Introduction

The importance of peer interaction is reflected in the definitions of tasks since they all center around communication of meaning (e.g., Long, 1985; Nunan, 1989). There is consensus among researchers that conversational interaction plays an important role in SLA (e.g., Gass, 1997; Long, 1996; Pica, 1994) by providing comprehensible input and negative feedback, which shows learners’ structures that are not acceptable in L2 (Long, 1996). Another significance of interaction is that it creates a scope for learners to produce output which has three main functions as proposed by Swain (1995, 2005): noticing holes in L2 knowledge, testing hypotheses against target language norms, and reflecting on their own and their partners’ output. The latter function occurs through languaging, a process allowing students to verbalize their deliberations over linguistics problems they are faced with (Swain, 2006). This usually takes place in dialogues for making meaning and constructing new knowledge, which are referred to as “collaborative dialogues” (Swain, 2000). These dialogues can promote L2 learning because knowledge jointly constructed through them can then be internalized by L2 learners and changed into individual knowledge (Swain, 2006). Recent research on focus on form (i.e., techniques designed to attract learners’ attention to form in communication, Ellis, 2015) has sought ways to draw learners’ attention to linguistic form, without isolating these from their meaningful context (e.g., Doughty & Williams, 1998). This usually takes place through tasks involving the generation or modification of written output that set the ground for the production of collaborative dialogues (García Mayo, 2007; Swain & Lapkin, 2001). In order to capture instances where learners draw each other’s attention to form, collaborative dialogues are studied through language-related episodes (LREs) known as occurrences where learners talk about the language they produce, question their language use, or correct themselves and others (Swain & Lapkin, 1998).

Researchers working from a sociocultural perspective have shown that learners’ ability to generate and resolve LREs through collaborative dialogues can be affected by task type (e.g., De la Colina & García Mayo, 2007 Mayo, 2007; García Mayo, 2002a, 2002b; Ismail & Samad, 2010; Nassaji & Tian, 2010; Storch & Wigglesworth, 2007; Yanguas & Bergin, 2018; Zabihi, 2022). As far as the effect of task type on focus on form is concerned, the small number of studies investigating task effects has not clarified the link between LREs and task type (Yilmaz & Granena, 2010). This means that while these researchers confirm that different tasks can mediate the frequency of LREs, they have obtained contradictory findings even when they compared the very same tasks (e.g., García Mayo, 2002a, 2002b; Ismail & Samad, 2010; Yilmaz & Granena, 2010). Another problem regarding previous research is that scholars have used a wide range of tasks while investigating their relation with LREs, making comparisons relatively difficult. Thus, there seems to be a need for classifying tasks based on the nature of the tasks under a specific classification where all can be compared and analyzed. We believe that this classification could be based on input orientation which will be elaborated on in the literature.

Literature Review

The Importance of Collaboration in Peer-Peer Interaction

From a psycholinguistic perspective, Long’s interaction hypothesis (1980, 1996) states that learners’ attention may turn to linguistic form when they encounter breakdowns during the communication of meaning. When learners do not succeed in understanding their interlocutor, they usually resort to the negotiation of meaning which helps them gain mutual comprehension. In an effort to achieve such mutual understanding, learners execute a wide range of strategies, such as asking an interlocutor to confirm message content or requesting the interlocutor to explain something further. These interactional moves promote L2 development in the context of peer interaction (McDonough, 2004; for recent reviews, see Philp et al., 2014, and Sato & Ballinger, 2016). The process by which two or more interlocutors identify and then attempt to resolve a communication breakdown is referred to as negotiation of meaning (Ellis, 2003). Negotiation of meaning is believed to enhance the comprehensibility of L2 input, making it more conducive to L2 development (Bygate, 1999; Foster & Ohta, 2005).

