The effects of task design variables and corrective feedback on EFL learners’ writing complexity and accuracy

Abstract

Task-based language teaching (TBLT) has occupied an important place in the field of language education; however, some of TBLT dimensions that pertain to the interaction between task design features, written corrective feedback (WCF), and learners’ performance have not received adequate attention in past studies. To fill this gap, the current study investigates how task complexity, task condition, and their interaction determine language learners’ gain from WCF. To conduct the study, 223 participants were randomly assigned into three experimental and one control groups. Participants in the experimental groups received a pretest, followed by three treatment sessions, during which they completed simple or complex writing tasks either individually or collaboratively. They received feedback on their performance and finally completed two posttests. Participants in the control group received pretest, posttest, and regular classroom instruction (instead of the treatment), but they did not receive WCF. Results of statistical analyses demonstrated that task implementation condition had a more highlighted role than task complexity in determining learners’ gain from WCF, but the interplay between the two variables didn’t affect participants’ writing complexity and accuracy. These findings lend partial support to Skehan’s Tradeoff Hypothesis and to Vygotsky’s sociocultural theory of language development. Implications for language instructors and syllabus designers will be discussed.

Keywords:

• Writing task
• task complexity
• task condition
• error correction
• task-based education

REVIEWING EDITOR:

• John Chiang, Communication and Media, State University of New York, United States

SUBJECTS:

• Language & Linguistics
• Language Teaching & Learning
• General Language Reference

Introduction

To acquire writing skills in English as a foreign language (EFL) classes, learners should develop their linguistic and communicative competence; therefore, many of them find this process to be a demanding undertaking. But, this challenging endeavor—as Cumming (2001) argues—has a special status in second and foreign language education. Carson (2001) believes that classroom activities engaging learners in negotiation of meaning are more likely to foster their written performance. Task-based language teaching (TBLT) is an innovative language teaching approach that strives to provide the opportunity for negotiation and interaction by involving learners in meaning-focused activities. Second language acquisition (SLA) researchers (e.g. García Mayo, 2007; Skehan, 1996; Van den Branden, 2006) have provided sufficient theoretical and experimental support for TBLT, but selecting appropriate tasks and sequencing them is still an unresolved issue for researchers and practitioners (Baralt et al., 2014). This issue is more critical in the case of L2 writing, mainly because it has received scant attention in TBLT-related research.

One of the big challenges facing SLA researchers concerned with the issue of task sequencing in TBLT framework is gauging the influence of task design features and variables on L2 learners’ performance (Robinson, 2001). Thus far, the majority of TBLT studies have primarily examined the interaction of task variables in second language (L2) learners’ oral production and have neglected their written performance (Carless, 2012; Cook, 2009). Marginalizing writing is occurring despite the fact that this skill, owing to its problem-solving nature, has a rich capacity to foster second language learning and use. This wide focus on oral production has negatively influenced TBLT research and has retarded the broadening of its conceptual and empirical horizons. Researching into other language skills (e.g. writing) can expand the rich educational potentials of TBLT (Byrnes & Manchón , 2014).

Task complexity is a task design feature that results from information processing demands of a task and manipulating it can affect language performance (Wood, 1986). Moreover, the circumstances under which a task is implemented can cause considerable changes in language learners’ performance in a specific task (Larsen-Freeman & Long, 1992). Similarly, the interaction between task complexity and task condition may affect L2 learners’ linguistic output (Michel et al., 2007); nevertheless, as Rahimi and Zhang (2017) contend the effect of this interaction on L2 writing production has not been researched exhaustively. In the present study, task complexity is manipulated by increasing the number of elements and task condition is manipulated by asking participants to do the writing task either individually or cooperatively. In previous research, “elements” have referred to number of options or considerations to take into account when making a decision. Therefore, if an L2 learner is asked to perform a writing task, which includes choosing an apartment from among several ones and he/she is required to select it considering factors such as location, price, rent, and parking space he/she should do the task taking four elements into account.

Furthermore, the consensus among SLA researchers is that TBLT-based syllabi should focus on creating meaning-oriented tasks and at the same time implicitly target specific lexicogrammatical forms (Nunan, 2004; Van Compernolle, 2014). Corrective Feedback (CF) is a common pedagogical intervention that directs learners’ attention to both form and meaning. Numerous studies have explored the effect of CF on the written performance of EFL learners, but this type of feedback has not received due attention in TBLT framework. What’s more, the interplay of CF with task complexity and task condition, especially in the written modality, has not been investigated sufficiently by SLA researchers so far. As Manchón (2014) argues CF has not been a component of TBLT theoretical accounts. She, therefore, suggests that the synergies between CF and TBLT research and theory should be further scrutinized.

Past research (e.g., Baralt, 2013; Granena & Yilmaz, 2021; Sasayama, 2011) has investigated the effect of task complexity, task condition, and CF on L2 learners’ written performance; however, most of the previous papers have dealt with only one or two dimensions of the issue and failed to consider the interaction between all three variables and their impact on L2 learners’ writing. The current study considers the whole story and seeks to examine if the effect of WCF on EFL learners’ writing is mediated by the complexity of the tasks and by the conditions under which they are implemented. Although writing can be evaluated by measuring various indices, following the tradition established by previous studies investigating this issue in TBLT framework (e.g., Ishikawa, 2007; Tavakoli & Skehan, 2005), it was decided to evaluate the writing performance of the participants by gauging the complexity and accuracy of their written output.

The current study can have some important implications for researchers and practitioners. First, language teachers can use the findings of this study in designing the writing tasks and sequencing them. Second, curriculum planners and materials developers can use the results of the study in preparing pedagogically sound curriculum and textbooks for language learners. Third, SLA researchers can receive new insights into the interaction between task design factors and CF and consider them when suggesting new theories about second language development.

Literature review

Task complexity

One topical issue in TBLT-framed research is assessing the influence of different task design variables, including task complexity, on learners’ language output. Despite the fact that different models and frameworks have been suggested by SLA researchers (e.g., Candlin, 1987, Ellis, 2003, Long, 1985) for gauging task complexity level, Cognition Hypothesis (Robinson, 2001, 2003, 2009) and Tradeoff Hypothesis (Skehan, 1998, 2009) are two theoretical models that have gained the most acceptance in the field of language education.

Following Schmidt’s (1990) and VanPatten’s (1990) arguments regarding constraints in the processing capacity of learners, Skehan (1998) presented the first psycholinguistic account for the effects of task demands on L2 learning and performance. Skehan’s (1998, 2009) Tradeoff Hypothesis, emphasizing the single resource model of attention, assumes that cognitively complex tasks put L2 learners under pressure such that tradeoff effects between form and meaning occur. Then, limited attention for form generates another tradeoff between complexity and accuracy with the result being that increasing task complexity will foster either accuracy or complexity of linguistic output—but not both of them. Conversely, less complex tasks leave more attentional resources that can be allocated to processing linguistic code, and consequently, they provide more gains from focus on language forms.

