Collaborative writing in face-to-face settings: A substantive and methodological review

Highlights

• The study reviews 94 quantitative studies on collaborative writing in face-to-face settings.

• Results indicate a heavy focus on adult learners, intermediate proficiency, and dyads.

• CW research is a domain informed by multiple substantive/theoretical domains within L2 research.

• Findings reveal a number of concerns regarding data analyses and reporting practices in the primary literature.

• The paper provides empirically grounded suggestions for future research on collaborative writing.

Abstract

Collaborative writing (CW) has received much attention in recent decades. To help elucidate the existing scope of inquiry and guide future research efforts, this study presents a methodological and substantive review of 94 quantitative primary studies on collaborative second language (L2) writing implemented in face-to-face settings. Each study was coded for study context, demographic features, research focus, and measurements of the target features. In addition, each study was coded for methodological features including research design, analyses, and a number of reporting practices associated with transparency. The results indicated (a) a heavy focus on adult learners, intermediate proficiency, and dyads; (b) more attention to task-related variables, meaning-focused tasks, and L2 interactional features; and (c) substantial variability in the measurements of written output. Our results also revealed concerns about research and reporting practices such as inadequate reporting of pre-task training and reliability estimates. The findings imply that CW is a classroom-based domain of research informed by multiple related research areas (i.e., L2 writing, task-based language learning, second language acquisition). Drawing on our findings, we provide a number of empirically-grounded suggestions for future research in this domain.

Introduction

Defined as a task in which two or more learners interact with each other throughout the writing process to co-produce one single document, collaborative writing (CW) has been thriving in the last two decades (Storch, 2019). Following early work on CW in second language (L2) settings (e.g., DiCamilla & Antón, 1997; Kowal & Swain, 1994), CW has been introduced to and investigated in diverse language classrooms. This growing interest in CW has primarily been motivated by (a) the theoretical support afforded by the interaction hypothesis (Long, 1983), the output hypothesis (Swain, 1993), and sociocultural theory (Vygotsky, 1978); (b) the increasing popularity of communicative language teaching methods and task-based language teaching (TBLT); as well as (c) the fact that CW unites elements that are critical for L2 learning and writing development (e.g., intermediate peer feedback, negotiations of form and meaning, and L2 written output).

Previous research on collaborative L2 writing mainly falls under two strands. The first strand of studies, which has focused on examining the potential of CW for L2 learning, has used either descriptive or quasi-experimental methods to investigate whether CW offers language learning opportunities, promotes the development of L2 writing proficiency, and facilitates L2 performance or learning gains (e.g., Kim, 2008; Shehadeh, 2011; Wigglesworth & Storch, 2009). Findings in this area have generally confirmed the benefits of CW as a pedagogical task. Building upon such findings, the other strand of research has explored variables that may influence the potential of CW for L2 learning. Researchers have investigated, for example, how variation in learner traits, task design, task implementation, or learning settings may influence the different aspects of CW relating to L2 learning, including co-constructed texts, learner interaction, and task perception.

The two subsequent sections first review the different variables that have been investigated in CW and then review how prior research has analyzed evidence of learning, learners’ co-constructed texts, collaborative dialogues, and task perception in CW. A third section involves the issue of how such investigations are reported. It should be noted that while CW tasks have been widely employed in both traditional and computer-mediated communication (CMC) settings, this study exclusively focused on CW tasks implemented in face-to-face settings, as research in CMC settings has been reviewed fairly recently in Li (2018).

Variables that have been investigated in face-to-face CW research, in essence, can be divided into two categories: task-related variables and learner-related variables. The former includes variables concerning the design and implementation of CW tasks (i.e., task design and task implementation), whereas the latter involves variables relating to the traits of learners (e.g., proficiency, individual differences).

In terms of task-related variables, prior CW research has probed into task design variables such as task type and task mode. For instance, García Mayo (2002) and Alegria la Colina and García Mayo (2007) compared the L2 learning opportunities in both meaning-focused (e.g., jigsaw) and form-focused tasks (e.g., dictogloss, text reconstruction). Rouhshad, Wigglesworth, and Storch (2016) and Tan, Wigglesworth, and Storch (2010) examined the role of task modality by comparing how L2 learners performed differently in CW tasks designed for face-to-face settings and those designed for CMC settings.