From a sociocultural theoretical perspective, primarily building on the work of Vygotsky (1978), it has been argued that learner-learner interactions contribute to language development because through such interactions learners find opportunities for languaging (Swain, 2006). When these verbalizations result in correctly resolved LREs, there is a higher possibility that they retain the co-constructed knowledge for subsequent independent use (e.g., Fernández Dobao, 2014). All the affordances created by peer-to-peer interaction emerge when students establish mutual relations grounded in sociocultural and psycholinguistic concepts. According to sociocultural theories inspired by Vygotsky’s work (1978), learning refers to a process that is socially mediated and that emerges in interaction either with peers or with experts, creating favorable conditions for languaging (Swain et al., 2011). Languaging manifests itself when learners get involved in completing communicative tasks through collaborative dialogues. Various studies (Kobayashi, 2003; Lapkin et al., 2002; Morris & Tarone, 2003; Storch, 1998, 2004; Swain & Lapkin, 1998; Tin, 2003; Zabihi & Bayan, 2020; Zabihi & Ghahramanzadeh, 2022) have demonstrated that student collaboration, frequently referred to as small-group work, is of great significance in second language (L2) acquisition, social aspects of the classroom (i.e., how learners join together), and language pedagogy since working collaboratively enables learners to have a relatively better performance than working alone (e.g., Storch, 1999). One way for benefiting from collaboration is through a specific kind of interaction known as collective scaffolding (Donato, 1994), where learners can pool their cognitive and linguistic knowledge and work together to solve their language-related problems.

Collaborative dialogues are primarily studied through LREs because they serve as overt indications that at a given moment students are focusing on form (grammar, lexicon, mechanics, or discourse; Fortune & Thorp, 2001; Loewen & Basturkmen, 2005), thereby establishing what may be a definitive connection with learning (Swain & Lapkin, 1998).

Task Type and Focus on Form in Peer-Peer Interaction

Inspired by Swain’s output hypothesis (1995, 2005), a growing body of research (e.g., Kowal & Swain, 1994, 1997; Leeser, 2004; Malmqvist, 2005; Williams, 1999, 2001) has used LREs to analyze occurrences where learners direct each other’s attention to form. In doing so, research has documented that the occurrence, resolution, and the type of LREs can be affected by task type (e.g., García Mayo, 2002a, b; Ismail & Samad, 2010; Swain & Lapkin, 2001; Yanguas and Bergin, 2018). Despite an agreement regarding the mediating role of tasks, results are still far from a firm conclusion due to variations in findings. For instance, two studies (Ismail & Samad, 2010; Yilmaz & Granena, 2010) compared dictogloss and jigsaw tasks but they reached different conclusions. That is to say, while the former study found dictogloss to be superior in terms of the frequency of LREs, the latter study found no difference in the frequency of LREs when comparing the two tasks. Such contradictory findings are also evident in other studies. García Mayo (2002a) administered two tasks (dictogloss and text reconstruction) to pairs of Spanish L2 learners. Quantitative and qualitative results showed that pairs generated more LREs and higher degrees of attention to form during the completion of the text-reconstruction task than when performing the dictogloss task. In another study, García Mayo (2002b) administered five tasks (multiple choice, cloze, dictogloss, text editing, and text reconstruction) to pairs of Spanish L2 learners. Analysis of her data illustrated that text editing generated a higher number of LREs compared to the dictogloss task. Looking at these studies, one could realize that in addition to contrasting results, comparing studies in terms of the frequency of LREs seems to have become more difficult as they choose different tasks such as text composition, cloze test, text editing, narrative writing, dictogloss, jigsaw, and text reconstruction (e.g., García Mayo, 2002a, b; Kaivanpanah & Miri, 2017; Swain & Lapkin, 2001; Zabihi, 2022). In order to solve this problem, there seems to be a need for a new method to classify tasks that would focus on their nature and would make comparisons possible. In order to do so, we believe that it is advisable to see these tasks from the lens of input orientation (i.e., the provided amount of linguistic input and its method of delivery) which makes comparisons much easier. Such vision indicates that an audio file (dictogloss) is either similar with a set of pictures (jigsaw) when it comes to the frequency of LREs (Ismail & Samad, 2010; Swain & Lapkin, 2001; Yanguas & Bergin, 2018) or is superior (Yilmaz & Granena, 2010); however, an input in the form of audio generates fewer LREs when compared to a text given to students (either editing or reconstruction). Last but not least, some recent studies (Kaivanpanah & Miri, 2017; Zabihi, 2022) have tried something relatively new in their selection of tasks and compared an input in the form of a text to a topic (such as narrative writing or text composition) given to learners where they play a major role in creating the input. These two studies have indicated that the latter tasks, where students have to generate input in collaboration, lead to the production of more LREs. Since the tasks used in all previous studies have their specific level of input, we aim to choose tasks with different ranges of input to see what results might emerge. In order to take the analysis of LREs a step further, we have opted to include the outcome and the type of LREs in addition to their frequency. This is because different tasks require students to pay attention to the different aspects of the language, such as grammar (text editing), vocabulary (cloze task), and sometimes a combination of these (e.g., narrative writing).

