Introduction

Writing has been conceptualised as a complex activity invoking cognitive, metacognitive, linguistic, affective, and motivational processes (Graham, 2006; Hayes, 1996; Zhang, 2022). These processes are recursive in nature, whereby writers execute metacognitive monitoring in planning, translating, and reviewing their writing process and product. Such understanding highlights the significant role of metacognition in each process of writing (Lee & Mak, 2018). Further supported by Hacker et al. (2009), the significance of metacognition is manifested in the view that writing is applied metacognition. Metacognition is usually understood as learners’ awareness and understanding of their own cognitive processes, enabling learners to monitor and regulate these processes in pursuit of a learning goal (Flavell, 1976). Achieving academic success relies not only on what teachers and programmes offer them; it also depends on learners’ development of metacognitive skills that can empower them to manage or self-regulate their own learning (Anderson, 2005). Indeed, metacognition is essential in L2 learning and higher-order human cognitive processes, including writing in a foreign language (Flavell, 1979; Wenden, 1998; Xu, 2023). Although several scholars have established few correlations between metacognition and writing performance (Qin & Zhang, 2019; Teng et al., 2022; Teng & Zhang, 2016; Yang, 2014), studies investigating the impact of metacognitive instruction on L2 writing remain insufficient. The mechanisms underpinning metacognitive instruction for L2 writing development are not fully understood, especially in the context of collaborative writing.

Byrnes and Manchón (2014) regarded writing as a multifaceted cognitive process involving various factors, such as the learner, instructor, task, and available resources, all contributing to the complex meaning-creating process that holds potential for L2 learning and development (Zhang, 2013, 2021). Two factors that appear to have a significant impact on the potential of writing for language learning and development are metacognitive instruction (Teng & Zhang, 2020; Xu, 2023) and collaborative writing (Li & Zhang, 2023; Li et al., 2020). Recent studies have revealed that metacognitive instruction helped L2 writers improve their writing performance, effective strategy use, learner autonomy, and transferring the learnt strategies effectively to other situations (Alfaifi, 2022; Forbes & Fisher, 2020; Nguyen & Gu, 2013; Teng & Zhang, 2020). Collaborative writing, where multiple individuals work together to create a text with shared responsibility for the co-constructed products, has the potential to support L2 learning and writing development (Storch, 2011, 2019). It can also encourage learner interactions and facilitate gap noticing, peer feedback, pushed output, and languaging. Whilst writing collaboratively, learning is not confined to an individual’s efforts but becomes a social endeavour that intertwines with the learning environment, tools, and overall context (Bueno-Alastuey et al., 2022).

For decades, scholars have been seeking the development of complexity, accuracy, and fluency measures of language performance in response to specific instructions and stimuli to understand the relationship between teaching and learning in a more organic manner (Lu, 2010; Norris & Ortega, 2009; Xu et al., 2023; Xu & Zhang, 2023). Amongst the limited number of studies investigating metacognitive instruction on L2 writing, few, to our knowledge, have explored the effect of metacognitive instruction on the effectiveness of collaborative writing in facilitating writing development. Therefore, this research intends to assess the feasibility of incorporating explicit metacognitive instruction with collaborative writing interventions for L2 writing development in lexical complexity, accuracy, and fluency.

Literature Review

Collaborative Writing and Writing Performance

In the L2 writing literature, collaborative writing is characterised as an interactive and educational activity where two or more learners engage in the writing process to jointly create a single text (Li & Zhang, 2023; Storch, 2019). Studies on collaborative writing have tended to be driven by social-cultural and cognitive perspectives, based on socio-cultural theory (Vygotsky, 1978), the interaction hypothesis (Long, 1996), and the output hypothesis (Swain, 1993). From a socio-cultural theoretical perspective, collaborative writing is viewed as a pedagogical activity facilitative of language learning, taking advantage of appropriate forms of support, deliberating about languaging, and collective scaffolding during meaningful social interactions (Donato, 1994; Lantolf & Thorne, 2006; Swain et al., 2009). From a cognitive perspective, the interaction within the collaborative writing process allows learners to participate in various activities, including negotiating meaning, providing and receiving feedback, and engaging in “pushed output” where they are impelled to notice the gap between their current language proficiency and the target language norms (Gass & Mackey, 2007; Long, 1996; Pham, 2023). Research evidence suggests that collaborative writing can potentially enhance L2 learning and writing development (Kim, 2008; McDonough & Fuentes, 2015; Wigglesworth & Storch, 2009; Zhang & Plonsky, 2020). Numerous studies also indicate that the pairs or group writers reported having positive experiences and attitudes towards collaborative writing activities (Storch, 2005; Fernández-Dobao, 2020).

