Abstract
Although second language (L2) collaborative writing research has demonstrated that texts composed collaboratively are more accurate than individually-written texts, few studies have explored whether collaborative prewriting yields similar benefits. This study investigated whether collaborative prewriting, i.e. interacting with peers during the prewriting phase followed by individual writing, led to higher accuracy, complexity, or analytic ratings than individual prewriting. It also explored the relationship between these text features and student talk during collaborative prewriting. English L2 university students in Thailand (n = 57) were randomly assigned to write a problem and solution paragraph with either collaborative or individual prewriting. Their texts were analysed in terms of accuracy (errors/word) and complexity (coordination and subordination), and were rated using analytic rubrics (content, organization, language). Transcripts of the collaborative prewriting discussions were analysed in terms of the topic of student talk (content, organization, language, task management, off-task talk). The results showed that the collaborative prewriting texts were more accurate and received higher ratings than the individual prewriting texts. Furthermore, there was a significant correlation between prewriting time and accuracy. Implications for the use of collaborative prewriting tasks in settings for English as a foreign language (EFL) are discussed.
Keywords
I Introduction
Peer interaction has an important place in second language (L2) classrooms due to the benefits attributed to it from both theoretical and pedagogical perspectives. Numerous theoretical L2 acquisition perspectives have offered a rationale for providing students with opportunities to collaborate with their peers, ranging from a cognitively-oriented skill acquisition perspective that acknowledges the benefits of meaningful practice for the development of proceduralization and automaticity (DeKeyser, 2007) to more socially-situated perspectives that emphasize the co-construction of knowledge (Lantolf, 2012). Pedagogically, peer interaction has been integrated in L2 classrooms for a variety of purposes, ranging from serving as primary vehicle for language development, as in task-based language teaching, to playing an important role in the practice phase of more traditionally-oriented approaches, such as present–practice–produce. A considerable body of research has provided empirical evidence for claims about the benefits of peer interaction for L2 learning (Philp, Adams & Iwashita, 2013; Sato & Ballinger, 2016), with many studies focusing on L2 learners’ interaction during oral tasks rather than writing tasks (Mackey, 2007).
Although exclusively oral tasks have been more widely researched, tasks that include a writing component, such as dictogloss tasks, have also been investigated since the 1990s (e.g. Swain & Lapkin, 1998). In addition, the nature of students’ oral interaction while planning, composing, and revising written texts has been the primary focus of collaborative writing research since the early 2000s (e.g. Storch, 2002). Collaborative writing, as defined recently by Storch (2013, p. 3), refers to an activity where students have shared responsibility for creating a single text and compose that text through a negotiated decision-making process. As Storch has pointed out, unlike cooperative writing where there is a division of labor among individual students, collaborative writing involves all students working together simultaneously. Empirical studies that tested the benefits of collaborative writing by comparing texts written by individual students to those composed by students in pairs or small groups have shown that collaborative texts are more accurate. However, the majority of collaborative writing studies have found no differences in measures related to grammatical complexity, such as t-unit measures (e.g. words or clauses per t-unit), subordination (dependent clauses/clauses), or nominal and relative clauses (Fernandez Dobao, 2012; McDonough & García Fuentes, 2015; Storch, 2005; Storch & Wigglesworth, 2007; Wigglesworth & Storch, 2009). Studies that included analytic ratings of student texts have reported more mixed findings, with only one study reporting that collaborative texts received higher ratings (Shehadeh, 2011) while three studies reported no differences in ratings (McDonough, De Vleeschauwer & Crawford, 2018; McDonough & García Fuentes, 2015; Storch, 2005). In sum, collaborative writing studies to date have provided a robust finding for accuracy benefits, but greater variation in terms of text ratings and null findings for grammatical complexity.