Alternatively, following multiple resource model of attention, Robinson (2001, 2003) postulated Cognition Hypothesis and the Triadic Componential Framework (TCF) that accommodates the fundamental pedagogical tenets of this hypothesis. Based on this hypothesis, task complexity has two dimensions: resource-directing variables and resource-dispersing variables. Robinson (2003) reasons that increasing task complexity along resource-directing variables (e.g., ± intentional reasoning) places greater conceptual demands on the learner and channels his/her attentional resources to linguistic structures required for performing the task and also to the provided CF. This is a process leading to the facilitation of interlanguage development and simultaneous improvement in the accuracy and complexity of language production. Contrarily, increasing complexity along resource-dispersing variables (e.g., ±planning time) places greater demands on the ability to access currently established and developing L2 knowledge and results in decreased accuracy, complexity, and fluency in the linguistic output.

Both Skehan (1998) and Robinson (2003) concur that complexifying tasks along resource-dispersing variables jeopardizes complexity, accuracy, and fluency of language output. However, according to Cognition Hypothesis, language problems are not a result of resource limitations, but rather they stem from interference problems and lack of control over attention. Due to this premise, the two models divaricate when it comes to discuss the effect of resource-directing variables. In contrast with Robinson (2003), Skehan (1998)—not dividing task complexity factors into resource-directing and resource-dispersing—holds the view that merely increasing task complexity along resource-directing variables will not cause a concurrent increase in both complexity and fluency of the language output. Rather, he maintains that manipulating other task features and their implementation condition may bring out this dual improvement.

Examining the influence of task complexity on language learners’ writing performance has been the focus of some studies conducted by SLA researchers (e.g. Ishikawa, 2007; Johnson 2017; Kormos, 2011). Ishikawa (2007) divided 52 Japanese language learners into two groups, each of them performing either the simple or the complex version of a narrative writing task. Complexity was manipulated by adopting here and now/there and then variable. He found out that learners doing complex task gained more in terms of accuracy and complexity, but their fluency decreased while performing the task. Frear and Bitchener (2015) studied the effect of increasing the reasoning demands of the tasks and the number of elements on written output of intermediate EFL learners. They concluded that task complexity had a negative effect on syntactic complexity of the texts, but it made a positive effect on their lexical complexity. The findings of Révész et al. (2017) indicated that complexifying the task did not affect L2 learners’ fluency, but it improved the lexical and syntactic complexity of their writing.

Operationalization of task complexity in the current study was achieved by manipulating the number of elements that participants considered while performing the treatment task. According to the Cognition Hypothesis (Robinson, 2001), a task that requires considering many elements invites more syntactically complex structures and more varied and specific lexical items. Classifying the number of elements as an important task design feature, Ellis (2003) conceptualizes that processing a task which includes few concrete elements whose relationships are well-defined is simpler than processing a task which has many abstract elements that are not organized into a specific and easily distinguishable structure. Viewing task difficulty inherent in tasks themselves, Skehan (2014) elucidates that the number of elements in a task may impact task complexity; however, this impact can be mediated by other factors such as concreteness/abstractness of the elements and nature of pedagogic context. He suggests that to predict the complexity of a task, the interaction between task design features (e.g., number of elements) and other factors that might create different interpretations for that task be considered.

Task condition

Task condition is another task design feature whose manipulation can bring about considerable changes in task performance. Task condition contributes to task variation because the same task administered under different conditions will yield different outcomes (Larsen-Freeman & Long, 1992). Robinson (2007), in his Triadic Componential Framework (TCF), places task conditions under two categories: 1. Participation variables (e.g., ± convergent solution) which make interactional demands and 2. Participant variables (e.g., ± same proficiency) which make interactant demands. Therefore, the criterion for classifying task conditions is interactional demands they impose on language learners.

On the contrary, Skehan’s (1998) framework incorporates factors such as time pressure, scale, modality, and opportunity for control under the category of communicative stress. Skehan (2014) conceptualizes that variables contributing to communicative stress along with factors falling under the categories of code complexity and cognitive complexity may affect the use of attentional resources of task performers and influence their linguistic output. In his framework for investigating second language task performance, he regards task conditions as factors that relate to implementation of the task such as pre-task variables (e.g., planning, repetition), during the task variables (e.g., time pressure), and post-task variables (e.g., post-task activities).

A number of scholars have studied the effect of task condition on language learners’ performance. Ong and Zhang (2010) investigated the impact of providing more planning time on L2 learners’ writing performance. They concluded that more planning time brought significant gains with respect to fluency and lexical complexity in the written output. Similar findings were reported by Ellis and Yuan (2004). They found that increasing the time for pre-task planning promoted the fluency and accuracy of the texts composed by the L2 learners. Regarding the effect of task structure and time perspective, Wang and Skehan (2014) state that completing a task involving speaking about past events without having accompanying supporting input is more demanding than doing a task that involves speaking about the present events supported by input, but the latter condition is more open to negotiation than the first one. In addition, some studies (Storch, 2016; Wigglesworth & Storch, 2009) explored the effect of collaboration on completing the task, and they confirmed the beneficial impact of this task condition on learners’ writing performance.

In the present study, task condition was manipulated by asking learners to complete the writing task either individually or collaboratively. During collaborative writing, purposeful interaction happens at every stage of composition process and students can benefit from the advantages of collaborative learning accentuated in Swain’s (1985) Output Hypothesis and Donato’s (1994) collective scaffolding. Producing a text collaboratively provides opportunities for learners to share their knowledge, test their hypotheses, discuss the unclear points, and achieve a consensus (Swain & Lapkin, 1995). Most studies conducted on this topic have confirmed the beneficial impact of collaborative writing on accuracy and complexity measures (DiCamilla & Anton, 1997; Meshkat & Ghasemi, 2021; Sang & Zou, 2023; Storch & Wigglesworth, 2007). Bueno-Alastuey et al. (2022) reported that treatment sessions involving collaborative writing and providing CF improved the complexity and accuracy of written output of students learning Spanish as a foreign language.

Interaction of task complexity and CF

Although SLA researchers have divergent ideas about definition of task and sequencing the tasks, most of them concur that in addition to requiring L2 learners to engage in meaning-making activities, tasks should include a Focus on Form (FonF) element (Robinson, 2007). Focus on form has been emphasized in TBLT research because as VanPatten (1990) argues when L2 learners are performing a task, their limited attentional resources are primarily directed at elements that carry message meaning. Therefore, formal features of language are usually neglected. Skehan and Foster (1999) confirm this stance by asserting that when L2 learners face the constraints of memory capacities, they prioritize ideational concerns at the expense of linguistic accuracy and may even disregard textual concerns. CF is one instructional procedure that draws learners’ attention to the linguistic forms.