When investigating the effects of collaborative writing on L2 writing development, scholars have either compared the texts written collaboratively with those produced individually, or adopted a pre-test, and a post-test design with writing tasks to examine the impact of collaborative writing on subsequent individual writing (Khatib & Meihami, 2015; Storch, 2005; Villarreal & Gil-Sarratea, 2020). Written texts were usually analysed using either established writing rubrics or assessments based on linguistic complexity, accuracy, and fluency. Differentiated results have been reported when comparing the complexity, accuracy, and fluency writing performance of collaboratively written text with that produced individually (Zhang & Plonsky, 2020). This suggests that one-short designs might fail to establish conclusive support for the theoretically presumed advantage of collaborative writing, which drives researchers to investigate further whether the impact of collaborative writing can endure over time and be observed consistently across various writing experiences, especially on subsequent individual writing (Elabdali, 2021).

Until recently, only a few scholars have investigated the effect of collaborative writing on subsequent individual writing skills (Bueno-Alastuey et al., 2022; Chen, 2019; Khatib & Meihami, 2015; Shehadeh, 2011). Amongst the scarce research, Bikowski & Vithanage (2016), and Khatib & Meihami (2015) found that the benefits of collaborative writing can be transferred into later individual writing performance measured through writing rubrics, such as content, organisation, grammar, vocabulary, and mechanics. However, much research has reported that collaborative writing only improves the accuracy, and fluency scores of the participants’ individual writing, not the complexity scores (Bueno-Alastuey et al., 2022; Chen, 2019). The reason was that participants focused on grammatical accuracy when engaging in the language learning opportunities (e.g. “languaging”, peer feedback, language-related episodes on vocabulary, grammar issues) whilst writing collaboratively.

The inconsistent findings regarding the effects of collaborative writing as a pedagogical activity have encouraged scholars to investigate the factors that might mitigate the learning outcomes by influencing the collaborative writing process (Chen & Hapgood, 2021; Zhang, 2018). Under such an initiation, the metacognitive approach offers a beneficial way to understand the link between collaborative writing and L2 learning (e.g. why some learners improve their L2 whilst others fail to) (Sato, 2023). This is because learners’ knowledge of themselves as learners, the task they are engaged in, and the strategies available to them have a direct effect on their engagement with the language learning opportunities during collaborative writing (Wenden, 2001; Teng & Zhang, 2024). This claim has been supported by Chen & Hapgood (2021), revealing that the knowledge of collaborative writing was positively correlated with the patterns of interaction and languaging opportunities manifested in the writing processes. Chen & Hapgood’s (2021) findings showed that learners with more knowledge about collaborative writing revealed a collaborative pattern of interaction (high equality, high mutuality), a much higher number of language-related episodes, and self-resolving such language-related episodes more frequently. Although their research suggests a significant correlation between metacognitive knowledge of collaborative writing and the interaction process during writing collaboratively (participation and learning) (Chen & Hapggood, 2021), they did not examine the direct relationship between the training of metacognitive knowledge of collaborative writing and the actual learning outcomes. Following this line of research, several scholars have tried to investigate how metacognitive instruction in collaborative writing might contribute to L2 writing development (Teng, 2021; Teng & Huang, 2023). Before reviewing these studies, we explain what metacognitive instruction entails.

Metacognitive Instruction

Metacognition is closely related to the process of self-regulated learning, “an active, constructive process whereby learners set goals for their learning and then attempt to monitor, regulate, and control their cognition, motivation, and behaviour, guided and constrained by their goals and the contextual features in the environment” (Pintrich, 2000, p. 453). Many struggling learners encounter challenges in learning due to their lack of pertinent knowledge and skills which has hindered them from selecting relevant strategy packages for effective self-regulation (Sato, 2023; Schraw & Gutierrez, 2015). This drives scholars and foreign language practitioners to take the initiative in training learners with knowledge and regulation of metacognition, known as metacognitive instruction (Lee & Mak, 2018; Xu, 2023).

