session reinforced strategy awareness and encouraged learners to transfer regulatory practices across writing tasks.
In the individual learning condition, learners completed all activities independently. In the collaborative learning condition, learners engaged in the same lessons and tasks while seated in small groups. Although learners progressed through the WeChat program individually, they used group chats and document-sharing functions to exchange ideas, discuss task requirements, compare plans, and provide peer feedback.
The FPI theory structured the functions of this WeChat program into five stages: (1) Activation – activating students’ prior knowledge of SRLL strategies; (2) Demonstration – presenting SRLL strategies for each writing stage; (3) Application – implementing strategies in class tasks and authentic writing; (4) Integration – applying learned strategies to authentic prompts like IELTS tasks; and (5) Reflection – focusing on consolidating learning through quizzes, performance review, and peer discussion via WeChat group discussions.
The SRSD theory guided instructional design, including (1) Develop Background Knowledge – discussing prior SRLL knowledge and setting goals; (2) Model, Discuss, Memorize, Support – demonstrating and guiding SRLL strategy use; (3) Independent Performance – responding to authentic writing prompts; and (4) Reflection – evaluating writing and reflecting on learning progress.
3.2 Functions of the program
The WeChat-based program incorporated several modules (Figures 2–5) to support learners’ SRLL processes.
“Watch,” “Learn,” and “Try” functions.
“My Prior Experience” function.
“My Comment” function.
“My Zone” functions.
As shown in Figure 2, the “Watch It” function allows learners to watch an introductory video clip before each lesson. The video provides an overview of the content and learning objectives for each lesson, helping to activate learners’ prior knowledge of SRLL. The “Learn It” module presents the instructional materials for each lesson, guiding learners to apply SRL strategies. Learners can learn at their own pace and review the instructional materials as needed. The “Try It” function delivers quizzes designed to enhance learners’ understanding of the SRLL strategies covered in each lesson. Learners receive immediate feedback and points upon completing the quizzes, which include true or false and multiple-choice questions. The administration system records learners’ first score when they take the quiz, while allowing them to repeatedly review the quiz.
“My Prior Experience” (Figure 3) function guides students to recall and share their previous experiences with specific SRL strategies that will be instructed in the lesson. This function can activate students’ background knowledge and help prepare them for the upcoming instruction.
After each lesson, “My Comment” function (Figure 4) allows learners to self-assess their learning progress, providing feedback ranging from “very satisfied” to “not satisfied at all.”
“My Zone” module (Figure 5) tracks learners’ learning behaviors, including the progress bar function, points, and the leaderboard. The progress bar monitors task completion, the points system provides rewards for learners’ performances in quizzes, and the leaderboard fosters a competitive learning environment by displaying rankings based on points and cumulative learning time.
Additionally, the instruction program leveraged the WeChat group chat feature. This allowed students to post questions, engage in discussions, share learning activities, and exchange experiences throughout the self-regulated English writing instruction processes.
4. Methods
4.1 Participants
A total of 63 first-year EFL learners (53 males and 10 females), aged between 17 and 21, participated in this study. They were recruited from two intact classes at a comprehensive university in southwest China. Intact classes were used to implement the WeChat-based SRLL program in an authentic classroom context and to preserve learners’ natural instructional experiences. Participants were randomly assigned to either the control group – WeChat-based individual learning group (WIG; N = 30) – or the experimental group – WeChat-based collaborative learning group (WCG; N = 33). In the WCG condition, learners were organized into small groups according to preestablished criteria, including heterogeneous English proficiency and a basic level of mutual familiarity to facilitate effective collaboration (Su et al., Reference Su, Zou and Xie2024). In total, 11 heterogeneous groups of three learners were formed. During in-class activities, the instructor monitored group interaction by encouraging the less active learners to participate and guiding more active members to support their peers when necessary.