Regarding the interactive effect of task complexity and CF, Skehan’s (1998) Tradeoff Hypothesis postulates that performing cognitively challenging tasks places cognitive load on L2 learners, leading them to allocate more attention to meaning and less attention to form and to the feedback provided for them. Conversely, believing in multiple resource model of attention, Robinson (2001) theorizes that when learners perform complex tasks, they can expand their attentional capacity to cater to the higher demands of the task. Hypothesizing that conveying more complex ideas requires more complex syntactic resources, Robinson reasons that performing complex tasks leads learners to employ specific forms to meet the demands of the task, and this situation guides them to be more receptive of the CF.

In recent decades, exploring the role of task features in language learners’ performance has been the focus of many SLA studies but examining the interaction of task complexity with CF has been limited to a few studies looking into the issue in oral modality (Révész & Han, 2006). Nuevo (2006) studied the effect of giving CF in learners’ performing simple and complex oral tasks. She found more L2 improvement for learners who did simple version of the task and received recasts on their performance. Révész (2009) studied the combined effect of task complexity and recasts on L2 morphosyntactic development. Oral and written picture description tasks were utilized to assess the impact of the treatment sessions on EFL learners’ progress in the use of the target form. The results indicated that complex tasks plus recasts were more beneficial than simple tasks combined with recasts in promoting the morphosyntactiac development in L2 learners. Révész et al. (2014) explored the effects of task complexity, input frequency, and feedback on L2 learners’ acquisition of past counterfactual constructions. Learners who received recasts on simple tasks (− reasoning) while having balanced input outperformed those who received feedback on complex tasks (+ reasoning) while having skewed input. Kim (2012) examined the impact of task complexity on the interaction-driven learning opportunities and question development in EFL context. Her findings showed that complex tasks promoted the exchange of CF in pairs and facilitated learners’ L2 development.

Interaction of CF and collaborative writing

Studies that have examined the effect of CF on L2 learners’ writing performance (e.g., Hyland & Hyland, 2006; Sheen, 2007) verify the beneficial effects of this instructional procedure. Nevertheless, the interaction between CF and condition under which writing task is performed has remained an unresolved issue. The effect of CF on individual and collaborative writing can be explained by the arguments of Vygotsky’s (1978) sociocultural theory of learning, which describes learning as a social process and the origination of human intelligence in society or culture. The major theme of Vygotsky’s theoretical framework is that social interaction plays a fundamental role in the development of cognition. Ohta (2005) reasons that CF provided either by a peer or by the teacher constitutes a form of interaction and scaffolding emphasized in Vygotskian theory, and it is in fact applying the zone of proximal development (ZPD) to language learning. The beneficial influence of feedback on performance of learners who carry out the writing task collaboratively can also be justified by theoretical models of L2 writing processes, including by Flower and Hayes’s (1980) model of writing.

Flower and Hayes’s (1980) seminal model of writing delineates the mental processes through which writers generate texts. In this framework, writing process is divided into three subprocesses: planning, translating, and reviewing. Planning includes subprocesses of generating ideas, organizing those ideas logically in one’s mind, and goal setting perhaps by modifying the generating and organizing activities. Translating means transforming the conceptual plan for the document to a text. In reviewing stage, the text produced so far is read, possible errors are corrected, and modifications are made to improve the quality of the text. Monitoring is the process which appears to dominate the other three writing processes. It coordinates and organizes planning, translation, and reviewing. As some researchers (Ede & Lunsford, 1990; Storch, 2016) describe interaction at all stages of writing from decision making to constructing and editing the text enhances the writing accuracy of learners and helps them sharpen their thoughts by extracting ideas through knowledge sharing.

Not many studies have examined how writing performance of L2 learners can be impacted by individually or collaboratively processing the CF. Kassim and Luan (2014) reported beneficial effects for collaborative processing of feedback on learners’ accuracy in using specific grammatical forms. The findings of their study revealed that collaboratively discussing the feedback with a partner made learners deliberate over the CF for a longer time and this raised their awareness of gaps that existed in their interlanguage; as a result, their accuracy in using selected grammatical forms improved. Mujtaba et al. (2021) conducted a study to compare the effects of individual and collaborative processing of WCF on participants’ writing accuracy and revision behavior. Participants in collaborative group made fewer grammatical and lexical choice errors and outperformed the other group in terms of resolved, unresolved, and abandoned errors. Carr and Weinmann (2018) concluded from their research that merely providing direct WCF failed to generate a significant effect in participants’ linguistic knowledge; therefore, they emphasized on the need for learners to collaborate during text generating and processing of WCF because they could co-construct their ZPD through such interaction.

Overall, the majority of past studies examining the effect of task design variables on L2 learners’ performance focused on oral modality of language production (See, Gilabert, 2007; Lee, 2002; Nuevo, 2006), and few of them investigated the influence of task features on learners’ written performance. Additionally, CF and its interplay with task variables have not received ample consideration in the studies mentioned to in this article so far. The current study attempts to fill these gaps by examining the effect of task complexity and task condition alongside CF on Afghan EFL learners’ writing complexity and accuracy.

Accordingly, the following research questions are formulated:

1. What is the combined effect of task complexity, task condition, and WCF on the accuracy measure of EFL learners’ writing performance?

2. What is the combined effect of task complexity, task condition, and WCF on the complexity measure of EFL learners’ writing performance?

Method

Design

The research method established to examine the joint effects of task complexity, task condition, and WCF on EFL learners’ writing complexity and accuracy was a pretest-posttest-delayed posttest design. Task complexity and task condition, each with two levels, together with WCF were the independent variables of the study, and EFL learners’ improvement in writing accuracy and complexity from the pretest to the posttest to the delayed posttest was the dependent variable of the research. Participants were divided into five groups: four experimental groups and one control group. Participants placed in experimental groups performed the simple or the complex version of the writing task either individually or collaboratively and received WCF on their performance, but the participants in the control group did not receive feedback on the complexity and accuracy level of the texts they produced.

Participants

Initially, six intact classes comprising of 247 Afghan undergraduate students who were taking a 12-week academic writing course in a language institute affiliated with Kateb University were invited to participate in the research study. The classes were held two times each week in the afternoon, and Longman Academic Writing Series (book 4), a coursebook designed for intermediate level EFL learners, was used as the main teaching material for the program. The data collection sessions which were held separately from the regular classroom sessions had been planned to be conducted in 10 sessions spread over a period of 9 weeks (see Table 1 for more information about the timing of the data collection process). They were to take place after the regular classroom hours, and the learners who participated in all 10 sessions received bonus scores for the final exam of the semester. At the outset of the research, the participants were given Nelson English Proficiency Test and the writing pretest. The learners whose scores in one test or both were more than one SD below or above the mean scores of all learners were excluded from the study. This measure allowed us to ascertain that selected participants were homogeneous in terms of English proficiency and writing competency. In addition, five other learners were excluded from the research process because they did not attend all treatment sessions or did not take the pretest or posttests. Therefore, a total number of 223 students (82 male and 141 female) participated in this research study. Their age ranged from 17 to 35 years old (M = 23.47, SD = 2.61), their native language was Persian, and none of them had lived or studied in an English speaking country.