In addition to comparing text features, prior research has also described the focus of student talk that occurs when students carry out writing tasks, in part to explore whether peer interaction helps students focus on language form. As a result, many researchers have narrowly examined how students talk about language, specifically by analysing the occurrence of grammatical and lexical language-related episodes (e.g. Fernandez Dobao, 2012; Storch & Aldosari, 2013). However, other researchers have examined peer interaction more broadly by classifying a wider variety of student talk, including language, content, organization, task management, and off-task talk (McDonough, Crawford & De Vleeschauwer, 2016; McDonough et al., 2018; Storch, 2005; Storch & Wigglesworth, 2007; Wigglesworth & Storch, 2009). Although novice writers on English as a foreign language (EFL) focused more on language than students in settings of English as a second language (ESL) and English for academic purposes (EAP), their discussions about content, organization, and language were positively correlated with analytic text ratings (McDonough et al., 2016).
Despite the benefits associated with collaborative writing tasks, namely generating opportunities for students to talk about language, develop global aspects of writing, and increase their accuracy, there may be pedagogical issues with their use in some L2 contexts. In the EFL context reported here (see Sections II.1), collaborative writing has been used as a summative assessment task to evaluate how well students achieved instructional objectives related to writing conventions and language use. Students worked in pairs to write texts under examination conditions, and instructors assigned the same grade to the two students who collaboratively produced a single text. Although this effectively reduced the number of texts instructors had to assess, which was an important consideration for classes with large enrollments and for instructors who taught multiple classes, there were some concerns about whether each student contributed equally to the writing process and if collaborative writing was an effective means of helping each individual student develop their writing skills. Similar concerns about how to fairly assign grades to students for group projects have also been expressed in other instructional settings (Kagan, 1995; Strauss & U, 2007).
As an alternative to collaborative writing, where students work together during the planning, composing, and revising phases, collaborative prewriting has also been implemented in L2 classes as a way to potentially preserve some of the benefits of collaborative writing while addressing some of the drawbacks. Along with collaborative revision (i.e. students compose individually but then interact with their peers to revise and edit their texts), collaborative prewriting is one of the most common collaborative activities in the L2 writing classroom (Fernandez Dobao, 2012; Storch, 2005). In collaborative prewriting, students interact to generate ideas and plan, but separate to write their texts individually after they finish collaborating. Like individual planning, collaborative prewriting helps students generate ideas, organize ideas, and set goals, which may free up attentional resources that can be used while composing (Kellogg, 1988, 1990), including a focus on language (Ong, 2014). However, the L2 studies that have tested claims about the benefits of planning for L2 writing have largely focused on individual prewriting time and instructions, rather than collaborative prewriting (e.g. Ellis & Yuan, 2004; Johnson, Mercado, & Acevedo, 2012; Ojima, 2006; Ong & Zhang, 2010).
Even though collaborative prewriting may be frequently employed in L2 classrooms and may generate comparable benefits to individual planning, relatively few studies have investigated the type of student talk it elicits or demonstrated a link between student talk and text characteristics, such as analytic ratings or linguistic measures of accuracy and complexity. In an early classroom-based study, Shi (1998) compared the student talk that occurred during peer and instructor-led prewriting discussions in an ESL context, focusing narrowly on the occurrence of negotiation of meaning. In comparing student talk in the two contexts, Shi found more student negotiation of meaning during collaborative prewriting discussions than teacher-led discussions, although the negotiation moves used by teachers elicited more student responses. Student perception data revealed that more students preferred the collaborative prewriting task because it allowed them to generate and defend their ideas, but they reported difficulty selecting and organizing ideas for writing. Although Shi’s findings provide insight into the nature of student talk during collaborative prewriting discussions, it was beyond the scope of her study to analyse the students’ texts or compare the texts written under the two prewriting conditions.
In addition to analysing student talk during collaborative prewriting discussions, Neumann and McDonough (2015) also examined whether there was a relationship between student talk and analytic ratings of the students’ texts. Situated in an EAP setting over a 13-week semester, university students engaged in several collaborative prewriting discussions and then wrote paragraph-length texts. Students in Experiment 1, who carried out unstructured tasks that did not contain explicit instructions about evaluating their peers’ ideas or focusing on organization, tended to focus on content without critically engaging or evaluating their ideas. However, students in Experiment 2, who received a structured task handout with instructions to provide feedback to their peers and to create a bullet outline, engaged in more critical discussions about their ideas and discussed organizational features of their texts. In addition, the Experiment 2 findings revealed a positive relationship between the students’ critical content discussions and analytic ratings of their paragraphs for four out of six writing tasks. Students reported that the collaborative prewriting tasks were helpful for generating ideas, selecting ideas, and developing a better understanding of the topic. Furthermore, a second study in the same context (Neumann & McDonough, 2014) revealed that analytic ratings were higher when students carried out collaborative rather than individual prewriting tasks. Thus, when given more explicit instructions about what to focus on during collaborative prewriting tasks, EAP students critically engaged with their peers’ ideas, discussed the organization of their texts, and composed texts that received higher analytic ratings.