Both groups were instructed by the same teacher, who held a doctoral degree in the field of TELL, and followed identical instructional materials and procedures. The only difference between the groups was the mode of SRLL implementation (individual SRLL for WIG and collaborative SRLL for WCG). Pretest results on IELTS writing performance and SRLL strategy use confirmed baseline equivalence between the two groups, thereby mitigating potential selection bias. Throughout the intervention, the teacher monitored the WeChat-based program and offered assistance when necessary. All participants provided written informed consent to participate voluntarily and to allow the use of their data for research purposes.
4.2 Research procedures
This seven-week quasi-experiment followed the procedures shown in Figure 6. In Week 1, participants completed an argumentative writing task on “Environmental Problem” as the pretest and responded to the SRLL strategy use questionnaire.
Research procedures.
From Weeks 2 to 6, all learners received short videotaped SRLL lessons and completed two writing tasks. Each SRLL lesson targeted a specific stage of the writing process (pre-writing, while-writing, and post-writing) and corresponding regulatory strategies. To ensure experimental control, learning sessions were inaccessible after class. Both groups received identical instructional content; the only difference was the mode of SRLL implementation, with learners in the WIG completing tasks individually and those in the WCG engaging in collaborative learning and writing. This four-week practice phase aimed to consolidate SRLL knowledge and facilitate its integration into learners’ writing processes.
Across the four instructional weeks, learners completed writing tasks (Topics: “Global Village”) based on the WeChat-based SRLL instructions. During Weeks 2 and 3 (Lessons 1–2), instruction focused on pre-writing strategies, including planning, organizing, and drafting content. Example activities included brainstorming, mind mapping, and structuring ideas accordingly.
In Week 4, learners completed Lesson 3, which emphasized applying SRLL strategies during the while-writing stage. Tasks focused on organizing pre-writing content and drafting essays, such as completing three “tree tables” to structure body paragraphs.
In Week 5, Lesson 4 addressed post-writing strategies related to self-evaluation and feedback processing. Learners were asked to evaluate their drafts using specified writing criteria and revise their work accordingly.
In Week 6, Lesson 5 was the booster session based on Harris and Graham’s (Reference Harris, Graham, Fidalgo, Harris and Braaksma2017) SRSD, aiming to enhance learners’ application of SRLL strategies they had learned in the previous three weeks. Learners were asked to review all SRLL strategies and complete a new writing task on the topic of “Recycling: A Necessary Action.”
In the last week, all learners applied SRLL strategies to individually compose their writing on the topic of “Aim Achieving: Luck Matters?” as the posttest, and they responded to the SRLL questionnaire again.
4.3 Data collection and analysis
4.3.1 SRLL questionnaire
Learners’ SRLL strategy use was measured using Writing Strategies for Self-Regulated Learning Questionnaire (Teng & Zhang, Reference Teng and Zhang2016; see Appendix A in the supplementary material). The instrument consists of 40 items (α = 0.90) covering four categories: “cognitive strategies” (N = 9, α = 0.70), “metacognitive strategies” (N = 9, α = 0.74), “motivational regulation strategies” (N = 15, α = 0.90), and “social behavioral strategies” (N = 7, α = 0.70). All items were rated on a 4-point Likert scale (1 = completely disagree to 4 = completely agree).
The questionnaire was administered in its original English version without modification, as it was designed to measure EFL learners’ writing SRLL strategies, which aligned with the aims of the present study. Participants were familiar with English-medium instruction and therefore able to complete the questionnaire in English. To ensure clarity, the instrument was reviewed by an expert in English language education, and brief in-class explanations were provided prior to administration. Independent-samples t tests indicated no significant pretest differences between groups across all SRLL dimensions (ps > 0.05).
4.3.2 Writing performance
Learners’ writing proficiency was assessed through pretest and posttest writing tasks. Writing topics were selected from IELTS Academic Writing Task 2. All prompts from Tests 4–18 were reviewed, from which eight topics were preliminarily shortlisted. Learners then voted to select four preferred topics – (1) Environmental Problem, (2) Global Village, (3) Recycling: A Necessary Action, and (4) Aim Achieving: Luck Matters? – which were finalized for writing tasks. Topics (1) and (2) were used for the pretest and posttest, respectively, while the remaining two were used during the practice phase. For each task, learners were required to produce an argumentative essay of at least 250 words within 40 minutes.