Table 1. Summary of the activities conducted during data collection sessions.

Week	Day	Activities
1	Saturday	Learners received proficiency test
2	Saturday	Learners took writing pretest
Week	Day	Activities
		Experimental groups	Control group
3	Saturday	1st treatment session	Following regular syllabus
3	Wednesday	1st feedback processing session	Following regular syllabus
4	Saturday	2nd treatment session	Following regular syllabus
4	Wednesday	2nd feedback processing session	Following regular syllabus
5	Saturday	3rd treatment session	Following regular syllabus
5	Wednesday	3^rd feedback processing session	Following regular syllabus
6	Wednesday	Posttest	Posttest
9	Wednesday	Delayed posttest	Delayed posttest

Assessment tasks

Assessment tasks evaluated participants’ gain from treatment sessions and included three writing tests administered to the participants as the pretest, posttest, and delayed posttest respectively. In each task, participants were given information about a product (e.g., a laptop computer) produced by seven different companies (A, B, C, etc.). The prompt of the writing task asked participants to choose one of the seven products for a hypothetical friend named Sara. They had to read the information in a table (see Appendix for sample tasks) and to decide which laptop satisfied Sara’s six criteria (price, weight, storage capacity, screen resolution, dimensions, and battery life). None of the laptops met all of Sara’s criteria. Yet, the participants had to choose the most suitable product based on Sara’s preferences, to discuss their selection, and to defend and justify it. They had 50 min to perform the task and to include at least 200 words in their argumentative texts. Pretest and posttest tasks differed in terms of topic, but they were comparable in respect of cognitive complexity.

Assessment tasks were medium in terms of complexity because participants considered six criteria to complete the task, but the number of criteria in simple and complex treatment tasks were four and eight, respectively. Determining complexity based on this criterion was in line with Halford et al. (2007) who asserted that our memory can process four items at a time and above this level processing becomes demanding for learners. The complexity of treatment and assessment tasks as well as their validity were checked and confirmed during a pilot study conducted before the main stage of the study.

Treatment tasks

Treatment tasks included writing tasks that participants in the four experimental groups completed during the three treatment sessions. They were similar to assessment tasks in terms of structure and instructions given, but their level of complexity was manipulated by changing the number of criteria that participants had to consider to perform the task. In the simple task, participants were asked to recommend a product (e.g., a laptop) to Sara (their hypothetical friend) considering four criteria, but in the complex task, they had to perform the same task considering eight criteria. Additionally, in simple task, participants were to choose an item from among the products made by five different companies, but in the complex task, they had to select it from products offered by seven different companies. The information about the different products appeared in tables. When participants completed the writing tasks, they received WCF on their performance. As stated earlier, the complexity level and validity of the treatment tasks were evaluated and assured during a pilot study.

Procedure

During the first and the second sessions of data collection procedure, all initial participants received Nelson English Proficiency Test and the writing pretest, respectively. The selected participants (N = 223) who met the proficiency criteria were randomly assigned into one of the five groups: four experimental groups and a control group. The treatment tasks that participants in the experimental groups received varied in terms of (a) cognitive complexity (simple/complex) and (b) task implementation condition (individually/collaboratively). Consequently, the four groups received simple individual (SI), complex individual (CI), simple collaborative (SC), and complex collaborative (CC) treatment tasks respectively. All participants in these groups received WCF on their L2 writing. Participants placed in the control group received regular classroom instruction, did pretest and posttests, but did not receive WCF on their writing.

In Saturday of weeks 3 to 5, participants in experimental groups received three treatment sessions. During each session, participants were given 50 min to produce an argumentative text in response to the specific prompt given to them. The topic of the prompt was identical for all groups, but they differed in terms of level of complexity and task condition. The teacher of the course read participants’ written outputs and gave them indirect unfocused feedback on their performance. It means that the teacher underlined the erroneous structures, but he did not provide the correct forms for the errors. Moreover, this type of feedback is targeted at all grammatical and lexical structures, and not at a few pre-selected constructions. In Wednesday of the three mentioned weeks, participants in the experimental groups received three feedback processing sessions. At each session, the instructor of the course returned the texts to the participants, asked them to pay particular attention to the underlined sections and to provide their correct forms within 20 min. Participants in the individual groups processed the WCF in isolation, but those in collaborative groups could discuss the errors with their partners and reach an agreement over the correct forms.

In Saturday and Wednesday classes of these three weeks participants in the control group followed the syllabus planned for their regular writing course. At the beginning of the Saturday class, which lasted for 50 min, the teacher delivered one or more lessons from the book, and then asked the learners to complete the practices included in the coursebook. At the second half of the session, learners were assigned a topic such as “What are the characteristics of a good laptop?,” and they were required to produce a short text as their response. The writing task assigned to all learners was similar in terms of topic and level of complexity and it was required to be performed individually. No type of feedback or score was given to the control group for completing this task. In Wednesday sessions, which lasted for 20 min, the instructor delivered some other lessons from the coursebook.

A week later, the participants in experimental and control groups took the posttest. Their written productions were coded and analyzed in terms of accuracy and complexity measures by the researcher of the study. Three weeks later, delayed posttest was administered, and participants’ written outputs were coded and analyzed to discover if possible effects of the treatments had been retained. It should be noted that both posttests were carried out individually by the participants. Table 1 provides a summary of the activities performed during each session of the data collection process.

Data coding

Consulting the recommendations of Ellis and Yuan (2004) and following the measures employed in some previous studies (e.g., Ruiz-Funes, 2014; Skehan & Foster, 1999), accuracy was measured by calculating the proportion of error-free clauses to all clauses. To code error-free clauses, each text was divided into clauses, and errors in lexis, morphology, and syntax were marked. Errors pertaining to spelling, punctuation, and capitalization which did not alter the meaning of the clauses were ignored. The syntactic complexity measure chosen was mean number of clauses per T-unit. Following Young (1995), single clauses, main clauses with their subordinate clauses, two or more phrases in apposition, and fragments of clauses produced by ellipsis were all considered T-units, but coordinate clauses were considered independent T-units. Participants’ texts were coded and scored by the author of the study who has an MA degree in applied linguistics and has taught English in different language centers for about 10 years. An independent expert colleague was recruited to code and score 15% of the texts written by the participants. To code the texts, he was told to identify the phrases, clauses, T-units, and sentences in the texts produced by the participants. And to rate the texts, he was requested to calculate the accuracy and structural complexity of written outputs based on the process that the author of the study had followed for coding and rating the texts. This procedure was done to ensure the reliability of coding and scoring mechanism. Inter-coder and inter-rater reliability coefficients were .91 and .93, respectively.