As early as 1986, Palincsar (1986) illustrated that metacognitive instruction, aiming to affect learners’ interactions with learning, should go beyond letting students learn a repertoire of strategies. More significant is the learners’ ability to select the best strategies that fit the learning targets, overcome challenges they may face, and align with their capabilities and limitations. The learners also need to possess the capacity to monitor and regulate their strategy use and evaluate the effectiveness of the strategies selected in achieving learning goals so that adjustments can be made in strategy selection for future learning tasks or situations (Palincsar, 1986). Effective metacognitive instruction improves self-regulation, allowing learners to “set goals for their learning and then attempt to monitor, regulate and control their cognition” (Pintrich, 2000, p. 453). It is hypothesised that metacognitive instruction has the potential to enhance learners’ higher-order thinking abilities, supporting cognitive processes in learning behaviours and achievement (Sato, 2023).

Metacognitive Instruction and L2 Writing

Previous studies have indicated that metacognitive instructions facilitate learners’ effective use of learning strategies, enhance learning outcomes, and cultivate their potential to become more accomplished language learners (de Boer et al., 2018; Ku & Ho, 2010; Xu, 2023; Zhang, 2010). As scholars have advocated, metacognitive instruction necessitates the interplay of various factors, encompassing the teacher, learner, and tasks (Sato, 2023; Sato & Loewen, 2018). This argument reiterates the suggestion that exploring the potential synergies between metacognitive instruction and collaborative writing is a valuable avenue for investigation (Zhang & Zhang, 2019).

Although recent scholars have acknowledged the benefits of combining metacognitive instruction with collaborative learning (Xu, 2023; Zhang & Zhang, 2018), several meta-analyses reported divergent results when strategy instruction was combined with group work (De Boer et al., 2014; Dignath & Büttner, 2008). Dignath and Büttner (2008) reported that only secondary school students benefited from strategy instruction in collaborative learning situations, whilst primary school students showed no gains from such combining practice. However, in another meta-analysis, De Boer et al. (2014) found that the cooperation group failed to show any superior benefits from strategy instruction compared to the group without cooperation.

In the context of L2 writing, much metacognitive instruction research has been conducted with learners in individual writing conditions (Alfaifi, 2022; Forbes & Fisher, 2020). Nevertheless, learners who possess metacognitive knowledge in individual writing may encounter challenges when transitioning to collaborative writing, as they may lack the know-how to effectively engage in tasks such as incorporating and offering feedback, negotiating meaning and forms, and reaching agreements with peers to complete writing task during the collaborative writing process (Sato, 2023). Both cognitive and metacognitive factors are essential in the effectiveness of collaborative writing on L2 learning (Sato, 2022; Zhang & Zhang, 2018). Metacognitively speaking, learners might actively engage in activities facilitative of L2 development when they understand the benefits of collaborative writing. Whilst engaging in cognitive activities during collaborative writing (e.g. processing input, offering and uptaking feedback, pushed output) has a direct effect on L2 learning, metacognitive processes such as planning, monitoring, and evaluation support those cognitive activities whilst writing collaboratively through metacognitive regulation (Sato, 2023). To maximise the effectiveness of collaborative writing in supporting L2 learning, Sato (2023) suggested that metacognitive instruction, designed in accordance with the three stages of planning, drafting, and revising, can be offered to learners so that higher-order thinking can be triggered in supporting their engagement in the cognitive process of collaborative writing.

Previous attempts have been made to investigate the impact of metacognitive instruction on the effectiveness of collaborative writing for L2 writing development. These attempts focus on the knowledge of metacognition, and regulation of metacognition in L2 writing (Teng, 2021; Teng & Huang, 2023). For example, Teng (2021) examined the effects of four different instructional combinations, metacognitive guidance with collaborative writing, metacognitive guidance without collaborative writing, collaborative writing without metacognitive guidance, and individual writing, on Chinese EFL students’ academic writing skills development. Metacognitive guidance encompasses metacognitive knowledge (knowledge of task, strategies, and self) and metacognitive regulation (planning, monitoring, and evaluation) for effective L2 writing. The study revealed that collaborative writing with metacognitive guidance triggered the participants to engage in higher levels of information processing, cultivate a more profound understanding of metacognition, and monitor and reflect on their writing processes, thus significantly improving their academic writing skills over time.

Teng and Huang (2023) further studied the effects of metacognitive instruction and collaborative writing on Chinese EFL learners’ writing performance under four instruction types: metacognitive instruction with collaborative writing, metacognitive instruction with individual writing, collaborative writing, and individual writing. Participants’ writing performance was measured by syntactic complexity, accuracy, and fluency. Metacognitive instruction was conducted with the method of metacognitive prompts, focusing on metacognitive knowledge, and metacognitive regulations. Results showed that participants benefited from metacognitive instruction and collaborative writing conditions in writing more accurately, with a null effect for complexity and fluency measures (Teng & Huang, 2023).