Learners’ writing was evaluated using the IELTS Writing Assessment Criteria (2023 old version; see Appendix B in the supplementary material), which includes four dimensions: task response, coherence and cohesion, lexical resource, and grammatical range and accuracy. Each dimension is scored on a 9-point scale, and the overall writing score is calculated as the mean of the four subscores, also ranging from 0 to 9. To ensure scoring reliability, three external EFL writing instructors with experience in teaching academic writing independently rated all essays. Each script was anonymized using numerical codes to prevent rater bias. Prior to scoring, the raters were trained on the assessment criteria and jointly scored five sample essays to establish a shared understanding. They then rated the remaining essays independently, achieving high interrater reliability (α = 0.93). Any uncertainties were resolved through discussion. The final writing score for each essay was calculated as the average of the three raters’ scores. Independent-samples t tests showed that learners in both groups did not differ in writing performances in the beginning (t = 1.37, p > 0.05).
4.3.3 Codes of learning behaviors
The WeChat learning program was designed in accordance with SRL theory (Zimmerman & Moylan, Reference Zimmerman, Moylan, Hacker, Dunlosky and Graesser2009), with each function mapped to a corresponding behavioral code to support subsequent analysis. Tracking points were embedded in all functions so that every learner interaction was automatically recorded by the backend system. After the intervention, all clickstream data generated during WeChat use were exported directly from the system. As each function was assigned a unique code, all available features were operationalized as distinct behavioral indicators. Consequently, learners’ interactions with the WeChat program were systematically translated into analyzable behavioral codes (Table 1). To ensure coding reliability, we consulted an expert in the TELL field.
Codes of learning behaviors
Specifically, S_P operationalized forethought-phase regulation by capturing learners’ engagement in task analysis and strategic planning; L_L and T_Q reflected performance-phase strategy enactment through engagement with guidance. Codes relating to learning monitoring, including C_P, C_L, C_T, and C_R, were indicators of self-observation and self-monitoring, which are central components of SRL during task performance. In the self-reflection phase, S_E captured learners’ self-judgment and self-reaction with performance outcomes.
To examine differences in learning behaviors between groups, lag sequential analysis (LSA) was conducted on WeChat-based learning logs. In LSA, Z scores (adjusted residuals) indicate whether observed behavioral transitions occur more frequently than expected by chance, given the overall behavior distribution. Adjusted residual tables were generated for each group’s behavioral sequences. Transitions with Z scores greater than 1.96 were considered statistically significant (p < .05; Hwang et al., Reference Hwang, Chien and Li2021).
5. Results
5.1 SRLL strategy use levels
An ANCOVA was conducted to compare learners’ post-instruction SRLL strategy use between groups. Shapiro–Wilk tests confirmed the normality of SRLL scores in both groups (ps > 0.05). Levene’s tests indicated homogeneity of variance for overall SRLL, F(1, 61) = 1.11; metacognitive, F(1, 61) = 1.04; cognitive, F(1, 61) = 0.76; motivational, F(1, 61) = 2.89; and behavioral, F(1, 61) = 3.38; dimensions (ps > 0.05). As shown in Table 2 (Appendix C in the supplementary material), significant group differences were found in overall SRLL strategy use, F(1, 60) = 11.86, p < 0.01; metacognitive strategy, F(1, 60) = 7.19, p < 0.01; cognitive strategy, F(1, 60) = 6.45, p < 0.05; and behavioral strategy, F(1, 60) = 43.92, p < 0.001; achieving large effect sizes (Cohen, Reference Cohen1988). However, no significant difference was observed in motivational strategy use, F(1, 60) = 1.94. These results indicate that learners receiving WeChat-based collaborative self-regulated writing instruction demonstrated significantly higher SRLL strategy use than those in the individual learning condition across most strategy dimensions.