Data analysis

The collected data were analyzed using SPSS 25. Results of normality tests, Kolmogorov–Smirnov statistic and assessment of Histogram, Normal Q-Q, and box plots confirmed that the scores were normally distributed. Consequently, parametric tests were employed to analyze the data. To answer the research questions, first, the descriptive statistics for each group’s written performance in pretest and posttests were calculated. Since ANOVA was to be utilized for data analysis, assumptions underlying this statistical test including normality and homogeneity of variance were checked, and no serious violations were detected. To ascertain that writing accuracy and complexity of the groups were at similar level at the outset of the study, two one-way ANOVAs were run and the five groups were compared in respect of these two measures. Next, two other separate one-way ANOVAs were performed to compare the accuracy and complexity scores of the five groups among pretest, posttest, and delayed posttest. Post-hoc comparisons were also run to discover which pairs of groups significantly differed from each other. Moreover, to assess the interaction of task complexity, task condition, and WCF on participants’ written productions, two separate two-way ANOVAs were run for the posttests. The significance level was set at .05, and to interpret the effect size in the analyses, Cohen’s (1988) suggestions were followed. As he recommends, η2 = .01 is small, η2 = .06 is medium, and η2 = .138 is large effect sizes.

Results

The effect of task variables on writing accuracy

To answer the first research question, descriptive statistics (mean and standard deviation) for accuracy of the produced texts were calculated. As Table 2 indicates, participants who performed the simple task in pairs and received WCF obtained the highest mean score in the posttests, but participants who were placed in the control group and did not receive feedback on their writing had the worst performance in this measure.

Table 2. Descriptive statistics for the accuracy variable in the pretest and posttests.

Group	n	Test	M	SD
Simple individual	45	Pretest	.53	.09
		Posttest	.69	.10
		Delayed posttest	.66	.13
Complex individual	45	Pretest	.52	.06
		Posttest	.62	.07
		Delayed posttest	.62	.10
Simple collaborative	44	Pretest	.50	.08
		Posttest	.81	.08
		Delayed posttest	.78	.13
Complex collaborative	44	Pretest	.51	.09
		Posttest	.76	.10
		Delayed posttest	.73	.17
Control	45	Pretest	.51	.09
		Posttest	.58	.08
		Delayed posttest	.56	.16

Having checked the assumptions underlying ANOVA, a one-way ANOVA was performed to detect if the five groups were comparable in terms of accuracy at the beginning of the study. Results of the analysis did not reveal any significant difference between the groups in the pretest, F (4, 218) = .15, p = .94. Thus, any probable gains in accuracy in the posttests could be ascribed to receiving treatment sessions. Two other one-way ANOVAs were run to compare the writing accuracy of the groups in the posttests. Results revealed a statistically significant difference between the groups in posttest, F (4, 218) = 14.21, p = .001 and also in delayed posttest, F (4, 218) = 13.06, p = .001.

So as to find out which groups were significantly different from each other in the posttest, the Scheffe test was conducted. According to the results, the writing accuracy of the complex individual group (M = .62, SD = .07) was significantly less than that of the simple collaborative (M = .81, SD = .08) and complex collaborative (M = .76, SD = .10) groups. Also, the accuracy of simple collaborative group was significantly higher than the accuracy of the control group (M = .58, SD = .08). The participants in complex collaborative group (M = .76, SD = .10) produced texts that were significantly more accurate than those produced by the control group. Moreover, the mean scores of accuracy for the simple individual (M = .69, SD = .10) and complex individual groups were significantly less than that of simple collaborative group but significantly higher than that of the control group. This finding indicates that participants who did the simple task individually and received WCF showed more gains in the accuracy of their written output than the participants in the control group; nevertheless, they showed less gain in accuracy in comparison with the participants who performed the simple version collaboratively and received the same type of WCF. The other pairwise comparisons were not statistically significant and hence did not indicate differences between the groups.

Another Scheffe test was run to compare the writing accuracy of the groups in delayed posttest. The results indicated that participants in complex individual group (M = .62, SD = .10) wrote significantly less accurate texts than the simple collaborative (M = .78, SD = .13) and complex collaborative (M = .73, SD = .17) groups. This finding shows that participants who did the complex task individually and received WCF made less gain in accuracy compared with participants who did the simple or complex versions of the task collaboratively. The analysis also revealed that the accuracy of the simple individual group (M = .66, SD = .13) was significantly less than that of simple collaborative group, but significantly higher than the control group (M = .56, SD = .16). Participants in simple individual group made more gains in accuracy than the participants in the control group, yet they made less gain than participants who completed the same version of the task jointly. Furthermore, the difference between the simple collaborative and control groups together with the difference between complex collaborative and control groups were statistically significant, demonstrating that those who completed the treatment task collaboratively outperformed those who did the task individually and did not receive WCF. There was no statistically significant difference between the other pairs of groups.

In addition, a two-way ANOVA was conducted to discover if the influence of WCF in participants’ writing accuracy was mediated by task complexity, task condition, and their interaction. Results of analysis for posttest indicated there was a significant main effect for task complexity, F (1,174) = 4.71, p = .03, η2 = .05, and also for task condition, F (1, 174) = 36.24, p = .00, η2 = .14. Thus, the effect sizes for task complexity and task condition were medium and large respectively. On the other hand, the interaction effect between task complexity and task condition did not reach statistical significance, F (1, 174) = .25, p = .54, η2 = .001 and the effect size for interaction was small. Results of two-way ANOVA for delayed posttest provided a significant main effect for task condition, F (1, 174) = 31.32, p = .01, η2 = .16. But, the main effect for task complexity did not yield statistical significance, F (1, 174) = 1.98, p = .13, η2 = .02. Like posttest, again the interaction between task complexity and task condition did not produce statistically significant difference, F (1, 174) = .02, p = .89, η2 = .00. These findings suggest that task condition mediated the positive effect of receiving WCF on producing more accurate texts in both posttests. Contrarily, although task complexity mediated the beneficial effect of WCF in the posttest, its effect size in the delayed posttest was small and non-significant.

The effect of task variables on writing complexity

To answer the second research question, first, descriptive statistics, including mean and standard deviation for the syntactic complexity of the texts in the pretest and posttests were calculated. Table 3 provides these statistics.

Table 3. Descriptive statistics for the syntactic complexity variable in the pretest and posttests.