However, it might be challenging to draw a firm conclusion based on such a limited number of studies into the impacts of metacognitive instruction in collaborative writing on L2 writing development. Furthermore, the metacognitive training presented in those earlier studies, whether through guidance or prompts, focused primarily on strategies tailored for individual writing situations. This has led to concerns that these approaches may not be suitable or effective in collaborative writing contexts. This holds particularly true for Chinese EFL writers, who are well-acquainted with individual writing but possess limited experience or knowledge in collaborative writing. However, an exception was Chen and Hapgood (2019), which encompassed metacognitive instruction with both metacognitive knowledge and metacognitive regulation of collaborative writing and was tailored for Chinese EFL learners, thus we adopted their training package in this research.

Considering the limited amount of research investigating the combined effect of collaborative writing and metacognitive instruction, and the scarcity of research focusing on post-graduate students, we aim to investigate the impacts of metacognitive instruction on the effectiveness of collaborative writing on a cohort of Chinese EFL post-graduate students’ writing performance development assessed by lexical complexity, accuracy, and fluency. Accordingly, we address the following research question: Does metacognitive instruction for collaborative writing in conjunction with collaborative writing interventions facilitate Chinese post-graduate students’ writing development in lexical complexity, accuracy, and fluency?

Method

Participants

This study, conducted in China, highlights the challenges faced by Chinese university teachers in teaching EFL writing, emphasising difficulties arising from test-driven and product-oriented pedagogical practices (Li et al., 2020; Zhang et al., 2016). In this study, we recruited two classes of post-graduate students majoring in Arts from a university in central China through convenience sampling (Dörnyei, 2010). All participants signed written consent forms before participating in this study. They were enrolled in the same compulsory English course with the goal of enhancing their overall English proficiency during this study. Although they were taught by the same instructor, a professor in Applied Linguistics, their classes were held in separate classrooms at different time. According to the pre-test results, they were homogeneous in their written proficiency at the onset of this study. The participants consisted of 21 male and 41 female students, ranging in age from 21 to 28, with an average age of 23. These two classes were randomly assigned into either an experimental group (N = 32) or a control group (N = 30). To randomly allocate the two classes to different treatments, we assigned each class a random number, either 1 or 2, and then distributed them to the respective conditions (1 for the experimental group, 2 for the control group) without following any predetermined pattern or bias. Participants from the experimental group received explicit metacognitive instruction and collaborative writing interventions, whilst the control group students were instructed according to the regular teaching schedule, following a genre-based approach to writing.

Instruments

Writing Materials for Collaborative Writing

We provided participants in the experimental group with three collaborative argumentative writing tasks to complete as part of in-class collaborative writing interventions. The rationale for adopting the argumentative genre for the interventions was that argumentative writing is a significant genre taught in the teaching syllabus and the textbook for the participants, as well as a prevalent writing proficiency test in appearance in language tests for Chinese university students, both nationally and internationally (Huang & Zhang, 2020; Li & Zhang, 2022). Also, we believed that improving their ability to write appropriate argumentative writing might enhance their motivation to participate in this study. Details of the topics of the writing tasks can be found in Appendix I.

Pre-, Post-, and Delayed- Post-tests of Writing

In order to assess the progression in the writing performance of the participants following the interventions, three argumentative writing tests were incorporated, and adapted from the IELTS writing examinations. The topics of the three tests were purposefully selected in accordance with the teaching schedules. They were piloted on a group of non-participants with similar demographical information to the participants in this study to ensure that the three tests were similar in their difficulty level. The tests were administered in a traditional paper-and-pen format, adhering to the 40-min duration allocated to test-takers in official IELTS tests. Appendix II provides the topics for the three tests.

Metacognitive Instruction for Collaborative Writing

We adopted the training package from Chen and Hapgood (2021) to familiarise participants with metacognitive knowledge of collaborative writing, with the aim to cultivate their metacognitive knowledge and metacognitive regulations, supporting the pairs to write in a collaborative relationship as reflected as a willingness to share ideas, respecting partner’s contributions, and taking equal turns (Storch, 2002, 2019). The training package includes materials illustrating the definitions of collaborative writing, its benefits and shortcomings (declarative knowledge), knowledge about how to implement collaborative writing and necessary steps and skills for collaborative writing (procedural knowledge), and examples of successful collaborative writing (conditional knowledge). The training package intends to facilitate learners to understand the process of collaborative writing, the learning benefits of collaborative writing, and the skills to maximise the learning benefits of collaborative writing. For detailed information on the training materials, please see Chen and Hapgood (2021). The instructional sessions and collaborative writing interventions occurred within the regular classroom schedule, whilst the control group received normal instructions for an equivalent duration. This was conducted to ensure that both groups had a similar amount of learning time. Based on the first author’s in-class observations, the control group received classes that incorporated model text analysis focusing on the structures and linguistic characteristics of essays. Additionally, the teacher provided various modal phrases and sentences tailored to each section of the essays. Participants in the control group did not engage in any collaborative writing or cooperative work during this study. They were only tasked with the pre-, post-, and delayed post-test writing as in-class practice.