5.2 Writing performance
An ANCOVA was conducted to compare learners’ post-writing performance across learning conditions, with pre-writing scores entered as a covariate. Shapiro–Wilk tests confirmed the normality of post-writing scores in both the WIG and WCG (p = 0.09; p = 0.15). Levene’s test indicated homogeneity of variance between groups, F(1, 61) = 0.01, p = 0.09. As shown in Table 3 (Appendix C in the supplementary material), learners in the WCG significantly outperformed those in the WIG on post-writing performance, F(1, 60) = 4.85, p = 0.03, achieving a moderate effect size (partial η2 = 0.08; Cohen, Reference Cohen1988). These results suggest that the WeChat-based collaborative self-regulated writing program was more effective in enhancing EFL learners’ writing performance than the individual learning program.
5.3 Learning behavior analysis
LSA was conducted using GSEQ to identify and compare learners’ behavioral patterns across the two groups. Tables 4 and 5 (Appendix C in the supplementary material) present the adjusted residuals for the WIG and WCG, showing that 14 behavioral sequences reached statistical significance in each group. Based on these results, behavioral transition matrices and between-group differences were visualized in Figure 7, where numbers on the arrows represent Z scores and arrow directions indicate behavioral transitions.
Comparison of behavior transitions between WIG and WCG.
Thirteen statistically significant sequences were shared by both groups (L_L→L_L; L_L→S_P; S_P→S_P; S_P→T_Q; T_Q→T_Q; C_P→C_L; C_L→C_P; C_P→S_E; S_E→C_P; C_P→C_R; C_R→C_P; C_R→C_T; C_T→C_R). Two group-specific sequences were also identified: T_Q→C_P in the WIG and C_T→C_P in the WCG.
Across both groups, learners began by engaging with SRLL instructional contents (L_L), followed by transitions to goal setting and strategy planning (L_L→S_P) and quiz completion to assess understanding (S_P→T_Q). The recurrent self-transitions (L_L→L_L; S_P→S_P; T_Q→T_Q) suggest learners in two groups engaged in the same SRL processes, reflecting their continuous self-monitoring and evaluation of whether current goals and strategies needed adjustment. This pattern reflects core SRL processes and aligns with the SRSD framework, which emphasizes sustained review and scaffolded practice prior to independent performance (Harris & Graham, Reference Harris, Graham, Fidalgo, Harris and Braaksma2017).
A clear difference between individual and collaborative contexts emerged following quiz completion. In the WIG, learners transitioned from quiz completion to checking learning profiles (T_Q→C_P), a sequence absent in the WCG. This pattern suggests that learners in the individual context relied on outcome-based feedback to interpret learning results, placing greater emphasis on individualized performance summaries. In both groups, the behavior of checking profiles was the center of other related behaviors, with learners subsequently monitoring learning progress (C_P→C_L), conducting self-evaluation (C_P→S_E), and reviewing ranking (C_P→C_R), before returning to the profile checking function (C_L/S_E/C_R→C_P). This clustering around C_P indicates active engagement with feedback for monitoring performance, diagnosing learning status, and regulating subsequent actions, consistent with the monitoring and reflection phases emphasized in SRL, SRSD, and FPI theories.
Notably, after reviewing rankings (C_R), learners in both groups examined learning time (C_R→C_T) and then returned to profile checking (C_T→C_P). However, the transition C_T→C_P occurred exclusively in the WCG. This pattern suggests enhanced time regulation and feedback awareness among learners in the collaborative context, reflecting more deliberate management of learning effort (see Figure 8 and Appendix D in the supplementary material). From an FPI perspective, this pattern reflects learners’ in-depth engagement with the integration and reflection functions of the learning environment, where learners regulate not only outcomes but also learning progress. Similarly, within the SRSD framework, time monitoring represents advanced self-management and readiness for independent strategic performance.
Data logs of WCG learners’ collaboration.