Group	n	Test	M	SD
Simple individual	45	Pretest	1.27	.14
		Posttest	1.25	.16
		Delayed posttest	1.25	.17
Complex individual	45	Pretest	1.29	.15
		Posttest	1.29	.13
		Delayed posttest	1.27	.16
Simple collaborative	44	Pretest	1.25	.10
		Posttest	1.26	.10
		Delayed posttest	1.28	.13
Complex collaborative	44	Pretest	1.32	.12
		Posttest	1.35	.11
		Delayed posttest	1.39	.10
Control	45	Pretest	1.25	.14
		Posttest	1.20	.09
		Delayed posttest	1.22	.12

A one-way ANOVA was performed to compare the syntactic complexity of the five groups’ written productions in the pretest. It was found out that the groups were comparable in this measure at the pretest, F (4, 218) = 1.41, p = .18. But, participants’ writing level changed in the posttest and the difference between the groups reached statistical significance, F (4, 218) = 6.01, p = .002. The results of ANOVA conducted for delayed posttest also revealed a significant difference between the groups, F (4, 218) = 5.95, p = .003. As Table 3 shows, participants in complex collaborative group obtained the highest syntactic complexity score in both posttests.

Then, Scheffe test was conducted to compare the mean values of complexity among different pairs of groups in posttest. Results revealed a significant difference between the simple individual (M = 1.25, SD = .16) and complex collaborative (M = 1.35, SD = .11) groups. Moreover, the difference between complex collaborative group and control group (M = 1.20, SD = .09) together with that between simple collaborative (M = 1.26, SD = .10) and control group were significant as well. These findings suggest that participants who carried out either the simple or the complex treatment task in pairs and received WCF on their performance wrote more structurally complex texts than those participants in the simple individual and control groups. In addition, the difference between simple individual group and control group and that between complex individual and control group reached statistical significance. The other two-by-two comparisons did not indicate statistically significant difference. The results of post-hoc comparison using Scheffe method for delayed posttest indicated a significant difference between complex collaborative (M = 1.39, SD = .10) and simple individual (M = 1.25, SD = .17) groups as well as between complex collaborative and control group (M = 1.22, SD = .12) groups. Besides, the difference between simple collaborative group (M = 1.28, SD = .13) and the control group together with that between complex individual (M = 1.27, SD = .16) and the control group indicated statistical significance. Thus, participants who completed the complex treatment task collaboratively and received feedback on their performance produced more syntactically complex texts than those who did the complex version of the task by themselves and also than those who did not receive any treatment tasks. The difference between other pairs of participants was not statistically significant.

Additionally, a two-way ANOVA was conducted to assess the impact of task complexity, task condition, and their combined effect on the efficacy of WCF in influencing the syntactic complexity of the texts. Analysis of data for posttest produced a significant main effect for task complexity, F (1, 174) = 10.46, p = .00, η2 = .136 as well as for task condition, F (1, 174) = 4.89, p = .01, η2 = .05. As the values of eta squared indicate the effect sizes for both of these variables were medium. However, a significant effect was not found for the interaction between task complexity and task condition, F (1, 174) = .37, p = .63, η2 = .005. Similar results were obtained from the two-way ANOVA run for delayed posttest. There was a significant main effect for task complexity, F (1, 174) = 9.14, p = .006, η2 = .07. Likewise, a significant main effect was detected for task condition, F (1, 174) = 11.41, p = .005, η2 = .10. Contrarily, the interaction between task complexity and task condition was not significant, F (1, 174) = 1.15, p = .24, η2 = .01. These results denote that the positive effect of receiving WCF on syntactic complexity of writing was mediated by the complexity level of treatment task and by the conditions under which the task was performed.

Discussion

The first research question addressed the effect of task complexity and task condition together with WCF on participants’ writing accuracy. Findings indicated that task complexity affected the accuracy of the texts in posttest but this positive effect was not retained until delayed posttest. Moreover, it was found out that the main effect of task condition on accuracy was significant in both posttests. These results clearly indicate that providing WCF for the participants prompted them to produce more accurate texts and that this positive effect was mediated by the condition under which the writing task was carried out. Therefore, doing simple version of the task collaboratively alongside receiving WCF resulted in more improvement in accuracy measure for L2 learners.

The second research question aimed to explore if task complexity, task condition, and their interaction mediated the effect of WCF on writing syntactic complexity. Statistical analyses produced significant main effects for task complexity and task condition in both posttests but not for their interaction. Participants who did complex version of treatment task and received WCF on their written output outperformed the learners in simple groups in terms of producing structurally complex texts. Furthermore, participants who performed the simple and complex versions of the task collaboratively wrote more complex texts than their counterparts who did the tasks with the same level of complexity (simple/complex) individually. These findings suggest that both task complexity and task condition moderated the positive and beneficial effects of WCF on producing structurally complex texts.

These results are similar to those of Golparvar and Rashidi (2021) who reported an inverse relationship between task complexity and writing accuracy and to those of Rahimi and Zhang (2017) who stated that raising task complexity affected accuracy negatively. Nevertheless, our findings are not in line with those of Kim (2022) who found that L2 learners receiving CF in complex tasks wrote more accurate texts. Révész and Han (2006) reported similar results for oral production of language learners. However, their study revealed that the learners who performed simple version of the task experienced the benefits of receiving recasts in notes-primed condition. Considering the similarities between notes-primed task and the writing task utilized in the present study, the findings of the two studies seem similar in this case. Our findings regarding the positive impact of collaboratively processing the WCF on improving accuracy is consistent with Mujtaba et al. (2021) who reported that collaborative processing group made more gains in terms of accuracy compared with individual processing group.

Although the findings of the study do not provide strong support for Skehan’s Tradeoff Hypothesis or Robinson’s Cognition Hypothesis, it seems that they partially back up the first theory rather than the latter one. As explained previously, Cognition Hypothesis postulates that increasing task complexity along resource-directing variables (e.g. +/- number of elements) causes a change in L2 learners’ interlanguage, with the consequence being that learners stretch their linguistic repertoire and they increase both accuracy and complexity level of their L2 production. If this hypothesis held true in relation to the findings of the present study, a simultaneous and considerable increase in accuracy and complexity measures would have been observed in the complex groups (complex individual and complex collaborative), but such a similar advancement in terms of the two measures would not have been witnessed in the simple groups (simple individual and simple collaborative). The results of this study indicated that in both posttests, it was the simple collaborative group which received the highest mean score in the accuracy measure. Besides, each of the two simple groups made more gains in this measure compared with its counterpart complex group. Conversely, when it comes to investigating the grammatical complexity measure, it can be seen that in the posttest the two complex groups received the highest mean scores in complexity. Additionally, in the delayed posttest, each of the complex groups made more gains compared with its counterpart simple group. Therefore, while giving complex version of the task to some participants led them to compose texts that were both accurate and complex, it did not result in their producing the texts that were both the most accurate and the most complex ones. In other words, participants in the complex groups exhibited their optimal performance in producing more complex texts but not in producing more structurally complex written outputs. These results reject the claim of the Cognition Hypothesis, according to which engaging in complex task provides an advantage for L2 learners and leads them to an optimal performance simultaneously in the areas of accuracy and complexity.