Research Questions

As shown above, few studies to date have attempted to assess the effectiveness of task type on the frequency of LREs (e.g., García Mayo, 2002a) or on the outcome of these episodes (e.g., Yilmaz & Granena, 2010). Although several studies have found empirical evidence for the mediating role of tasks on the generation of LREs through using different types of these activities, to the best of our knowledge, there is a rarity of studies applying the same tasks (dictogloss and text reconstruction or cloze task and composition task) in different contexts to see if similar results emerge. For example, while García Mayo (2002a) used dictogloss and text reconstruction, Swain and Lapkin (2001) adopted jigsaw and dictogloss tasks, and Ismail and Samad (2010) used dictogloss and opinion gap tasks to claim that LRE production can be affected by task type. In addition, given that these tasks vary in terms of the level of input orientation, it can be considered a mediating factor in the generation of LREs that, as far as we know, has not been investigated in any study. Another point to note is that although the aforementioned studies reached the same conclusion regarding the link between task type and LREs, their primary focus was on learning vocabulary (Kaivanpanah & Miri, 2017), attention to form (García Mayo, 2002b), or learning grammar (Zabihi, 2022). This means that there was no room for a more thorough analysis of LREs but their frequency and therefore, it is still not clear what types of LREs dominated students’ conversations and what outcomes they reached. With that in mind, we use two tasks that are rarely put together (i.e., text editing and dictogloss) to see if the results converge with previous studies, and in order to shed more light on the effect of input orientation on task performance, we chose argumentative writing as the third task since it has the lowest level of input among our tasks and makes the concept of “having tasks with varying levels of input orientation” possible. This research will use three tasks with varying levels of input orientation to examine the frequency (i.e., the number of times when LREs emerge), type (i.e., lexical or grammatical), and outcome (i.e., solved correctly or incorrectly and unresolved) of LREs produced while students work collaboratively:

  1. 1.

    Does task type have a significant effect on the frequency of LREs produced in L2 peer interaction?

  2. 2.

    Does task type have a significant role in the type of LREs (lexical or grammatical) learners produce in L2 peer interaction?

  3. 3.

    Does task type have a significant role in the outcome (solved correctly, solved incorrectly, and unresolved) of LERs learners produce in L2 peer interaction?

Method

Participants

Twenty-four English as a Foreign Language (EFL) learners (female, n = 18; male, n = 6; Mage = 22.79, range = 18–32) agreed to take part in this research by signing informed consent forms and were offered extra course credit for their participation. It is worth noting that the small number of participants was due to the restrictions caused by the COVID-19 pandemic. These learners were first-year college students with at least 5 years of formal English instruction at school. They were put into 12 self-selected pairs and were asked to complete three tasks (argumentative writing, dictogloss, and text editing). All students were native speakers of Farsi and were selected according to English proficiency level based on their scores on a version of the Oxford Placement Test (OPT, Dave, 2004) and the proficiency reports we received from the institution where they studied English. Out of a total of 50 EFL learners who took the OPT, 24 independent users (B1 and B2) were selected based on the levels defined by the Common European Framework of Reference for Languages (Council of Europe, 2001) as the desired participants of the study.

