Writing Process Research: L2 Writing

Abstract

Over the past 40 years, L2 educators and assessors have realized the importance of investigating the writing process to inform teaching, learning, and assessment practices.
Understanding what L2 writers do and think can link the writing task, the underlying construct, and the cognitive mechanisms involved.
This understanding improves feedback from teachers and ensures that score inferences in assessments are linked to linguistic knowledge and intended mental architecture.
Despite research in L1 and L2 writing processes, many studies provide only lists of strategies and observed phenomena.
This paper reviews L2 writing process research by presenting influential theoretical approaches and evaluating their contributions through empirical research.
The paper also explores the investigation of writing processes in assessment contexts.
It discusses how a theory-based understanding can be achieved using innovative assessment tools like scenario-based assessment (SBA) with computer technology.
Keywords: cognition, metacognition, SBA, technology, writing strategies, writing process.

Introduction

Studies of L2 writing performance are evolving to incorporate findings from writing ability and cognitive theories.
The process approach emerged in the early 1980s, shifting focus from the writing product to the writer and the writing process.
Unlike product-oriented approaches, the process-oriented approach considers linguistic, cognitive, and metacognitive aspects, including brainstorming, planning, generating ideas, revising, and reflecting.
Writing process is the combination of strategies used by language learners to produce text, including physical actions and underlying cognitive and metacognitive processes.
In L2 assessment, researchers aim to link assessment tasks to cognitive mechanisms and ensure score-based inferences depend on linguistic knowledge and mental cognitive architecture (e.g., working and long-term memory).
Writing, being complex, places affective and cognitive demands on learners, particularly under assessment conditions.
Research into writing processes in assessment contexts is scarce but valuable for understanding mental processes leading to the final product.
This paper reviews L2 writing process literature by presenting theoretical models, reviewing methodologies and findings of empirical research, and examining recent research in assessment contexts.

Theoretical Models of Writing Process

L2 writing research often uses L1 process-oriented models (e.g., Bereiter & Scardamalia, 1987; Flower & Hayes, 1981; Hayes, 1996, 2012; Kellogg, 2001) to describe mental processes writers use to compose texts.
Flower and Hayes (1980) proposed a framework dividing the writing process into planning, translating ideas into words, and reviewing, all monitored by a control process.
These subprocesses interact recursively with the task environment, control level, and resource level.
Criticisms of the Hayes-Flower framework include its failure to consider learners’ linguistic knowledge and external social factors.
Despite these criticisms, the model has been widely used in L2 process studies, and several features remain valid in modern representations of writing.
Bereiter and Scardamalia (1987) suggested two composing models: knowledge telling and knowledge transforming.
Knowledge telling uses maximum language competence through natural social experience, while knowledge transforming requires reprocessing knowledge independently from social interaction.
Beginner writers use knowledge telling to retrieve content from memory and quickly produce discourse.
Advanced writers use knowledge transforming strategies to develop their knowledge by reviewing their writing.
Bereiter and Scardamalia’s model was not as widely used in L2 studies because L2 writers’ strategies differ from those of L1 writers.
The model was also criticized for not explaining the transformation from knowledge telling to knowledge transforming.
Grabe and Kaplan (1996) proposed a process-oriented model based on Chapelle et al.’s (1993) communicative language performance model.
This model comprises both internal processing and external factors, including situation variables (participants, setting, task, text, topic) and performance (textual output).
Internal processes involve processing activities in verbal working memory.
Writers set goals based on context or internal goal setting, which activates a cycle of operations in verbal processing, mediated by metacognitive factors.
Language competence and knowledge of the world are used to produce internal output, which is compared to task goals and revised if necessary.
The processing sequence is similar to the Hayes-Flower model and Bereiter and Scardamalia’s framework, but Grabe and Kaplan’s model incorporates contextual influences, task specifications, and different processing sources.
Kellogg (1996) proposed a model of writing consisting of six basic processes: planning, translating, programming, executing, reading, and editing.
Kellogg’s model differs from the Hayes-Flower model in the working memory demands specified for each subprocess.
For example, generating and organizing ideas pose similar demands on working memory.
Translating includes linguistic encoding (grammatical, phonological, and orthographic encoding), which translates activated concepts into words and sentences.
Kellogg et al. (2013) argue that research supports the assumptions made in Kellogg’s model, although working memory resources may differ depending on the task.
Deane et al. (2008) proposed a writing competency model that includes contextual and social factors as well as critical thinking abilities.
The model comprises three strands: language and literacy skills, writing process management skills, and critical thinking for writing.
Strand I includes skills such as drafting and editing.
Strand II includes writing-process management skills such as planning and evaluating.
Strand III refers to critical thinking abilities that enable reasoning about content and social context, allowing the writer to explain, hypothesize, support, refute, synthesize, report, narrate, or describe.
The elements of this model can vary based on the requirements of the specific genre and task, allowing for different combinations of skills and strategies.
This comprehensive model can be applied to the investigation of L2 writing process in innovative tests such as scenario-based assessments.
Theoretical frameworks outline the cognitive, metacognitive, and physical processes that writers go through to produce text.
The major components (planning, generating ideas, writing, organizing, and editing) are shared among different models but differ in their emphasis on linguistic, social, and contextual features.
Writing involves complex cognitive mechanisms and problem-solving abilities.