Turning to the EFL context investigated here, McDonough et al. (2018) also explored the nature of student talk during collaborative prewriting and its relationship to text features. Their goal was to describe the student talk that occurred during collaborative writing and collaborative prewriting tasks and compare the features of the students’ written texts. EFL writers were assigned to write a problem and solution paragraph under one of three conditions: collaborative writing, collaborative prewriting, and no collaboration (control), but were not given any specific instructions about what to focus on during peer interaction, as in Experiment 1 in Neumann and McDonough (2015). The analysis of student talk indicated that students in the two collaborative conditions spent a similar amount of time planning prior to beginning to compose, and oriented to content while they were collaborating, with little talk about organizational features. The students’ paragraphs were analysed for error rate (errors/word) and subordination (dependent clauses/main clauses), and rated using an analytic rubric created from the grading criteria used in the students’ EFL class. The results showed that collaborative writing texts were more accurate than individual and collaborative prewriting texts but had less subordination, while there were no differences in analytic ratings. Thus, the findings confirmed that collaborative writing texts had higher accuracy, but failed to identify any advantages for collaborative prewriting tasks compared to the no collaboration control condition, although the unstructured design of the task materials may have contributed to the findings.
In summary, collaborative writing has been associated with greater accuracy, varying effects on analytic ratings, and largely null findings for various complexity measures (Fernandez Dobao, 2012; McDonough & García Fuentes, 2015; Storch & Wigglesworth, 2007; Wigglesworth & Storch, 2009). To provide students with peer collaboration while still requiring students to compose individually, researchers have implemented collaborative prewriting in ESL (Shi, 1998), EAP (Neumann & McDonough, 2014, 2015) and EFL settings (McDonough et al., 2018). These collaborative prewriting studies have shown that students negotiate, discuss content, and talk about organization, particularly when task materials help them focus on evaluating and organizing their ideas. However, studies to date have not demonstrated any benefits for collaborative prewriting as compared to individual prewriting, as least when an unstructured task was administered (McDonough et al., 2018). Using a collaborative prewriting task modeled after the structured task implemented in previous research (Neumann & McDonough, 2014, 2015), the current study further investigates the student talk generated during collaborative prewriting discussions and explores whether it has any relationship to the features of paragraph-level texts written by Thai EFL learners. The research questions were as follows:
Is there a difference in the analytic ratings, accuracy, or complexity of paragraphs written by Thai university EFL students following either individual prewriting or collaborative prewriting?
Is there a relationship between student talk during collaborative prewriting and text features?
II Method
1 Participants and instructional context
The participants were 57 Thai university students (30 women, 27 men) studying in undergraduate degree programs at a public university in Northern Thailand. They were native speakers of Thai who ranged in age from 19 to 23 years, with a mean age of 20.7 years (SD = 1.2). They reported studying English previously for a mean of 15.4 years (SD = 2.4), which occurred through traditional instruction in primary and secondary schools. Only two students reported having spent more than 30 days in countries where English was spoken as a medium of communication. They were enrolled in two critical reading and writing EFL classes that were required for their undergraduate degree. The classes met for two, 75-minute class periods per week for a total of 30 class periods in a 17-week semester (including holidays and exams). The class targeted all four English language skills, but emphasized critical reading and writing skills. In terms of writing, the course included paragraph-level writing skills such as writing introductory and concluding sentences, giving reasons with supporting details, and using discourse markers, and focused on different types of paragraphs (descriptive, cause/effect, problem/solution, and opinion). Power was estimated at .82 based on the smallest group size (27), the alpha level (.05), and estimated medium effect sizes.