Tasks

The tasks used in this study were argumentative writing (Appendix A), dictogloss (Appendix B), and text editing (Appendix C). These tasks were assigned since they have varying levels of input orientation. In fact, input orientation increases from argumentative writing to dictogloss and finally text editing. Simply said, while argumentative writing gives participants only a topic to write about, in a dictogloss task, students listen to an audio for two times and collaboratively write down what they have understood. The text editing task has the highest level of input orientation because students have a text at their disposal containing flaws that should be corrected. Following Storch’s (2008) argument that engaging learners in constructing a joint written text can open up noticeable opportunities for collaborative dialogues, an argumentative writing task was assigned to students. This task requires learners to jointly write about a given topic (The prompt was: How far do you agree with “It’s not government’s fault that young people can’t find jobs, it’s because they are not qualified”), discuss their arguments, and give their reasons. We decided to use an argumentative writing task because previous scholars have claimed that an unstructured task like argumentative writing can draw learners’ attention to lexis (Adams, 2007), promote vocabulary learning (Kaivanpanah & Miri, 2017), and encourage the generation of more LREs (Zabihi, 2022). The second task, i.e., dictogloss, was carried out in three phases: In the first phase, students listened to a recorded audio without being allowed to take any notes; in the second phase, the audio was played again and this time students could take notes if they deemed it necessary; in the last phase, all pairs were asked to collaboratively write down the audio they heard as precisely as possible based on what they understood and their notes. The dictogloss task has widely been used in previous studies (e.g., García Mayo, 2002a, b; Ismail & Samad, 2010; Yilmaz & Granena, 2010) because this task has proved to be effective as far as encouraging learners to reflect on their output is concerned. In addition, this task propels learners to explain their choices and work together to reconstruct a text, which may deepen their awareness of language form (Kowal & Swain, 1994; Lapierre, 1994; Swain, 1998; Swain & Lapkin, 2000, 2001). The third task was text editing which has been used in previous research (e.g., García Mayo, 2002b) and has proven to be effective in eliciting learners’ attention to a range of grammatical choices in a given text (Storch, 1997, 2007), leading to a more effective learning experience in both the short and long terms (Zabihi, 2022). Therefore, a text was given to students that comprised 300 words containing flaws in five different areas within their level of proficiency (prepositions, conditional clauses, collocations, passive structures, and past tenses), demanding learners to find errors and correct them while interacting with their pair members.

Procedure

Data Collection

A repeated measures (within-subjects) design was adopted in this study in order to avoid the risk of individual differences getting involved. One week before assigning the tasks and after taking OPT, 24 intermediate students from among a total of 50 were selected as participants of the current study. Pairs carried out the three tasks in three consecutive sessions. Before data collection began, ethics approval was granted by the language center and informed consent was obtained from all participants. The tasks (text editing, dictogloss, and argumentative writing) were assigned in three consecutive sessions, and in order to control for any possible effects of order of presentation on task performance, tasks were counterbalanced across three groups of four pairs. Pairs 1 to 4 were put into group 1, pairs 5 to 8 were put into group 2, and pairs 9 to 12 were put into group 3. Each group followed its specific order for carrying out the tasks, with group 1 doing argumentative writing, dictogloss, and text editing; group 2 doing dictogloss, argumentative writing, and text editing; and group 3 doing text editing, argumentative writing, and dictogloss. Learners were asked to prioritize speaking in English when doing tasks. They were allowed to use their L1 (Persian) only when they were faced with a communication barrier. This is because translanguaging leads to more effective communication among learners, while speaking only in L2 could discourage students from discussing linguistic issues (García, 2009). All conversations across twelve pairs were recorded and then transcribed verbatim for analysis.

Data Coding and Analysis

All the transcripts were coded for LREs. In this study, instances, where pair members faced a linguistic problem and tried to mutually resolve it through talking about the language they produced, questioning the language they produced, and giving justifications for their linguistics choices, were considered LREs (Swain & Lapkin, 1998). Two raters independently identified and coded LREs. The first rater was an assistant professor of applied linguistics who had prior knowledge of coding LREs. The second rater was an M.A. student of applied linguistics whose research was related to LREs and has been trained to code these episodes. Based on the Spearman-Brown Prophecy Formula, the inter-rater reliability was high (α = .88). The two raters then discussed and resolved all discrepancies. All disagreements between raters were discussed and resolved. LREs were sorted into three categories: frequency, type, and outcome. Frequency is the number of times when an LRE emerges; type refers to the nature of LREs being lexical or grammatical. In lexical LREs, learners talk about meaning, spelling, or the choice of lexical items. The following is an instance of Lexical LRE:

  1. (1)

    S1: aval mikhasti begi ala raghme, are? (At first, you wanted to say “despite,” right?)

  2. (2)

    S2: ohoom, apart from manzooram bood. Doroste? Nazari? (Yeah, I meant to say “apart from.” Is that correct? Any idea?)

  3. (3)

    S1: despite ham mishe goft (You can also say “despite.”)