Investigating L2 Writing Process: L2 Empirical Studies

The investigation of L2 writing processes is motivated by pedagogical purposes, with the assumption that it could provide a sound theoretical basis for teaching practices.
Many L2 writing process studies investigate the effectiveness of L1 process theories for L2 writing instruction (Cumming, 1989; Silva, 1993; Zamel, 1982).
Influenced by the models of Flower and Hayes (1980) and Bereiter and Scardamalia (1987), many studies focus on the differences between L1 and L2 writing, addressing skilled vs. unskilled writers and the effects of fluency (e.g., Chenoweth & Hayes, 2001).
Some studies investigate the writing process at a global level (e.g., Raimes, 1987) or explore specific sub-processes such as planning or revision (e.g., Cumming, 1989; Krapels, 1990b; Roca de Larios et al., 2001).
Other studies examine differences in time allocation (Roca de Larios et al., 2008) or take a developmental approach and examine the development of writing strategies over time (e.g., Sasaki, 2002).
Studies by Zamel (1982, 1983) and Raimes (1985, 1987) are among the first to investigate the L2 writing process.
Zamel (1983) examined the composing processes of skilled and unskilled ESL students by observing their writing and interviewing them.
Zamel argued that ESL students use thinking, writing, and rewriting recursively.
Unskilled writers were focused on correcting grammatical and lexical errors, while skilled writers were focused on developing their ideas, similar to native speakers.
Zamel (1983) concluded that writing is a process of discovering ideas and finding the best framework to present and develop these ideas.
Raimes (1987) found the skilled-unskilled distinction in Zamel’s study to be blurry and argued that it can only be applied after considering their language level and L1 writing ability.
Using different tasks and think-aloud protocols, Raimes (1987) examined ESL students with different levels of instruction and reported that lower L2 proficiency students acted similarly to less skilled L1 writers in terms of the time spent on prewriting and planning.
More advanced L2 students spent more time reading and revising.
Raimes’s (1987) procedures and the use of different tasks and proficiency levels was a valuable contribution to the field, despite limitations due to a small sample size.
Cumming (1989) agreed that some L2 writing processes are comparable to L1 but questioned the claims that L2 proficiency does not affect the writing process.
Cumming (1989) argued that L2 proficiency and writing expertise must be treated as separate variables, and studies should examine the effect of tasks more systematically.
Cumming (1989) conducted a larger scale study and found that students from higher L2 proficiency and writing expertise scored higher, with the difference being more apparent for more cognitively demanding tasks.
Experts attended to word-level evaluations after making decisions about the gist and organization of their texts, while non-expert writers were focused on immediate planning and correction.
Cumming (1989) also reported that students applied their L1 writing expertise in their L2 writing, and L2 proficiency level did not affect the quality of their writing process and problem-solving behaviors.
In terms of methodology, the validity of verbal protocols was questioned for causing reactivity, which led researchers to use computer programs and video taping.
Sasaki (2002) examined Japanese L2 learners and asked them to explain their writing behavior while watching their tape-recorded composition sessions.
Sasaki’s (2002) study showed that expert writers spent more time planning and produced lengthier texts in a shorter time compared to novice counterparts.
Sasaki (2002) also found that novice writers spent more time on local planning as well as translating from their L1 into L2 due to low L2 proficiency.
A concurrent line of research examined the context of writing and the effect of a set time limit on L2 writing process.
Research findings suggested that L2 learners’ process and subprocesses differ during assessment conditions compared to non-assessment settings (e.g., Ellis & Yuan 2004; Hall, 1991).
Khuder and Harwood (2015) examined the writing process of ESL students writing an argumentative essay under both test and non-test situations and found that while translation and surface revision were used frequently in test situations, the students did more planning and evaluation in non-test conditions.
There were differences in time allocation and the role of task type on L2 writing process also started to receive attention (Cumming et al., 2005; Plakans, 2008).
These combined effects motivated the study of L2 writing process under assessment conditions.