2 Materials
The materials consisted of a writing task handout and an analytic rubric. The writing topic was about proposing solutions to the problem of video games addiction among Thai university students. The instructions stated that students should write a paragraph that included topic and concluding sentences, two solutions with supporting details, and appropriate discourse markers. To accompany the writing prompt, a task handout was created by the researchers modelled after the structured prewriting task used in previous collaborative prewriting research (Neumann & McDonough, 2014, 2015). It included two sections, one for content and one for organization. The content section contained three blank spaces labeled with the terms ‘causes of games addiction’, ‘negative effects of games addiction’ and ‘solutions’ where students could make notes. The instructions stated that the students should share ideas with a partner, and provide feedback about their partner’s ideas. After sharing ideas, the students used to space in the organization section to make a bullet list of the ideas they planned to include in their paragraphs. The organization instructions stated that they should ask their partner for feedback about their bullet list. The collaborative prewriting task is provided in Appendix 1. Students in the individual prewriting group received the same handout, except that the instructions were modified to state that they should critically evaluate their ideas and outline individually before beginning to write.
An analytic rubric previously created by the researchers for use in this EFL setting (McDonough et al., 2018) was used to assess the students’ paragraphs. The rubric was based on the grading criteria used in the EFL class and contained three categories: content, organization, and language (rubric is provided in Appendix 2). Each category could be scored from 10 (good) to 1 (poor). To encourage raters to use the full scale, descriptors were provided for each category along with suggested point ranges using the criteria articulated by instructors about content (i.e. ideas expressed in the introductory sentence, solutions, and concluding sentence), organization (presences of required elements), and language (range and frequency of errors involving vocabulary, grammar, and mechanics).
3 Design
A between-groups design was used to compare the texts written by Thai EFL university students following either collaborative or individual prewriting, with the two EFL classes randomly assigned to a prewriting condition. Whereas students in the collaborative prewriting group worked with a partner to brainstorm ideas and plan their texts before composing individually, students in the individual prewriting group worked alone to both plan and write their texts. For the text analysis, the dependent variables were analytic ratings, accuracy (errors/words), and complexity (coordinated phrases plus coordinated clauses/words and dependent clauses/main clauses). The student talk data from the collaborative prewriting groups was analysed in terms of the categories identified in previous collaborative writing research: content, organization, language, task management, and off-task talk (McDonough et al., 2016, 2018; Storch, 2005; Storch & Wigglesworth, 2007; Wigglesworth & Storch, 2009). Definitions and examples are provided in Table 1. Any words or phrases in the examples when the students spoke Thai have been translated to English and are indicated in italics.
Student talk episodes with examples.
4 Procedure
The writing task was administered by the first and second researchers during one regularly-scheduled class period (75 minutes) in each of the two EFL classes. After distributing the task handout, students in the collaborative prewriting group self-selected a partner and worked in pairs to complete the content and organization sections. When a pair finished discussing their ideas and outlines, both students were given lined paper for writing their texts by hand and asked to separate so they could write individually. While collaborating, students could interact with their partner, but conversation across pairs was discouraged. For the individual prewriting group, students received the task handout and worked through the content and organization sections by themselves. When they were ready to begin writing, they were given paper for composing their paragraphs. The students had the entire class period (75 minutes) to plan and compose their written texts. Across the two writing conditions, the amount of time students chose to plan prior to writing was similar, ranging from five to 15 minutes. All students had access to their task handouts with their notes while they were writing. However, they were not allowed to consult electronic devices such as phones or tablets. Individual audio-recorders were used to record the interaction between pairs in the collaborative prewriting group. All students were instructed to interact and take notes in English.
5 Analysis
The students’ hand-written paragraphs were converted into Microsoft Word documents by research assistants without making any changes to the formatting, spelling, punctuation, or language use, except for ignoring words and sentences that the students had crossed out. The electronic texts were independently rated by two research assistants with experience rating L2 texts using the analytic rubric. Their interrater reliability was .97. The means for each subscore and the total score assigned by the two raters were used for subsequent analyses. To facilitate comparison with previous collaborative writing studies (e.g. Fernandez Dobao, 2012; McDonough & García-Fuentes, 2016; Storch, 2005) and to reflect the EFL instructors’ goals of improving student accuracy and increasing their use of compound and complex sentences, the texts were coded by the third author for accuracy (errors/word) and two measures of grammatical complexity most relevant for these participants’ proficiency (Norris and Ortega, 2009): coordination (coordinated phrases plus coordinated clauses/words) and subordination (dependent clauses/main clauses). A second coder analysed a subset of the data (26%) and the interclass correlation coefficients were .99 for errors, .99 for independent clauses, .99 for dependent clauses, and .91 for coordination.