Procedure

Before data collection, participants were provided with detailed information about the study’s purpose and procedures, and signed consent forms were obtained. Subsequently, they were randomly allocated to either the experimental or control group. To ensure fairness in the teacher’s attitude towards the experimental and control groups, the first author received assurance from the classroom teacher that she would hold a position of neutrality towards both groups. At the onset of the study, both groups underwent a pre-test writing session. The first author of this research also met with the instructor to train her to familiarise her with the process of metacognitive instruction. Participants in the experimental group were then instructed to self-select pairs for collaborative writing and were given the option to change partners if necessary. Prior to the in-class training and practice session, participants were given a training package to review at home. Similarly, participants in the control group were tasked with completing standard homework assignments and reading articles from the textbook slated for study in the following week. The 2-hour training session commenced with a detailed discussion, led by the instructor, of the collaborative writing concepts outlined in the provided reading materials. Following this discussion, participants practised collaborative writing under the guidance of the instructor, who circulated the classroom, offering support and encouragement to the pairs to work in a collaborative manner. After the training session, the experimental group engaged in three collaborative writing interventions over three weeks. These interventions involved collaborative composition of three argumentative essays on topics selected by participants amongst several candidate topics relevant to the teaching materials in the textbook. After assigning the collaborative writing tasks, the instructor refrained from intervening in the process, stepping in only when students requested assistance. In contrast, the control group received normal instructions with a genre-based approach during both the training and intervention periods, and did not participate in any collaborative assignments. Immediately after the interventions, both groups completed a post-test writing session. Additionally, a delayed post-test was administered to both groups six weeks after the interventions concluded.

Data Analysis

The pre-test, post-test, and delayed post-test writing performance were analysed from the dimensions of lexical complexity, accuracy, and fluency. These metrics have been utilised to explore written and oral language performance, language proficiency, and language development in L2 learners (Bulté & Housen, 2014; Housen & Kuiken, 2009; Xu et al., 2023).

Lexical complexity refers to the sophistication and diversity of vocabulary employed by L2 learners in communications (Lu, 2012). It encompasses lexical density, lexical sophistication, and lexical variation. Lexical density was assessed by number of lexical words to the total number of words (Nlex/N). Lexical sophistication was assessed by the number of sophisticated lexical words to the total number of lexical words (Nslex/Nlex). Lexical variation was assessed through the corrected type token ratio (CTTR). We adopted the automatic computer software, L2 Lexical Complexity Analyser, for lexical complexity analysis (Lu, 2012).

Accuracy refers to the extent to which the language produced aligns with the native-like norms established in the target language (Pallotti, 2009). Fluency refers to the extent to which a learner demonstrates his/her ability to produce language in a fast and smooth manner without interruption (Pallotti, 2021). Accuracy was operationalised as the ratio of error-free clauses (EFC/C), whilst fluency operationalised as the number of words written since the participants were required to write for the same amount of time. We coded the errors based on Polio and Shea’s (2014) guidelines. To ensure the reliability of the coding, another coder, a PhD student from the same faculty majoring in Applied Linguistics, was invited. At the beginning, the coder and the first author collaboratively coded five texts to identify and address any discrepancies in their coding approaches, before they independently coded about 20% of the writing. The intraclass correlation coefficient (ICC) was used to calculate the interrater reliability of the coding. The results revealed an acceptable inter-coder reliability of .82. Then, the first author coded the rest of the written texts.

Before delving into the primary statistical analysis regarding the within and between-group effects of the two groups, assessments were conducted to verify assumptions related to normality, the presence of outliers, and missing data. Subsequently, mixed within-between subject ANOVAs were performed, followed by within-subject and between-subject simple effect analyses.

Results

Lexical Complexity

Table 1 illustrates the descriptive statistics for the lexical complexity scores in each group across the three tests. These statistics reveal that both groups’ participants seemed to exhibit slight enhancements in lexical density, lexical sophistication, and lexical variation performance. Further details of the inferential statistical analysis results can be found in the following sections.