  4. (4)

    S2: aha (Alright.)

  5. (5)

    S1: benazaret kodoom behtare begim, despite ya apart from? (Which one is better? “despite” or “apart from”?)

  6. (6)

    S2: harchi khodet mikhay ro bego. Man mikham begam ala raghme (Whatever you like. I just want to say “despite.”)

In grammatical LREs, learners talk about the grammatical aspects of the language they produce, such as singular or plural, past or present tense, gerund or infinitive, and the like. The following is an instance in which learners produce grammar-related LREs:

  1. (1)

    S1: hamin ke bazi az afrade khas daraye emtiazati hastan (Some people have privilege over others)

  2. (2)

    S2: bayad felesh mofrad bashe ya jam? Have ya has? (Should we use plural or singular verb? “Have” or “has”?)

  3. (3)

    S3: has dige (We should use “has.”)

Finally, the outcome of LREs was further compartmentalized into three subcategories following Swain (1998): solved correctly, problem not solved or disagreement about the problem solution (unresolved), and problem solved incorrectly. Correctly resolved LREs took place when the correct target language form or an appropriate explanation for it was provided. Here is an example of a correctly resolved LRE:

  1. (1)

    S1: Khob begim tashkhis midan? Mishnasan? (Should we say they identify? They know?)

  2. (2)

    S2: yani chi? (What do you mean?)

  3. (3)

    S1: baraye verbesh migam. Mishnasan estedadeshoono? Chi behtare? (For the verb. They know their talent? What is better to say?)

  4. (4)

    S1: recognize? (Recognize?)

  5. (5)

    S2: bale (Yes.)

An LRE was categorized as unresolved when learners dropped the topic because they either could not find an answer to the problem or could not reach a mutual solution. In the following example, student 1 (S1) tries to ask about the noun form of a verb, but the other student (S2) disagrees with her pair member on the ground that the whole idea is erroneous in her vision:

  1. (1)

    S1: Prefer mishe dekhalat, esmesh chie? (“Prefer” means interfere, what is the noun form of it?)

  2. (2)

    S2: chi? (What?)

  3. (3)

    S1: na, prefer nemishe. Dekhalat chi mi she? (No, that is not “prefer.” How do you say interfere?)

  4. (4)

    S2: kolan motevaje jomlat nemisham (I do not understand your sentence at all.)

  5. (5)

    S1: mikham begam dekhalate mardom va dolat tasir dashte (I want to say that interference from government and people has had an impact.)

  6. (6)

    S2: na, mardom ke dekhalat nemikonan (No, people do not interfere.)

Instances where one’s attempt to start an LRE was unnoticed by the other learner were coded as unresolved as well:

  1. (1)

    S1: All people were entering, bad az start ‘ING’ darim? When the door, all people start? Started? Badesh fel be che soorate? (All people were entering…, should we use “ING” after star? When the door, all people start? Started? What form of the verb should we use after this?)

  2. (2)

    S2: Ye lahze, inja chi dashtam mineveshtam? (Hold on a second, what was I saying in this part?)

An LRE was incorrectly solved when the learners agreed on a solution that was not target-like, agreed on an erroneous grammatical structure, or when they provided an incorrect explanation:

  1. (1)

    S1: Economic a miad ya an? (Should we say “a economic” or “an economic”?)

  2. (2)

    S2: A economic. Fekr mikonam ‘a’ bayad biari. Aha, bayad ‘a’ bashe. (A economic. I think you should use “a,” yeah, that is correct. We should use “a.”)

Results

Table 1 shows the descriptive statistics for the frequency of LREs produced across the three task types. In light of the repeated-measures design, learners’ LRE production for each task was summed and compared to determine whether there was a variation that could be attributed to task type. Before running ANOVAs, we checked the assumption of Homogeneity of Variance using Mauchly’s test and the result was not significant (p = .135), suggesting that the variances of differences were not significantly different and the data met the assumption required by ANOVA. Furthermore, results from a repeated measures ANOVA generally showed that the effect of task type on the frequency of LREs produced by the learners was statistically significant, F(2, 22) = 18.153, p = .000. The effect size for task type using partial eta squared (ηp2) was 0.623, representing a large effect (based on Cohen, 1988: small = .0099; medium = .0588; large = .1379). In addition, post hoc pairwise comparisons with Bonferroni correction showed that learners significantly produced more LREs while doing the argumentative writing task (p = .000) and the text editing task (p = .006) than when they did the dictogloss task. The mean values also indicated that learners produced more LREs when doing the argumentative writing task than when they did the text editing task. However, the difference was not significant (p = .227).