The audio-recordings were transcribed and verified by research assistants who spoke Thai as their first language (L1). Although the students in the collaborative prewriting group were instructed to speak English, they spoke both Thai and English while collaborating, as is typical in this EFL context. The transcripts were segmented into episodes based on the focus of student talk (content, organization, language, task clarification and management, and off-task talk), which were adapted from the coding categories used in previous studies (McDonough et al., 2016, Storch, 2005; Storch & Wigglesworth, 2007; Wigglesworth & Storch, 2009). Definitions and examples of the episodes were provided in Table 1. A research assistant who spoke Thai as a first language coded the entire dataset, and a subset (20%) was coded by the first researcher for interrater reliability. Interrater reliability was calculated as two-way random intraclass correlation coefficients for each coding category and the values were as follows: task management (.99), content (.95), language (.99), organization (.95), and off-task talk (.92). Alpha was set at .05.
III Results
The first research question asked whether there were any differences in the features of paragraphs written by Thai EFL students following either individual or collaborative prewriting. Prior to comparing the texts, we checked text length to assess comparability between the groups. The collaborative prewriting texts had a mean length of 116 words (SD = 47), while the individual prewriting texts were only 67 words in length (SD = 41), which was a statistically significant difference: t(55) = 4.18, p = .001, d = 1.11. However, the accuracy and complexity measures were normed (either by words or main clauses), which helped mitigate potential differences due to text length.
For the analytic ratings, the collaborative prewriting texts were rated higher in content, organization, and language than the individual prewriting texts (see Table 2), which resulted in higher total scores. Independent-samples t-tests (equal variance assumed) using an adjusted p value of .008 (.05/6) to account for all statistical comparisons (analytic rating and text features) indicated that the content, organization, and language subscores were all significantly higher for the collaborative prewriting texts, with large effect sizes (Plonsky & Oswald, 2014).
Analytic ratings by text type.
The collaborative prewriting texts also had lower error rates and greater coordination and subordination than the individual prewriting texts, as shown in Table 3. Independent-samples t-tests (equal variance assumed) showed that the collaborative prewriting texts had significantly lower error rates, but there was no statistically significant difference for either complexity measure. The effect sizes for accuracy and subordination were similar (.59 and .57, respectively), while the effect size for coordination was lower (.39).
Linguistic measures by text type.
To summarize the findings of the text comparison, the collaborative prewriting texts had lower error rates and received higher analytic ratings than the individual prewriting texts. To illustrate these differences, representative examples are provided in Table 4. Whereas the collaborative prewriting text (89 words, 19.5/30 text rating) contains more content development and fewer errors, the individual prewriting text (77 words and 12/30 text rating) has less cohesive and elaborated content information and contains numerous errors.
Collaborative and individual prewriting texts.
The second research question asked about the relationship between student talk during the collaborative prewriting task and the text features. In terms of how long the students talked prior to beginning to write, they spent a mean of 11.6 minutes collaborating (SD = 3.5) before they separated to write individually. For the amount and type of student talk, Table 5 summarizes the number of episodes across coding categories. The students talked about content most frequently (48% of the episodes), followed by task management, organization, language, and off-task talk.
Student talk episodes by type.
To assess the relationship between the amount and type of student talk and text features, Spearman rho correlations were obtained for text length (in words), ratings, accuracy, coordination, and subordination. Only one aspect of student talk was significantly correlated with text features: length of prewriting (see Table 6). More specifically, the amount of time that students spent together before separating to write individually was negatively correlated with accuracy. The correlation was negative because as prewriting time increased, the students’ error rates decreased. To check whether prewriting time was associated with lower error rates because of student talk about language, the Spearman rho correlation coefficient between prewriting time and language episodes was obtained, which was significant (rho = .52, p = .003).