With regard to the implementation of CW tasks, prior research has examined a variety of variables (e.g., group sizes, group-formation methods, language used for interaction, and pre-task modeling). For instance, Fernández Dobao, 2012, Fernández Dobao, 2014 examined how group size may influence the L2 written output, peer interaction, and vocabulary acquisition of intermediate Spanish learners in CW tasks. Mozaffari (2017) compared two CW task conditions in an Iranian EFL classroom: working in self-assigned pairs and working in teacher-assigned pairs. Unlike the studies examining the grouping of students, Zhang (2018a, examined how L1 or L2 use may influence EFL learners’ linguistic performance and peer interaction in CW.

Compared with task-related variables, the number of learner-related variables that have been examined seems to be more limited. One frequently investigated learner variable is proficiency. A number of studies have investigated how learners, when paired with peers of various proficiency levels, may assume differing roles, attend to language use differently, and form different relationships (Kim & McDonough, 2008; Storch & Aldosari, 2013; Watanabe & Swain, 2007). DiCamilla and Antón (2012) examined the relationship between proficiency level and L1 use in learners’ collaborative dialogues in CW. Another learner trait that has received growing attention is a learner’s linguistic background. For instance, a range of recent studies have examined how a pair member’s linguistic background (e.g., L2 learners vs. heritage language learners) may influence their interactional features (e.g., Fernández Dobao, 2020a; Henshaw, 2013; Torres & Cung, 2019; Walls, 2013, 2018) or perceptions of CW tasks (e.g., Fernández Dobao, 2020b). Also noteworthy is recent work by García Mayo and colleagues, which examined the potential of CW for a likely underrepresented learner population (i.e., young L2 learners) in Content and Language Integrated Language (CLIL) programs (e.g., Basterrechea & García Mayo, 2013; Calzada & García Mayo, 2020).

To understand the utility of CW in the language classroom in comparison with other tasks or in relation to the aforementioned variables, prior research has often investigated learning gains, learners’ jointly written texts, learner interaction, and task perception in CW. This section provides an overview of the measurement of these four aspects in prior research.

In detecting evidence for learning gains in CW, prior research has utilized a wide range of measures. For instance, Eckerth (2008) designed pre- and post-tests to measure gains in explicit knowledge of target grammatical features resulting from CW tasks. Henshaw (2013) considered learning gains through the incorporation of negotiated L2 forms in subsequent individual writing tasks. While all focusing on vocabulary learning in CW, Kim (2008) adopted a vocabulary knowledge scale to examine learners’ self-reported word knowledge in a pre-test and post-test, whereas Fernández Dobao (2014) and García Mayo and Zeitler (2017) compared learners’ usage of chosen lexis in a pre-test and a post-test following CW tasks. On the other hand, Shehadeh’s (2011) longitudinal study focused on writing development and measured improvement in learners’ writing quality after receiving CW as a semester-long treatment.

Similar to other L2 writing research, CW research has also employed the complexity, accuracy, fluency (CAF) framework to analyze written products (Housen et al., 2012). The measures of complexity (i.e., the use of diverse and sophisticated lexical and syntactic structures) in L2 writing, particularly syntactic complexity, have involved substantial variability (Lan, Liu, & Staples, 2019), ranging from T-unit, clause-related measures, and subordination-related measures, to specific lexico-grammatical features such as nominal structures (Biber, Gray, & Poonpon, 2011; Larsson & Kaatari, 2020; Norris & Ortega, 2009). Defined as “the extent to which an L2 learner’s performance deviates from a norm” (Housen et al., 2012, p. 4), accuracy has been mainly measured in three ways: holistic scales, number of error-free units, and the number of errors (Polio, 1997). Last, fluency, while often measured by text length, has also occasionally been considered as a temporal phenomenon and operationalized as words per minute (e.g., Chenoweth & Hayes, 2001). Overall, immense variability is characteristic of the metrics of CAF in previous L2 writing research.