6. Discussion
6.1 SRLL strategy use
The results showed that EFL learners who received WeChat-based collaborative SRLL instructions outperformed those who completed the tasks individually in terms of cognitive, metacognitive, social behavioral, and overall SRLL strategy use. This finding echoes Ma and Chiu (Reference Ma and Chiu2024), who reported that collaborative TELL environments facilitate both language learning and the development of self-regulatory knowledge. These gains can be largely attributed to the collaborative nature of the writing tasks. Collaborative learners were required to share ideas, articulate reasoning, and respond to peers, which likely increased the use of cognitive strategies such as elaboration, organization, and rehearsal. The need to coordinate collective progress and evaluate group outputs may have further promoted metacognitive regulation, while the social demands of collaboration, such as help seeking, feedback provision, and negotiation, supported social behavioral regulation that was largely absent in individual writing contexts. Unlike individual SRL, collaborative writing requires learners to align their regulatory processes with those of their peers. Accordingly, learners not only regulated their own writing behaviors but also engaged in co-regulatory actions, including prompting peers’ planning, clarifying task goals, and jointly monitoring progress (Law et al., Reference Law, Ge and Eseryel2016). Such distributed regulatory responsibility may explain the observed increase in metacognitive strategy use. Moreover, WeChat’s multimodal affordances (e.g., text messaging, voice notes, and document sharing) likely facilitated learners’ strategic use of technological resources, reinforcing both cognitive and social regulation and fostering shared responsibility for regulating individual and collective learning, thereby enhancing overall SRLL strategy use.
These explanations align with Zimmerman’s (Reference Zimmerman2002) SRL model. In the forethought phase, collaborative learners appeared more willing to engage in joint goal setting, knowledge sharing, and collective evaluation (Qi & Derakhshan, Reference Qi and Derakhshan2025). During the performance phase, WeChat-mediated interactions enabled discussion, negotiation, peer assistance, and feedback, likely promoting learners’ metacognitive awareness, and they processed information more effectively than working individually (Su et al., Reference Su, Zou and Xie2024). In the self-reflection phase, progress checking, peer feedback, and collective evaluation helped learners critically assess their learning outcomes and informed subsequent goal setting (Zhang et al., Reference Zhang, Zou and Cheng2026). This socially shared regulation enables learners to transfer SRL strategies and regulate each other’s metacognitive, cognitive, and emotional processes, in turn facilitating collaboration (Su et al., Reference Su, Li, Hu and Rosé2018).
However, no significant difference was found between learning contexts in motivational strategy use, consistent with Teng and Zhang (Reference Teng and Zhang2020), who reported that motivational regulation is less sensitive to instructional format. A key factor may be the explicit SRL strategy lessons designed in this study, which provided structured guidance on planning, monitoring, and effort regulation, enabling learners to manage their own learning. In the individual condition, learners may have relied on these strategies to sustain engagement and regulate effort during writing tasks to compensate for the absence of collaborative scaffolding. Consequently, learners in both individual and collaborative conditions demonstrated comparable levels of motivational strategy use, indicating that explicit SRL instruction can effectively support intrinsic motivation even in the absence of social interaction. Additionally, the novelty of learning via WeChat may have initially enhanced learners’ motivation across both conditions. Although WeChat collaboration provided emotional support and social presence that helped learners maintain positive emotions and enhanced their learning persistence (Zhang & Zou, Reference Zhang and Zou2024), motivational regulation involves complex self-beliefs and affective control, which may develop independently regardless of learning format (Zhang et al., Reference Zhang, Zou and Cheng2025). As a result, motivational strategy use appeared relatively stable across collaborative and individual contexts.
6.2 Writing performance
Consistent with Su and Zou (Reference Su and Zou2022), the present study found that learners in the collaborative condition outperformed those in the individual condition in writing performance. This finding aligns with prior research showing that collaborative learning fosters joint negotiation and collective regulation, leading to enhanced language performance compared to individual self-regulation (Ma & Chiu, Reference Ma and Chiu2024; Su et al., Reference Su, Li, Hu and Rosé2018). One possible reason is that the heterogeneity within collaborative groups, which exposes learners to diverse linguistic input and writing strategies, enables them to acquire a wider range of writing knowledge and integrate more accurate expressions into their writing. Higher-proficiency learners often modeled accurate language use, varied expressions, and persuasive argumentation, providing linguistic and rhetorical resources that less proficient learners could appropriate. This reciprocal exchange fostered an optimal zone of proximal development (Sharma et al., Reference Sharma, Nguyen and Hong2024), benefiting lower-achieving learners while consolidating higher achievers’ knowledge through explanation and justification of language choices.