Relationships among text features and student talk.
Note. * p < .05.
Thus, the correlation results indicated that the duration of student talk during collaborative prewriting was associated with lower error rates, which may be due to the positive relationship between prewriting time and language episodes.
IV Discussion
The primary goal of the current study was to explore whether collaborative prewriting had any of the advantages associated with collaborative writing, namely greater accuracy and higher analytic ratings, as compared to individual prewriting. The findings showed that students who collaborated during the prewriting phase wrote texts that were more accurate and received higher ratings than the texts written by students who carried out individual prewriting, although there was no significant difference for complexity measures (coordination and subordination). The results parallel the findings of previous collaborative writing research, which found collaborative texts to be more accurate and highly rated than individual texts (Fernandez Dobao, 2012; McDonough & García Fuentes, 2015; McDonough et al., 2018; Shehadeh, 2011; Storch, 2005; Storch & Wigglesworth, 2007; Wigglesworth & Storch, 2009). However, collaborative prewriting did not positively impact complexity, which also extends the findings of collaborative writing studies that reported null findings for a variety of complexity measures (Storch, 2005; Storch & Wigglesworth, 2007; Wigglesworth & Storch, 2009). Taken together, the findings suggest that collaborative prewriting may generate similar benefits as collaborative writing for novice EFL writers, at least in terms of analytic ratings and accuracy.
However, it is important to note that the findings that collaborative prewriting texts were more accurate and highly rated than individual prewriting texts diverge from those of a previous comparative study in a similar EFL context (McDonough et al., 2018), which found no significant differences between collaborative and individual prewriting. There are two possible explanations for the divergent findings. First, as described previously, the current study administered a structured task handout which provided students with more guidance about what to discuss concerning both content and organization and encouraged students to provide feedback for their partner’s ideas. In contrast, the prior study implemented an unstructured task in which students were simply asked to brainstorm and discuss ideas with a partner. By focusing students’ attention on specific aspects of the topic in terms of content (causes, effects and potential solutions), and organizational features of their paragraphs, the structured task may have encouraged students to collaborate in ways that more directly influenced their subsequent writing. A second possible explanation is variation in the students’ writing abilities, as the students in the current study were less skilled writers. Whereas students in the current study generated short texts (means of 67 and 117 words), the students in the prior study wrote longer texts (150 words) that received higher ratings (approximately 23/30 points) and had lower error rates (.05). Furthermore, these students took longer to collaborate before writing (M = 11.6 minutes) than students in the prior study (M = 7.8 minutes). There were also differences in the students’ degree programs. While these students were enrolled in a variety of programs across several faculties, the students in the prior study were all studying in the Faculty of Medicine, which has some of the highest requirements for admittance to the university. In sum, due to differences in both the task materials and the students’ academic skills between experiments, it is not possible to determine the exact reason for the divergent findings.
Turning to the findings about student talk, the results confirm that students discuss content most frequently when carrying out collaborative prewriting tasks (McDonough et al., 2018; Neumann & McDonough, 2015), which has also been reported for collaborative writing tasks (Storch, 2005; Storch & Wigglesworth, 2007; Wigglesworth & Storch, 2009). However, providing a structured prewriting task handout that included a section specifically about organization was an effective way to increase student talk about organization. As described previously, Neumann and McDonough (2015) found no organization episodes in their first experiment when the instructions did not explicitly state that students should plan and discuss the organization of their texts, but modifying the task materials in the second experiment to include an organization section elicited a total of 60 organizational episodes. Similarly, our previous study in this EFL setting that administered an unstructured prewriting task found that only 6% of the student talk episodes involved organization (McDonough et al., 2018), which contrasts with the 13% elicited by the structured prewriting task in the current study. Taken together, the findings from both the EAP and EFL settings suggest that the design of collaborative prewriting materials may play an important role in encouraging students to discuss organizational features of their texts.