Table 1 Descriptive statistics of lexical complexity
Full size table

Lexical Density

Mixed ANOVA results on the lexical density performance showed a significant main effect of time, F (2, 59) = 16.815, p < .001, Partial η2 = .363, but no significant main effect of group, F (1, 60) = .141, p = .709, Partial η2 = .002, and no significant interaction effect between the time × group, F (2, 59) = .191, p = .827, Partial η2 = .006. Further pairwise analysis revealed that participants from both groups improved their lexical density performance across time, and the results were maintained in the delayed post-test (experimental group: pre- vs. post-test, p < .05; pre- vs. delayed post-test, p < .05; post- vs. delayed post-test, p > .05. control group: pre- vs. post-test, p < .001; pre- vs. delayed post-test, p < .05; post- vs. delayed post-test, p > .05). Between-subject comparison revealed that the two groups showed similar performance in the pre-test (p > .05), post-test (p > .05), and delayed post-test (p > .05). We can see that both the interventions and normal instructions help improve participants’ lexical density performance.

Lexical Sophistication

Mixed-design ANOVA results for the lexical sophistication performance showed that there was a significant main effect of time, F (2, 59) = 103.966, p < .001, Partial η2 = .779, but no significant main effect of group, F (1, 60) = 4.012, p = .050, Partial η2 = .063, and no significant interaction effect between the time × group, F (2, 59) = .001, p = .999, Partial η2 = .001. Further pairwise analysis revealed that both groups benefit from consistently improving their lexical sophistication performance (experimental group: pre- vs. post-test, p < .001; pre- vs. delayed post-test, p < .001; post- vs. delayed post-test, p < .001. control group: pre- vs. post-test, p < .001; pre- vs. delayed post-test, p < .001; post- vs. delayed post-test, p < .001). No significant difference was detected in the lexical sophistication performance in pre-test (p > .05), post-test (p > .05), and delayed post-test (p > .05) for the two groups.

Lexical Variation

A significant main effect of time, F (2, 59) = 19.042, p < .001, Partial η2 = .392, a significant main effect of group, F (1, 60) = 7.166, p = .010, Partial η2 = .107, and a significant interaction effect between the time × group, F (2, 59) = 5.679, p = .006, Partial η2 = .161 were revealed in the mixed-design ANOVA results for the lexical variation performance. Further pairwise analysis revealed that only the experimental group participants improved their lexical variation performance across time (pre- vs. post-test, p < .001; pre- vs. delayed post-test, p < .001), and the results were retained in the delayed post-test (post- vs. delayed post-test, p > .05), whilst lexical variance performance for the control group participants remained stable across three tests (pre- vs. post-test, p > .05; pre- vs. delayed post-test, p > .05). Between-subject comparison revealed that the two groups showed similar lexical variation performance at the onset of the study (p > .05), however, after the intervention, the participants from the experimental group exhibited significantly better results than their counterparts from the control group in the post-test, p = .015 with small effect size, Cohen’s d = .454, but not in the delayed post-test, p > .05.

Accuracy

Table 2 displays the descriptive statistics for the writing test accuracy scores within each group across the three tests. These statistics underscore that the participants from both groups seemed to have slight improvements in the accuracy scores. Additional details concerning the results of the inferential statistical analysis are elaborated upon in the following sections.

Table 2 Descriptive statistics of accuracy
Full size table

Mixed-design ANOVA results for the accuracy performance showed a significant main effect of time, F (2, 59) = 11.451, p < .001, Partial η2 = .280, and a significant main effect of group, F (1, 60) = 4.658, p = .035, Partial η2 = .072, but a non-significant interaction effect between the time × group, F (2, 59) = 2.395, p = .100, Partial η2 = .075. Further pairwise analysis revealed that participants from the experimental group consistently improved their accuracy performance in the post-test, and the delayed post-test (pre- vs. post-test, p < .05; pre- vs. delayed post-test, p < .001; post- vs. delayed post-test, p < .05), whilst control group participants’ accuracy performance remained stable across three test times (pre- vs. post-test, p > .05; pre- vs. delayed post-test, p > .05). Between-group analysis revealed that at the onset of the study, participants were similar in their accuracy performance (p > .05), and the same results were found in the post-test (p > .05) after the interventions. It is interesting to figure out that experimental group participants had better accuracy performance in the delayed post-test compared to their counterparts from the control, p < .001, with a small effect size, Cohen’s d = .134.