Moreover, the social support inherent in collaboration likely sustained learners’ motivation and persistence, engaging them in iterative drafting and revision processes that are essential for producing high-quality writing. Additionally, during WeChat-based collaboration, learners frequently exchanged drafting thoughts, discussed writing organization, and negotiated language wording. These interactions required learners to plan content in advance, monitor structural coherence, and evaluate revisions against assessment criteria, thereby activating metacognitive strategies such as monitoring and evaluation (Teng & Zhang, Reference Teng and Zhang2020). Furthermore, providing and responding to peer feedback prompted learners to generate reasoned revision suggestions, engaging cognitive strategies such as elaboration, organization, and critical evaluation. At the same time, coordinating contributions, resolving disagreements, and maintaining constructive interactions drew on social behavioral strategies like criticism, turn-taking, and collaborative decision-making. Through this complex interplay of cognitive, metacognitive, and social regulation, learners can identify and address individual weaknesses, transfer knowledge within the group, and co-construct higher-quality written products than would be possible in individual learning alone.
6.3 Learning behaviors
We identified patterns about learners’ behavioral sequences within WCG and WIG, shedding light on how these behaviors reflect the enactment of SRLL strategies. In both groups, a core cyclical process emerged, in which learners progressed from lesson engagement to goal setting and strategic planning, followed by knowledge checks through quizzes. This pattern shows that learners did not passively complete writing tasks but continually regulated their writing processes. Such a recurring behavioral sequence underscores the central role of metacognitive strategies in error identification, progress monitoring, and decisions of subsequent learning actions (Zhang & Zou, Reference Zhang and Zou2024; Zhang et al., Reference Zhang, Zou and Cheng2025). This may be attributed to the combined effects of explicit SRL instructions and the familiar affordances of WeChat, which enhanced learners’ understanding of how technology can more effectively support writing regulation compared with prior writing practices that involved limited exploration upon task initiation. Consequently, learners demonstrated higher-level engagement in SRL learning and evaluation of knowledge acquisition, enabling the production of higher-quality writing outcomes. Additionally, this repetition of learning behaviors corroborates Zimmerman’s (Reference Zimmerman2002) SRL model, emphasizing the forethought and performance phases where learners plan their goals and enact learning strategies, and the self-reflection phase when they actively engage in metacognitive monitoring by evaluating outcomes and adjusting behaviors. Notably, the presence of this cycle in the individual condition indicates that SRLL can be performed independently when learners are provided with explicit SRL instructions and corresponding tasks. Under such conditions, learners can internalize self-regulatory routines in which planning and evaluation function as integral components of learning, while collaboration strengthens these regulatory processes.
Across both groups, checking learning profiles functioned as a central regulatory behavior linked to subsequent actions, including monitoring learning progress, conducting self-evaluation, and reviewing rankings. The cyclical transitions among these behaviors demonstrate a sophisticated self-regulatory pattern in which learners continuously collect performance-related information, evaluate their learning status, and benchmark progress. This reflects the multidimensional nature of SRLL strategies incorporating cognitive, metacognitive, and behavioral components (Law et al., Reference Law, Ge and Eseryel2016). Cognitively, the integration of profile checking with progress and ranking information provided a rich source of data for elaboration and comparison, allowing learners to identify specific content areas requiring further practice. Metacognitively, these cycles operationalize ongoing monitoring and strategic control, which are crucial for real-time adjustment of learning goals. Behaviorally, repeated access to performance data indicates that learners are actively managing their learning trajectories rather than passively adhering to instructional sequences. Such monitoring and evaluation behaviors have been consistently linked to improved writing and language performance (Teng et al., Reference Teng, Wang and Zhang2022). The frequent cycling through these behaviors suggests that learners actively adjust their learning based on immediate feedback from the WeChat-based program, thereby operationalizing the monitoring and evaluation of self-regulatory strategies.