Since after receiving the treatment sessions, participants in the complex groups did not perform optimally in both of accuracy and complexity measures, it seems that they could not attend to both areas at the same time. The reason for this lack of optimal dual improvement is that a trade-off or competition took place between these two aspects of language performance, with the consequence being that L2 learners assigned to the complex groups allocated more attention to complexifying their output and less attention to producing more accurate texts. This conclusion is consistent with the tenets of the Tradeoff Hypothesis (1996, 1998) which posits that when performing a complex task, L2 learners cannot attend to both accuracy and complexity, because attending to one area of performance takes attention away from the other dimension. Another finding of the study that further supports the Tradeoff Hypothesis is that nearly all participants made significant gains in respect of accuracy, but either they did not make any gains in respect of complexity or their advancement in this domain was not as noticeable as their gains in the accuracy measure. A reasonable explanation for these findings can be that learners’ interlanguage is a system consisting of different subsystems that may interact in negative ways and in competition with one another. When an L2 learner is performing a task, all performance areas (complexity, accuracy, and fluency) require attention and working memory involvement at the same time, but the system can commit the limited attentional resources towards some but not to all subsystems at the same time. The ultimate result of this condition is that L2 learners cannot produce oral or written output that reflects the optimal consideration of both accuracy and complexity.

Another prediction of the Tradeoff Hypothesis is that when L2 learners are engaged in the complex task, and they divide their attention between processing form and meaning, they cannot allocate sufficient attention to the provided CF. Our findings partially supported this claim because even though the participants in all groups (including complex groups) benefitted from WCF and improved the accuracy of their writing from the pretest to the posttests, the amount of this gain for each of the complex groups was less than that for each of the corresponding simple groups. It might be the case that since participants placed in the simple groups were not faced with the high level of cognitive load, and therefore they were not subject to the constraints of the attentional resources, they could absorb and internalize a greater amount of the WCF provided throughout the treatment sessions and use them to produce more accurate texts in the posttests. On the other hand, even though participants in the complex groups did focus part of their attention on improving writing accuracy, high level of cognitive load imposed by the tasks deterred them from obtaining the same amount of feedback as obtained by the simple groups. If their gain from the provided WCF had been at its optimal level, each of the complex groups would have received a higher score in the accuracy measure than its counterpart simple group. All in all, these results partially confirm the claims of the Tradeoff Hypothesis, according to which more complex tasks allow less attention to the meaning and by implication to the provided feedback. Emphasizing the limited attentional capacity of L2 learners, Skehan (1998) reasons that learners performing complex tasks have to divide their attention between the writing task prompted by many elements and the points they have received through feedback. So, they prioritize conveying meaning at the expense of form and they don’t benefit considerably from provided WCF.

As previously stated, increasing the cognitive load of tasks prompted participants in the complex groups to produce more structurally complex texts than participants in the simple groups. A number of previous studies (Abdi Tabari et al., 2023; Sasayama, 2011) reported that tasks requiring learners to process many elements push them to distinguish and compare all different elements, encouraging them to exhibit a higher level of structural complexity and lexical diversity in their texts. The positive effect of task complexity on writing syntactic complexity can also be justified by reference to Ishikawa’s (2007) argument which postulates that doing complex tasks entails a greater degree of information packaging and deeper semantic processing, resulting in greater use of subordination and embedded clauses. In current study, although the collaborative groups made improvement in terms of producing structurally complex texts from the pretest to posttests, the individual groups experienced a decrease in respect of this measure. This lack of improvement in complexity measure can be explained by Truscott’s (1996) claim that learners tend to avoid the categories that have been the subject of CF. In present research, participants received indirect unfocused CF. Probably, they were not able to produce the correct forms for erroneous structures during feedback processing sessions and thereby decided to produce shorter and less complex sentences in the posttests. Further research is needed to discover if the provision of explicit and direct CF might cause L2 learners to exhibit different performance in terms of producing accurate and structurally complex written outputs.

Better performance of simple groups on writing more accurate texts can also be accounted for by reference to the L2 writing models. One such model was presented by Roca de Larios et al. (2001). Taking into account different stages of composing process (planning, translating, and reviewing), they argue that when L2 learners receive a simple writing task, there is less burden on the central executive of their working memory, because they spend less time for retrieving and developing ideas and combining them into propositions. Besides, they perform the steps in planning stage more easily, and translate their ideas into sequences of lexically and syntactically organized linguistic units with more success. Therefore, they have free attentional resources in the review stage to process the CF they have received, to revise their texts, and to produce more accurate linguistic outputs.

The better performance of the collaborative groups in producing accurate texts substantiates Vygotsky’s (1978) sociocultural theory of language acquisition and Long’s (1996) Interaction Hypothesis, which underscore the role of social interaction and collaboration in L2 learning. In current study, the two learners placed in each of the collaborative groups could work jointly to process the feedback they had received from the teacher of the course. Besides, each member of the pair could receive WCF about his written performance from the other member. Interactional procedures that occurred in pairs pushed learners to do the feedback processing more profoundly and to modify their output in more targetlike ways (Nuevo, 2006). Donato (1994) is of the view that asking L2 learners to perform the writing task individually and providing them with indirect WCF (as with the present study) could not generate a significant shift toward acquiring correct linguistic forms within their ZPD, with the consequence being that they had little gain from WCF. It can be safely concluded that L2 writing tasks, because of their cognitively laborious nature and imposing constraints on writer’s composing capacity, make this group of L2 learners focus on meaning and neglect form as well as feedback given over writing accuracy. On the other hand, L2 learners who process the WCF in pairs experience less pressure on their attentional capacity, and this advantage encourages them to convey their intended meaning with more accuracy.

The results regarding the positive influence of collaboration on writing more structurally complex texts also add support to Vygotsky’s (1978) social cognitive theory of language learning, Swain’s (1985) Output Hypothesis, and Skehan’s (1998) Tradeoff Hypothesis. Considering the issue from a cognitive perspective, it can be argued that during collaborative writing, learners have the opportunity to combine efforts in different stages of composition, especially at planning and translating stages. This cooperation mitigates the cognitive load on individual learners and leaves them with more attentional resources that can be exploited for producing syntactically complex sentences and deliberating on teacher’s feedback and getting immediate feedback from their partners. Conversely, participants in individual groups did not have the chance to benefit from pooling of linguistic knowledge that happens in coauthoring a text. They had to make full use of their attentional resources at every stage of generating a text and so they did not gain considerably from treatment sessions to enhance the structural complexity of their writing. Results revealed that each of the variables—task complexity, task condition, and WCF—contributed to L2 writing complexity and accuracy in their own unique ways. Learners assigned to complex individual group performed a cognitively demanding task by themselves and then processed WCF provided by the teacher. Completing the task under this condition exerts significant pressure on learners’ limited attentional capacity, leaving them with less free mental resources for correcting erroneous structures. This finding contradicts Robinson’s (2001) prediction that increasing complexity along resource-directing variables leads L2 learners to expand their attentional capacity and produce more accurate and more complex language. Thus, writing the text jointly helps learners to bypass the limitations of the working memory.