Table 1 Descriptive statistics: production of LREs across the three tasks
Full size table

Table 2 shows the descriptive statistics for the frequency of types of LREs produced across the three task performances. Learners’ production of lexical and grammatical LREs for each task was compared to determine whether there was a variation that could be attributed to task type. First, Mauchly’s test of Homogeneity of Variance was run and the results were not significant for lexical LREs (p = .076) and grammatical LREs (p = .107), indicating that the data met the assumption required by ANOVA. Furthermore, results from a repeated measures ANOVA showed that the effect of task type on the frequency of lexical LREs was statistically significant, F(2, 22) = 39.95, p = .000, partial eta squared (ηp2) = 0.784: large effect size (see Table 3). Post hoc pairwise comparisons with Bonferroni correction revealed that learners significantly produced more lexical LREs while doing the argumentative writing task than when they performed either the dictogloss task (p = .000) or the text editing task (p = .000). The mean values also indicated that, although learners produced more lexical LREs during the completion of the text editing task than in the dictogloss task, the difference was not significant (p = .056). On the contrary, regarding grammatical LREs, a comparison of means showed that, among the three tasks, learners produced the highest number of grammatical LREs when completing the text editing task. ANOVA results showed a statistically significant effect of task type on the frequency of grammatical LREs, F(2, 22) = 19.71, p = .000, partial eta squared (ηp2) = 0.642: large effect size. In addition, ANOVA results showed that the number of grammatical LREs produced in the text editing task was significantly higher than that produced during the completion of the argumentative writing task (p = .009) and the dictogloss task (p = .000).

Table 2 Descriptive statistics: types of LREs produced across the three tasks
Full size table
Table 3 One-way repeated-measures ANOVA: tests of within-subjects effects
Full size table

Finally, Table 4 shows the descriptive statistics for the resolution of LREs produced across the three task types. Learners’ performances on the three tasks were compared regarding the resolution of LREs to see if there was a variation that could be attributed to task type. First, Mauchly’s test of Homogeneity of Variance was run and the results were not significant for correctly resolved LREs (p = .515), incorrectly resolved LREs (p = .062), and unresolved LREs (p = .897), indicating that the data met the assumption required by ANOVA. Furthermore, results from a repeated measures ANOVA generally showed that the effect of task type on the frequency of correctly resolved LREs produced by learners was statistically significant, F(2, 22) = 12.55, p = .000, partial eta squared (ηp2) = 0.533: large effect size. More specifically, post hoc pairwise comparisons with Bonferroni correction showed that learners significantly correctly resolved more LREs during the completion of the argumentative writing task (p = .000) than they did in the dictogloss task (p = .000). However, no significant difference was found either between the argumentative writing task and the text-editing task (p = .078) or between the text-editing task and the dictogloss task (p = .188) regarding the correct resolution of LREs. Our results also showed a non-significant effect of task type on the frequency of unresolved LREs (F(2, 22) = 2.65, p = .093) and incorrectly resolved LREs (F(1.361, 14.970) = 4.812, p = .053), indicating that none of the tasks performed by the learners has led them to resolve LREs incorrectly or leave them unresolved more often than the other two.

Table 4 Descriptive statistics: resolution of LREs across the three tasks
Full size table

Discussion and Conclusion

The first research question asked whether task type would impact the frequency of LREs. Statistical analyses revealed that, overall, learners produced more LREs when working on argumentative writing. Therefore, our study lends support to previous research claiming that task type could be a mediating factor in the production of LREs (e.g., Swain & Lapkin, 2001; García Mayo, 2002a, b). However, these findings are contrary to what Ismail and Samad (2010) found, as their research showed that dictogloss (a task with more input) increases the production of LREs, while here, a task with the least level of input (argumentative writing) generated the highest number of LREs. Also, our research defies Yanguas and Bergin’s (2018) claim that tasks (with varying levels of input) do not affect the production of LREs. One explanation for our findings might be that when students work on the argumentative writing, unlike other tasks, they need to talk more as this task requires providing reasons and supporting their claims. Therefore, it is not counter-intuitive to suggest that learners will have more chances to produce LREs when they talk more. Another explanation for such finding could be that argumentative writing provides learners with the least level of input, meaning that they need more negotiation for choosing words and structures that all require talking about their linguistic productions. In view of this, second language teachers can use argumentative writing or other tasks with lower levels of input to promote negotiation of meaning and communication between learners.