Fluency

Table 3 presents the descriptive statistics for the writing fluency scores within each group across the three tests. These statistics highlight that the participants from the experimental group demonstrated marginal improvements in fluency scores, whereas the participants from the control group maintained relatively consistent scores. Further elaboration on the results from the inferential statistical analysis is provided in the subsequent sections.

Table 3 Descriptive statistics of fluency
Full size table

Mixed design ANOVA revealed that there was a significant main effect of time, F (2, 59) = 10.264, p < .001, Partial η2 = .258, and a significant interaction effect between the time × group, F (2, 59) = 3.951, p = .025, Partial η2 = .118, but no significant main effect of group, F (1, 60) = 2.627, p = .110, Partial η2 = .042 for the participants’ fluency performance. Further pairwise analysis revealed that the interventions helped improve experimental group participants’ fluency performance across time (pre- vs. post-test, p < .05; pre- vs. delayed post-test, p < .001), and the results were maintained (post- vs. delayed post-test, p > .05). However, control group participants failed to show any significant variations in their fluency performance across three test time (pre- vs. post-test, p > .05; pre- vs. delayed post-test, p > .05). Between-subject comparison results showed that prior to the interventions, students in the two groups were quite similar in their fluency performance, p > .05. However, after the interventions, experimental group participants showed better fluency performance in the post-test, p = .001, with a large effect size, Cohen’s d = 35.545, but not in the delayed post-test, compared to their counterparts in the control group.

Discussion

This study set out to examine the impact of metacognitive instruction for collaborative writing on enhancing the effectiveness of three weeks of collaborative writing interventions. Results showed that Chinese EFL learners significantly improved their lexical variation, accuracy, and fluency performance following the collaborative writing interventions with metacognitive instruction. This section discusses these results whilst offering possible explanations by referring to prior literature.

Lexical Complexity

The results showed that the combination of metacognitive instruction and collaborative writing resulted in a significant benefit in improving participants’ lexical variation performance after the intervention, and the results were retained. Participants from the experimental group exhibited significantly better results than their control group counterparts in the post-test, although with a limited effect size. The significant improvement in the lexical variation performance might have resulted from the participants’ successful incorporation of the learning opportunities, resolved language-related episodes, and processing and offering feedback (Sato, 2023). Such learning opportunities might have been promoted for our participants from the experimental group, inspired by metacognition about collaborative writing. As Chen and Hapgood (2021) highlighted, learners who believe in the benefits of writing collaboratively to enhance language learning, might reveal more engagement and participation with collaborative writing. EFL learners actively participating in collaborative writing may experience advantages in constructing diverse perspectives, and talking about incorporating a more varied and precise range of language choices. Such talk may have enhanced their linguistic skills and contributed to a more nuanced and sophisticated expression of ideas (Li & Zhang, 2022). They might also have gained insights from peer conversations regarding word selection, encouraging the exploration of alternative terms. These interactions contribute to their capacity to acquire an extensive vocabulary for expressing meanings and internalise lexical knowledge, ultimately supporting their independent writing endeavours.

Our results showed that both interventions and normal instructions had a positive impact on participants’ lexical density and lexical sophistication performance, and these improvements were sustained in the delayed post-test. Surprisingly, collaborative writing interventions, coupled with metacognitive instructions, did not demonstrate any discernible advantage over normal instructions in enhancing participants’ lexical density and sophistication performance. Such results can be explained by the topics selected for the writing tests, which were closely related to their textbook’s writing topic. The English course they were enrolled in intended to improve their comprehensive English ability through listening, reading, speaking, and writing about several topics, revealed in several units. Consequently, exposure to such topics in English might have enhanced all participants’ lexical repertoire, which may have resulted in their improvement in lexical density, and lexical sophistication performance.

Accuracy

Participants benefited from metacognitive instructions and collaborative writing in improving their accuracy performance in the post-test and the delayed post-test, whilst normal instruction participants’ accuracy performance remained stable. The sustainable improvement in our participants’ writing accuracy performance is without surprise to our knowledge that a large sum of literature has shown that either metacognitive instruction or collaborative writing helps learners write more accurate sentences for communication (Chen, 2019; Pham, 2023). We align with earlier collaborative writing research, where the collective advantages of written performance by pairs compare with those of individually authored pieces (Zhang & Plonsky, 2020). We also support findings indicating the benefits of collaborative writing on subsequent individual writing accuracy performance (Bueno-Alastuey et al., 2022; Chen, 2019). We are also in line with the idea that writing collaboratively enables participants to co-construct knowledge and deliberate on their language use collaboratively, facilitating noticing of form, thus writing more grammatically accurate sentences (Li & Zhang, 2021a, 2021b; Wigglesworth & Storch, 2012). Furthermore, it is noteworthy that participants in the experimental group exhibited superior accuracy performance in the delayed post-test compared to their counterparts in the control group, albeit with a small effect size. This suggests that the benefits derived from the experimental group’s interventions can endure for a more extended period.