Investigating L2 Writing Process in Assessment Contexts

Discovering and understanding writing processes in assessment contexts can give us useful information about the cognitive mechanisms underlying performance in writing tasks.
This information assures that these mechanisms are congruent with the intended construct and relevant for successful performance (Purpura, 2013).
Assessment context can affect the writer’s engagement with the text (Wolcott, 1987).
Global and local writing processes have been investigated in assessment contexts, as well as across different task types, including the use of source materials (Barkaoui, 2016; Cumming et al., 2005; Gebril & Plakans, 2013; Révész et al., 2017).
Yang and Plakans (2012) found that L2 proficiency did not necessarily lead to better source use ability and a higher score and that writers had to activate a combination of self-regulatory strategies to perform successfully but test wiseness strategies had a negative effect, while discourse synthesis strategies had a significantly positive effect.
Barkaoui (2015) accounted for L2 proficiency and task-type differences and found that evaluation of language and local text was the most frequently observed activity in all tasks, writers interacted with the integrated writing task more often and generated source-based content, but revised language more frequently in the independent task and lower L2 proficiency level students did more planning and organization activities than high L2 participants.
Simulated recalls may not fully capture the broad range of writing activities that writers engage in during the test.
Increasing L2 process studies rely on new techniques such as keystroke logging programs.
Révész et al. (2017) examined the effect of task complexity on L2 writers’ fluency, pausing and revision behavior and the underpinning cognitive processes.
Test takers reported the simple task as requiring less mental effort and being less difficult in terms of content generation, paused less, and translated more than the complex task.
Barkaoui (2019) found that the participants completed the integrated task faster than the independent task, but L2 proficiency had no effect on the overall pause duration.
Michel et al. (2020) found that task type had a significant effect on the writers’ speed and fluency they had less frequent and shorter pauses in the independent task since they did not have to refer to notes from the listening and reading sections like they did in the integrated task and students relied heavily on translation, but they spent more time planning during the independent task, which was due to lack of access to source materials.
Choi and Deane (2021) explored the stability and predictive potential of writing process features of adult EFL learners and found that only stable features (such as total keystrokes and within-word pauses) correlated with the quality of the response and overall English proficiency.
The studies conducted on L2 writing process in assessment contexts suggest that assessment conditions affect a variety of writing behaviors and cognitive mechanisms such as time allocation, pausing, planning, monitoring, fluency, revision patterns and perception of difficulty level.
Most writing process studies in assessment contexts used samples of high-stake tests such as TOEFL iBT and showed that writing processes are also affected by the type of task and having access to content support or source materials. For example, integrated task types can be easier for higher L2 proficiency level writers by reducing planning processes and allowing them to focus on higher order processes such as global monitoring and revisions.
The effects of L2 proficiency level on writing process were further evidenced in studies showing that high proficiency students normally engaged in more planning, rescanning, revising, and editing.

Towards a Better Understanding of L2 Writing Process: Innovative Assessment Tools and Technologies

An investigation of the L2 writing process in assessment contexts has a number of benefits for language learners, educators, and assessors.
This information can be used to provide feedback and design writing tests that reflect the use of expected competencies during a task.
Scenario-based assessment (SBA) can provide the opportunity to investigate L2 writing process and strategies in a user-friendly assessment design (Deane et al., 2015).
SBA simulates an authentic, purposeful situation where students write to achieve a meaningful objective, leading to learning in addition to assessment and takes into account indicators of other skills necessary to accomplish a writing task successfully (e.g., reading comprehension).
Keystroke logging can be used in combination with other tools to identify writing process features relevant to SBA (e.g., source use, planning, revision, text production) and identify the errors and linguistic complexity of the final text.
This line of research can offer perspectives for the conceptualization of L2 writing process and validity evidence for new test designs.
In order to have multiple levels of analysis, the analysis of the writing process needs to be ideally rooted in a theoretical model of cognition and different types of cognitive and metacognitive strategies involved in writing strategies.
This can be accomplished by using SBAs as more authentic assessments of writing ability, where the L2 writing process can be better captured and analyzed in reference to particular cognitive and metacognitive processes underlying the writing activities.
Moreover, including a wider population of adult ESL learners from different backgrounds and L2 proficiency levels can potentially help us achieve an even better understanding of L2 process in assessment contexts.

Conclusion

This paper presented an overview of major theoretical frameworks and approaches to the conceptualization and investigation of writing process.
It then presented and discussed the contributions made by some representative empirical research in general writing and in assessment contexts.
Finally, it examined how SBA can be used in L2 assessment contexts to link L2 writing process of more varied population of language users to a theoretically grounded model of cognition and help us achieve a more valid analysis of writing process data and a better understanding of the concept of L2 writing process.
Developments in theoretical models and research methodologies have led the researchers to consider individual and social factors as well as the requirements of specific writing tasks and genres in writing process research (Deane et al., 2008).
More authentic task types and simulated assessment conditions are being used along with less intrusive tools such as keystroke logs and other computer software.
One shortcoming with previous research is that the majority of L2 writing process studies relied on high-stakes standardized tests assumed a relationship between the observable writing strategies and cognitive and metacognitive processes underlying these strategies.
It is, therefore, essential to address the issues and implications emerging from past and current L2 writing process research by investigating the writing behaviors and underlying cognitive mechanisms using innovative types of assessment that consider the wide range of contextual factors, and personal and social knowledge, skills and abilities needed to successfully fulfill a writing task.