In terms of the relationship between student talk and text features, there was a significant, negative correlation between prewriting time and accuracy. In other words, the more time that the students interacted before separating to write individually, the lower their error rates. Furthermore, there was a positive relationship between prewriting time and language episodes, which may help explain why longer interaction during prewriting was associated with lower error rates. However, it is important to note that language episodes accounted for only 11% of the student talk episodes, and there was little relationship between language episodes and accuracy (rho = −.12). Taken together, these findings suggests that it may have been interaction more generally, rather than discussions about specific language forms, that accounted for the benefits of collaborative prewriting. Rather than talk about language during the collaborative prewriting discussions, these students brainstormed ideas, evaluated and selected ideas, and discussed how to approach the writing task. As a result, their student talk contained more content (48%) and task management episodes (22%). Having had the opportunity to collaboratively plan their paragraphs, students may then have greater attentional resources available for linguistic accuracy while writing individually. In other words, for these novice writers, collaborative prewriting may have facilitated accuracy by helping them plan the content and organization of their texts so that they were able to devote more attentional resources to language while writing individually. This possibility should be considered speculative as the current study was not designed to explore the link between collaborative prewriting and accuracy specifically.
Although the findings provide insight into the benefits of collaborative prewriting, there were some limitations that may negatively impact the generalizability of the findings. First, while these novice English L2 writers benefited from the opportunity to collaborate with a peer prior to writing individually, more proficient writers may not experience the same effects because they may have sufficient linguistic and writing skills for individual prewriting. Additional comparative studies of collaborative prewriting tasks are needed to determine whether proficiency is a mediating variable. Furthermore, in contexts such as the one investigated here, where standardized proficiency test scores are not available, it may be useful to implement other measures of global proficiency, such as cloze tests (e.g. Bachman, 1985; Brown, 1983), elicited imitation tasks (for a recent study, see Gaillard & Tremblay, 2016), or for students with advanced proficiency, the LexTALE test (Lemhöfer & Broersma, 2012). However, elicited imitation tasks and the LexTALE test may not be feasible to administer in a classroom setting.
Insight into the nature of student talk during the collaborative prewriting task was obtained through the audio-recordings, but the current study did not elicit information from students in the individual prewriting group. The audio-recordings from the collaborative prewriting interactions indicated that the students spoke their L1 extensively while exchanging ideas with a partner, which has been previously reported in collaborative writing research (McDonough et al., 2016). For future classroom-based studies, interviews or focus groups with volunteer participants could be administered after class to learn more about what students in the individual prewriting group did prior to writing, and the extent to which they also drew upon their L1 when planning. Future research should also compare the longer-term benefits of collaborative and individual prewriting to determine which type of practice is most effective at promoting writing development. Our focus here was on paragraph-level writing, specifically problem/solution paragraphs, because it was relevant for the students’ EFL class. Additional studies of other paragraph types and longer texts are needed to identify whether the benefits of collaborative writing are mediated by text type or length. Finally, our measures of accuracy and complexity were selected for their ecological validity for teachers in this EFL context and for comparison with prior collaborative writing studies and linguistic complexity research (Bulté & Housen, 2014; Norris & Ortega, 2009; Pallotti, 2009) that used these measures. Future studies involving students from higher proficiency levels may benefit by using linguistic measures, such as phrasal complexity, that are associated with academic writing (Biber, Gray, & Poonpon, 2011).
To conclude, current study explored whether any of the advantages associated with collaborative writing, such as greater accuracy and higher ratings, would occur when students had opportunities to collaborate during the prewriting phase only. The findings indicated that collaborative prewriting resulted in longer, more accurate and highly rated texts, but did not impact coordination or subordination. Furthermore, the majority of the student talk during collaborative prewriting concerned content, with the total length of prewriting time associated with lower error rates. Our future studies aim to explore additional factors that may mediate the benefits of collaborative prewriting during L2 writing, such as task design, text type, text length, and proficiency, and further specify the nature of individual and collaborative prewriting. In addition, we plan to investigate the longer-terms benefits of collaboration in the writing process to determine when it most effectively leads to individual writing development.
Footnotes
Appendix 1
Acknowledgements
We would like to thank the research assistants for their help with transcription and data analysis: Phung Dao, Alexandre Dion, and Amornthep Jaidee.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We are grateful for the funding provided by the Canada Research Chairs program (950-231218).