A notable difference was learners’ unique tendency in WIG to check their personal learning profiles immediately after quizzes, a behavior absent in WCG. This indicates that individual learners place greater emphasis on self-assessment through reviewing personalized feedback and progress summaries. Such behavior embodies self-regulatory processes related to self-evaluation and reflection, which are core aspects of metacognitive strategy use associated with enhanced self-awareness and autonomous learning (Shen & Bai, Reference Shen and Bai2024; Zimmerman & Moylan, Reference Zimmerman, Moylan, Hacker, Dunlosky and Graesser2009). Frequent profile consultation likely helped learners identify strengths and weaknesses, enabling adjustment of learning goals and strategies. In the absence of peer scaffolding, individual learners appeared to rely more heavily on system-generated feedback to sustain regulatory cycles and guide subsequent decisions, thereby compensating for the lack of peer feedback. This finding aligns with Teng and Zhang (Reference Teng and Zhang2020), reporting that metacognitive engagement promotes greater control over task execution and outcome evaluation, which is particularly critical in individual SRLL where external feedback is minimal.
In contrast, a feedback loop involving ranking checks, which leads to monitoring learning time and then returning to profile check, was uniquely presented in the WCG. This sequence highlights an advanced level of behavioral regulation characterized by time management awareness, a critical behavioral strategy identified by Wang et al. (Reference Wang, Zhou, Chen, Tong and Yang2024) as foundational to successful technology-enhanced SRLL. Collaborative learners’ engagement with managing learning time suggests that they adopt a holistic regulation approach, showing their awareness of individual achievements and peer learning paces, likely due to the shared responsibilities and social accountability inherent in group work. Such regulation exemplifies socially shared regulatory processes in which learners collectively coordinate behavioral strategies to optimize group progress and task completion (Sharma et al., Reference Sharma, Nguyen and Hong2024).
7. Conclusion
The present study compared the impact of WeChat-based collaborative and individual SRLL instructions on EFL learners’ self-regulation strategy use, writing performance, and learning behaviors, revealing that the collaborative context facilitated learners’ SRL strategy application, improved their writing ability, and enhanced their social regulation behaviors.
These findings extend the understanding of socially shared regulation of learning in technology-enhanced collaborative learning contexts. The complex interplay of SRL strategies in WCG demonstrates that regulation can be co-constructed through negotiation and discussion in collaboration. In this setting, regulation shifts from being an individual cognitive action to a shared endeavor that emerges from collective contributions, where learners plan, monitor, and evaluate the group task in unison. The behavior transitions also suggest that SRLL in TELL environments is a dynamic and cyclical process rather than a stable skill set, emphasizing the adaptive nature of regulation, where strategies are applied flexibly as per task demands and social conditions (Su et al., Reference Su, Li, Hu and Rosé2018). Importantly, WeChat functioned not merely as an interaction space but as a mechanism that externalized learners’ regulatory processes through traceable, theory-driven behavioral logging. Every learner action within the WeChat-assisted program (e.g., planning, task execution, evaluation, and time monitoring) was automatically recorded and coded using predefined behavioral categories aligned with SRL theory. This design enabled regulatory processes to be captured at a fine-grained behavioral level through temporally sequenced actions. Furthermore, LSA identified multiple statistically significant behavioral transitions, indicating that learners’ regulatory behaviors were repeatedly activated and adapted across learning processes. These findings suggest that the technological environment did more than passively record regulation; it prompted ongoing regulatory engagement, thereby functioning as a “regulatory catalyst,” which facilitates cycles of SRLL that can drive language development (Tran & Ma, Reference Tran and Ma2025).