Fluency

Our results showed that collaborative writing and metacognitive instructions had sustainable benefits in enhancing participants’ fluency performance compared with normal instructions. Participants in the experimental group demonstrated enhanced fluency performance in the post-test, characterised by a large effect size, as opposed to their counterparts in the control group. However, this advantage did not persist in the delayed post-test.

Our results contradicted several previous studies showing that collaborative writing products tend to be less fluent than individual writing counterparts (Storch, 2005; Zabihi et al., 2013). However, these results were refuted by Pham (2021), who revealed that collaborative writing could help learners write more words within a given amount of time. The sustainable development of participants’ writing fluency scores in our study also echoes Schuster et al. (2020), revealing that metacognitive instruction for collaborative writing can facilitate the effective transfer of metacognitive strategies in near and far contexts. It reveals that learner self-regulation can be instructed and transferred to future tasks for enhanced task performance (Schuster et al., 2020). Our results seem to support that training learners to write collaboratively before collaborative writing interventions could improve their engagement and peer interactions with more chances of language use, idea sharing, and equal contributions, thus enhancing their ability to think of more illustrations with strong arguments (Zabihi & Bayan, 2020).

Conclusion

The present study was designed to investigate the impact of metacognitive instruction for collaborative writing in conjunction with collaborative writing interventions on Chinese EFL learners’ individual writing development. The results indicate that offering participants collaborative writing interventions alongside metacognitive instruction on writing collaboratively significantly improved their later individual written performance.

This research offers several pedagogical and theoretical implications. Pedagogically, the results of this study will be relevant to EFL teachers in that integrating metacognitive training into collaborative writing can be a valuable and essential component for teaching practitioners who aim to incorporate collaborative writing as a pedagogical approach in writing classrooms. Specifically, the current data highlights the significance of metacognitive instruction for collaborative writing as beneficial in enhancing the effectiveness of collaborative writing in improving later individual writing performance. Participants engaged in such pedagogical practice have significantly improved their writing performance in terms of complexity, accuracy, and fluency. Compared with previous studies, where only several dimensions of such metrics have been improved (Chen, 2019; Teng & Huang, 2023), our results are promising as they support the idea that learners with metacognitive awareness can effectively attend to multiple dimensions of their writing performance during collaborative writing. This emphasises the need for teachers to combine metacognitive instruction with collaborative writing, as suggested by Xu (2023), and Zhang and Zhang (2018), in harnessing the full potential of collaborative writing for writing development. Theoretically, this study contributes to our understanding of collaborative writing activities as effective pedagogical practices from a metacognitive perspective. Exploring metacognitive instruction and collaborative writing adds to our understanding of cognitive processes, metacognitive regulations, and the complex dynamics inherent in writing collaboratively (Sato, 2023). The results carry substantial weight in bolstering the metacognitive perspective of collaborative writing, thereby encouraging further investigations of the metacognitive processes integral to collaborative writing, and the metacognitive mechanisms underlying L2 learning achievement. This aligns with the observation of Zhang and Zhang (2019), who noted the mission of cultivating “lifelong learners, who are eager to show responsibilities” (p. 894) in L2 education.

Our study is also subject to several limitations, such as short intervention duration, lack of deep analysis of the collaborative writing process, and small sample size. Additionally, this study did not explore how metacognitive instruction and collaborative writing might have influenced participants’ metacognitive knowledge about collaborative writing, their understanding of collaborative writing, and their writing behaviours. Another limitation of this study lies in the absence of a collaborative writing only group, or an individual writing only group. However, considering that the learning gains of the experimental group surpass those of the control group, we believe that the instructional programme encompassing collaborative writing interventions with metacognitive instruction can be beneficial for L2 writing development. We hope that researchers and foreign language teachers continue to explore the potential of metacognitive instruction and collaborative writing, taking advantage of the development of new technologies, like the synchronous and asynchronous writing tools to monitor participants’ metacognitive processes involved during collaborative writing, and how such metacognitive mechanisms contribute to the final L2 attainment (Almalki & Storch, 2023).