Conclusion

The present study analyzed the interactive effects of task complexity, task condition, and WCF on language learners’ writing complexity and accuracy. Altogether, the results provided theoretical support for Skehan’s (1996) Tradeoff Hypothesis and Vygotsky’s (1978) Sociocultural Theory. Specifically, statistical analyses revealed that giving simple writing task to participants and asking them to complete it in pairs helped them benefit from WCF and produce more accurate texts. Participants in simple collaborative group experienced a less degree of cognitive pressure; moreover, they could join forces with another partner during different stages of composition (planning, translating, and reviewing). The combination of these two factors left learners with more attentional resources that could be allocated for processing WCF and improving linguistic knowledge. A similar relationship was found between complex tasks and collaborative condition, in that performing complex task collaboratively along with receiving feedback improved the structural complexity in texts produced by participants.

Considering the scarcity of studies scrutinizing the interplay between task variables and CF, our findings can have significant implications for theoreticians, language instructors, and syllabus designers. If writing instructors wish to improve learners’ writing accuracy, they can ask their students to perform simple writing tasks collaboratively, and then give them CF. But, if they want the students to produce more structurally complex texts, they can assign them complex tasks to be performed collaboratively and then provide feedback for them. In sum, if they want to see a balanced improvement in different dimensions of students’ writing performance, they should carefully consider the complexity level of tasks and then decide how (individually/collaboratively) they should be completed. Furthermore, SLA researchers can find the results of this study insightful in that they provide new pieces of information about the role of task complexity, task condition, CF, and their interaction in language development. Thus, they can use the methodology used in this article and improve upon it to carry out replication studies or conduct new studies exploring the role of other task design variables on EFL learners’ oral or written performance.

A number of limitations affected the findings of our study. In the first place, the time interval between pretest and posttests was not long enough. It would have been more desirable if the delayed posttest had been given to the participants at least 2 months after the first posttest. But, since the academic writing course in the language institute were to last for a semester (12 weeks), it was not feasible to administer the second posttest at a later time. The second limitation was that a number of learners were not comfortable with collaborative writing, and therefore, they did not cooperate in the composing process as much as it was anticipated. According to Willing (1994), these learners are analytical learners and like to study alone and find their own mistakes. Besides, some socio-cultural contexts (e.g., some Asian educational settings) do not encourage collaborative work (pair and group) in the classroom, but instead they promote individual work. It is vital that future researchers consider this personal characteristic in mind and select learners who are more communicative. The other limitation of the study was that participants were undergraduate university students who were taking academic writing course. So, further research is needed to discover if EFL learners at lower and higher levels of education might exhibit different performance if they received the simple and complex writing tasks given to the participants of the current study.

In current study, complexifying the task was achieved only by increasing the number of elements. Further studies should examine the effect of other task variables (e.g., spatial reasoning/causal reasoning) as well as that of simultaneous manipulation of multiple task complexity factors on EFL learners’ writing complexity and accuracy. Given the growing influence of technology in language classrooms, it is recommended that interested researchers explore the impact of task features and CF in digital platforms and computer-mediated communicative contexts. Using computational tools (e.g., Coh-Metrix) allows researchers to obtain various indices of accuracy and complexity with more ease and within a shorter time. Furthermore, scrutinizing the interaction between task complexity and other task implementation conditions (e.g. changing time limit needed to do the task) is another fruitful and promising area for future studies.

Ethical approval

All subjects gave their informed consent for inclusion before they participated in the study.

Author’s contributions

The author designed the study, collected the data and performed statistical analyses. He prepared the manuscript and approved the final version.

Acknowledgments

We acknowledge all the students who accepted to take part in the study.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available. But they can be demanded from the corresponding author of the study upon reasonable request.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Mohammad Bagheri

Mohammad Bagheri has an MA degree in applied linguistics from University of Tehran, Iran. He has been working as an EFL instructor and translator in recent years. His main research interests include second language acquisition, the use of technology in L2 instruction, and the relationship between culture and language identity.

Appendix A

A simple version of the treatment task

Your friend Sara wants to buy a laptop computer. She intends to buy a laptop which has a reasonable price, large amount of hard drive capacity, a low weight, and a small size. The information about some different laptops appears in the following table. None of the laptops meets all Sara’s criteria. However, you should decide which laptop is the most suitable one for her. You should write an argumentative text and discuss why you have selected a specific laptop for Sara. Please include at least 200 words in your text. You have 50 minutes to complete the task.

Table

Laptop	Price ($)	Hard drive capacity (GB)	Weight (KG)	Dimensions (cm)
A	1000	250	1.27	30.40 × 21.23 × 1.60
B	950	500	1.75	36.02 × 23.39 × 1.78
C	2100	2000	2.69	34.62 × 24.99 × 6.10
D	537	250	1.81	35.56 × 22.86 × 2.54
E	500	64	1.18	28.50 × 20.07 × 1.73

Appendix B

A complex version of the treatment task

Your friend Sara wants to buy a laptop computer. She intends to buy a laptop which has a reasonable price, large amount of hard drive capacity, a low weight, a small size, a large screen, a long battery life, high screen resolution, and is of the latest model. The information about some different laptops appears in the following table. None of the laptops meets all Sara’s criteria. However, you should decide which laptop is the most suitable one for her. You should write an argumentative text and discuss why you have selected a specific laptop for Sara. Please include at least 200 words in your text. You have 50 minutes to complete the task.

Table

Laptop	Price ($)	Hard drive capacity (GB)	Weight (KG)	Dimensions (cm)	Screen dimensions (Inch)	Battery life (hrs.)	Screen resolution (Inch)	Release date
A	1000	250	1.27	30.40 × 21.23 × 1.60	13	8.5	1440 x 900	2020
B	950	500	1.75	36.02 × 23.39 × 1.78	15.6	6.5	1920 × 1080	2022
C	2100	2000	2.69	34.62 × 24.99 × 6.10	14.4	5.5	1366 × 768	2018
D	537	250	1.81	35.56 × 22.86 × 2.54	15.6	6	1920 × 1080	2021
E	500	64	1.18	28.50 × 20.07 × 1.73	11.6	5	1366 × 768	2015
F	1200	500	1.90	35.51 × 22.91 × 1.88	15.6	5	2880 × 1800	2009
G	2000	1000	2.50	35.41 × 26.39 × 2.26	16	9	2560 × 1440	2017