Although WeChat-based SRLL instructions facilitated the use of SRLL strategies in both individual and collaborative contexts, the underlying strategic processes differed. In WIG, metacognitive strategies such as timely self-evaluation dominated, suggesting that learners benefited from SRLL instruction that primarily scaffolded individual self-regulation, particularly reflection and progress monitoring. In WCG, however, behavioral strategies such as time management were more evident, indicating learners’ stronger awareness of the “we” perspective characterized by co-regulation, including the coordination of actions, alignment of goals, and shared responsibility for task completion. These results highlight the need for differentiated pedagogical designs that facilitate transitions from self-regulation, collaborative regulation, and socially shared regulation, rather than treating SRLL as a uniform process. Such designs can enable learners to alternate between individual planning and monitoring, peer-supported co-regulation, and group-level regulation, leading to more balanced and flexible SRLL strategy use (Su et al., Reference Su, Li, Hu and Rosé2018). In this sense, co-regulation connects individual self-regulation and socially shared regulation, enabling learners to negotiate regulatory responsibility while maintaining individual autonomy. Moreover, these findings offer transferable insights for other digital platforms (e.g., Google Docs) that record interaction data and capture behavioral sequences. Ultimately, the effectiveness of technology lies not in its level of complexity but in how its features support learners’ engagement across each phase of SRLL.
Despite several meaningful findings, this study is not without limitations. Primarily, we focused on learners’ learning behaviors throughout the process but did not investigate weekly changes in their behavior patterns in the WeChat-based learning program. This was due to the relatively short intervention duration and absence of delayed posttests, which hindered in-depth understanding of whether SRLL skills developed in the long term and how different learning contexts impact learners’ behavior changes over time. Second, the imbalanced gender proportion may have influenced SRLL engagement and collaboration dynamics. Future research with more gender-balanced samples can consider continually tracking the longitudinal evolution of learners’ self-regulation strategies and examining gender as a potential moderator in T-CSRLL processes. Third, although participants from two intact classes were randomly assigned to control and experimental groups, the use of intact classes may have limited the strength of causal inference, and the small sample size may reduce statistical power for detecting between-group differences. Future research can consider larger samples and more rigorously controlled randomization.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0958344026100536
Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Acknowledgements
We sincerely thank all participants for their contributions to this study. We also appreciate the reviewers’ and editors’ valuable feedback, which helped us improve the quality of the paper.
Authorship contribution statement
Fan Su: Writing – original draft, Data curation, Methodology, Visualization; Ying Zhao: Conceptualization, Writing – original draft, Writing – review & editing, Data curation, Methodology, Formal analysis, Validation; Di Zou: Conceptualization, Supervision, Methodology, Validation, Writing – review & editing; Biyun Huang: Conceptualization, Supervision, Writing – review & editing.
Funding disclosure statement
This research did not receive any specific funding.
Competing interests statement
The authors declare no competing interests.
Ethical statement
All procedures were approved by Human Research Ethics Committee (HREC) at The Education University of Hong Kong. We confirm that all the research meets ethical guidelines and adheres to the legal requirements of the study country.
GenAI use disclosure statement
We acknowledge that ChatGPT Version 5.5 was used to polish the language of this paper.
About the authors
Fan Su is a lecturer at Shanghai Normal University. Her research interests include teacher education and technology-enhanced language learning. She has published over 10 research articles in international SSCI and ESCI journals, such as Computer Assisted Language Learning, SAGE Open, International Review of Applied Linguistics in Language Teaching, and Asia Pacific Education Review.
Ying Zhao received an EdD degree from the Education University of Hong Kong. Currently, she is a lecturer in the Faculty of Foreign Languages and Cultures at Kunming University of Science and Technology. Her research interests include second language acquisition, self-regulated language learning, and technology-enhanced language learning.
Di Zou is an associate professor at the Hong Kong Polytechnic University. Her research interests include AI in language education, TELL, etc. She has published more than 200 papers in international journals. Stanford University listed her as one of the World’s Top 2% Most-cited Scientists since 2021 for five years.
Biyun Huang is an assistant professor at Hong Kong Metropolitan University. Her research interests are gamification, AI in education, creative talent development, transdisciplinary STEAM, and teacher professional development.
Open access