The second research question asked whether task type would affect the type of LREs produced (i.e., lexical or grammatical) during peer interaction. Our analyses illustrated that learners significantly produced more lexical LREs while doing the argumentative writing task than when they performed the dictogloss task. This finding is different from what Yilmaz and Granena (2010) found because their study indicated that the dictogloss task contributed to the generation of more lexical LREs. The difference in the results of Yilmaz and Granena (2010) could be justified by the fact that, in their study, prior to carrying out the tasks, participants had the chance to practice doing the tasks in face-to-face mode. This could have increased learners’ familiarity with the tasks and thus, instead of finding dictogloss a barrier to producing lexical LREs, learners used this as an opportunity to talk more about lexis-related issues. The higher frequency of lexical LREs in our study might be due to the fact that argumentative writing, as the task with the lowest amount of input, requires learners to talk more about abstract words that can create more negotiation for choosing the right lexis. Concerning grammatical LREs, the text-editing task contained a significantly higher number of these episodes compared to argumentative writing and dictogloss. This finding corroborates earlier claims about the appropriateness of the text-editing task for eliciting grammatical type of LREs (e.g., Storch, 1997, 2007). Thus, it is reasonable to conclude that structured tasks like text editing can better shift learners’ attention to grammatical features compared to unstructured tasks such as argumentative writing that usually tend to draw learners’ attention to lexical features or even semi-structured tasks like dictogloss (Adams, 2007). This finding implies that if L2 educators want to draw learners’ attention to formal aspects of language, they can use text editing. Simply put, in order to provide some sort of practice after teaching grammar, teachers can use text editing or other tasks that provide a framework in the form of a text (input) so that learners can dedicate most of their attention to grammatical features.

The third research question asked whether the type of task used would mediate the resolution of LREs. Results indicated that learners correctly resolved more LREs while doing argumentative writing compared to the time when they performed the dictogloss task. This finding supports Yilmaz and Granena’s (2010) findings in that their analysis showed that jigsaw (as a task with a lower level of input) led to more correctly resolved LREs compared to the dictogloss task. However, what Yanguas and Bergin (2018) found is contradictory as they concluded that task type does not lead to a significant difference in the outcome of tasks. Our finding might be justified by the assumption that when learners work on tasks allowing them to choose words and structures by themselves, they usually use those that they know well; thus, they can easily explain their choices upon being asked by their interlocutors. As a result, those tasks that offer less input (e.g., argumentative writing) and let learners generate output can benefit students more because they create affordances for pair members to find gaps in their knowledge and fill them with the help of their interlocutors.

Finally, it is important to bear in mind that the empirical results reported herein should be considered in the light of some limitations that could be addressed in future research. First, since our study showed that tasks with less input (argumentative writing) promoted a higher number of lexical LREs compared to dictogloss which differs from what previous research (e.g., Yilmaz & Granena, 2010) found, more investigation is needed to ensure which task encourages more LREs and which type dominates interaction. Another point of note is to include learners of other proficiency levels to see if their conversations are dominated by grammatical or lexical issues and whether the same results emerge. In addition, given that our study focused on independent users of English (both B1 and B2 levels), one might argue that the participants’ mixed proficiency could downgrade the validity of the results, as previous research has shown that learners’ L2 proficiency may have a great impact on task performance, collaborative dialogues, and interaction dynamics (e.g., Storch & Aldosari, 2012; Vahdany et al., 2016; Watanabe & Swain, 2007). Finally, due to homogenizing students in terms of proficiency, one could argue that the study lacks ecological validity because it was not carried out in a regular L2 writing classroom, and that the findings may not be directly applicable to such a context. In view of this, future research could enhance the ecological validity of this line of research by exploring the effect of task type and input orientation in the context of L2 writing classrooms.