To better understand the nature of collaboration and interaction and their possible associations with learner writing, learners’ collaborative dialogues have also often been analyzed. Specifically, three parameters have been employed: (a) focus area (i.e., what learners discuss); (b) language learning opportunities (i.e., how learners negotiate form and meaning); and (c) the dynamics of collaboration. Regarding focus area, a predominant framework (Storch, 2005) identified both task-related focus areas (e.g., content elicitation, discussion on organization) and interaction-related focus areas (e.g., interpersonal communication, backchannels) (McDonough, Crawford, & De Vleeschauwer, 2016; Wigglesworth & Storch, 2009; Zhang, 2019a, 2019b). Language learning opportunities are often operationalized as language-related episodes (LREs), defined as any part of interaction during which learners “talk about the language they are producing, question their language use, or correct themselves or others” (Swain & Lapkin, 1998, p. 326). LREs have been further analyzed for type (e.g., Fernández Dobao, 2014), resolution (e.g., Kim, 2008), and level of engagement (e.g., Rouhshad et al., 2016). Last, prior research has often analyzed the dynamics of collaboration using Storch’s (2002, 2013) model, which distinguishes five collaboration types based upon a global analysis of learners’ involvement in and control over a CW task (e.g., Chen & Hapgood, 2019; Watanabe & Swain, 2007).

Aside from their L2 output in CW, learners’ task perception has occasionally been examined to afford additional insights on the utility of CW tasks in language classrooms from the learner perspective. Interviews and questionnaires are two frequently employed methodologies to gather such information in prior research (Fernández Dobao & Blum, 2013; Fernández Dobao, 2020b; Storch, 2005).

Aiming to understand collaborative L2 writing as a subdomain of L2 research and inform future studies in this area, one goal of the current study is to review how previous face-to-face CW research has analyzed the various L2 features of CW, including the co-constructed texts, collaborative talks, and task perception.

In addition to synthesizing the analyses in CW research (i.e., the substantive review), this paper is also interested in the methodological practices in this domain. The field of quantitative L2 research has witnessed a surge in methodological syntheses in the past decade. In addition to surveying individual research and reporting practices in quantitative L2 research (e.g., Al-Hoorie & Vitta, 2019; Derrick, 2016; Plonsky, 2013, 2014; Plonsky & Ghanbar, 2018), previous efforts have systematically examined methodological practices in a range of subdomains of L2 research, such as L2 interaction (Plonsky & Gass, 2011), task-based learner production (Plonsky & Kim, 2016), computer-mediated interaction (Ziegler, 2016), and written corrective feedback (Liu & Brown, 2015). Their findings have suggested considerable idiosyncrasies and inconsistencies in the reporting of statistical data (e.g., standard deviation, effect sizes, confidence intervals, assumption checking) that are critical for better interpreting results in primary studies and conducting further meta-analytic research (Liu & Brown, 2015; Plonsky, 2013, 2014). Derrick (2016) also found very limited reporting of instrument piloting and revision, which poses obstacles for research replications and potential threats to internal validity. Other researchers have raised concerns about the reporting of reliability estimates for both instruments and the coding of L2 features (Larson‐Hall & Plonsky, 2015; Plonsky & Derrick, 2016; Plonsky & Kim, 2016; Saito & Plonsky, 2019).

Drawing upon the argument in Plonsky (2013) that advancement in a field of inquiry relies on “sound research designs, principled data analyses, and transparent reporting practices” (p. 656), this study joins other L2 researchers’ efforts in promoting methodological rigor by reviewing the research and reporting practices in quantitative CW research. Although still a young domain of research, CW has received much attention in the past decades. A methodological and substantive review of face-to-face CW research will, we hope, elucidate the existing scope of inquiry, shed light on areas in need of more research, and provide empirically-grounded guidance to future research efforts. With these goals in mind, this paper answers four research questions (RQs):

• What are the study contexts and research design features in quantitative CW research in face-to-face L2 settings?

• How have the different aspects of CW (i.e., co-constructed texts, learner interaction, task perception) been investigated in quantitative CW research conducted in face-to-face L2 settings?

• What variables have been investigated in quantitative CW research in face-to-face L2 settings?

• What research and reporting practices have been employed in quantitative CW research in face-to-face L2 settings?