Task-modality effects on young learners’ language related episodes in collaborative dialogue

In adult learners’ collaborative dialogue, oral+written tasks have been found to promote a greater incidence and resolution of language-related episodes and to demand higher levels of accuracy than oral tasks thanks to the extratime learners have to reflect on their written outcome. No previous studies have tested whether asking learners to attend to accuracy in both modalities would yield similar results. The present study with twenty-three dyads of young English learners supports the superiority of the oral+written modality in the promotion of learning opportunities, even if learners are encouraged to focus on form in the oral modality, a result reinforced by the incorporation of target-like resolved episodes in the written product. However, the intragroup analysis reveals that young learners focus on meaning in equal terms, present low rates of target-likeness, and do not elaborate their resolutions, all of which can be ascribed to their younger age and developing metalinguistic awareness.


Introduction
Studies on collaborative dialogue, operationalized as language-related episodes (LREs) across task modalities (oral vs. oral+written), have received limited attention in the literature and the vast majority have examined adult learners (Adams & Ross-Feldman, 2008;Niu, 2009;Payant & Kim, 2019), except for García Mayo and Imaz Agirre (2019). All these studies have ascribed the greater incidence and higher number of resolved LREs in oral+written tasks to the extra processing time learners have to reflect on their production. However, no studies exist that have actually tested whether asking learners to attend to accuracy in both modalities would yield similar results. Likewise, very little empirical evidence exists as regards children's attitudes towards these tasks (Shak & Gardner, 2008).
This study will try to fill these gaps by analysing the amount and types of LREs produced by primary-school Basque/Spanish bilinguals learning third language (L3) English in a Content and Language Integrated Learning (CLIL) setting in two tasks with different modalities: a speaking task and a speaking+writing task. In addition, this study will shed more light on learner attitudes by measuring their motivation before and after performing these tasks.

Collaborative dialogue and Language-Related Episodes
Collaborative dialogue is considered a source of language learning and development (Brooks & Swain, 2009;Donato, 1994;Swain & Lapkin, 2002;Watanabe & Swain, 2007), as during this dialogue learners may "form and test hypotheses about appropriate and correct use of language, as well as reflect on their language use" (Swain & Watanabe, 2013, p. 3). In particular, LREs are the unit of analysis to operationalize the construct of collaborative dialogue. Swain and Lapkin (1998) defined LREs as "any part of a dialogue where the students talk about the language they are producing, question their language use, or correct themselves or others" (p. 326). The production of LREs has been positively correlated with subsequent performance on tailor-made post-tests (Kim, 2008;Swain, 1998;Swain & Lapkin, 2001).
Task-modality, which is the focus of the present paper, has been one of the factors that seems to affect the production, nature, and resolution of LREs. However, task-modality has received limited attention in the literature and the vast majority of studies have examined adult learners, as we will see in the next section.

Task-modality and LREs
Several studies have attested that some tasks draw learners' attention to form more than others. For example, in more structured tasks such as multiple choice and text repair, learners focus more on form than in less structured tasks such as dictogloss (Adams, 2006). Likewise, research dealing with task-based interaction has also examined the role of different production modes in language learning. It has been argued that speaking and writing impose unique cognitive demands (Payant & Kim, 2019): speaking is ephemeral and characterized by its immediacy, with very little planning and editing time for the learners, whereas writing is more visual and permanent and offers learners more time to think and more opportunities to reflect on their product. Likewise, in writing learners can more easily retrieve declarative knowledge due to additional processing time (Gass, Behney, & Plonsky, 2013, as cited in Payant & Kim, 2019). In other words, it aids in raising language awareness as it demands learners to express thoughts in a more precise way (Wolff, 2000). In addition, in classroom settings learners very often consider that teachers or peers could potentially assess their written products, which also leads to an increase in their orientation to form in writing (Adams, 2006).

Adams and Ross-Feldman (2008) examined 44 high-intermediate ESL learners
with L1 Spanish while collaboratively completing two tasks targeting past tense and locative prepositions. These tasks included speaking and writing components. Half of students performed the speaking and writing part simultaneously, while the remaining students carried out the speaking part in the first place and subsequently the writing part. LREs were codified in terms of their focus (meaning vs. form), complexity of focus (complex vs. simple), directness of focus (direct vs. indirect) and resolution (resolved vs. unresolved). The overall analysis of the results indicated that in the case of locatives, learners significantly produced more LREs in the writing component. As regards types, learners also produced more form and complex LREs as well as direct and resolved LREs, even though not significantly. In the case of past tense, the different categories analyzed did not yield statistically significant differences. However, the descriptive statistics showed a greater number of LREs, as well as more form, complex, direct, and resolved LREs in the writing part. When comparing dyads doing the tasks simultaneously to those doing them sequentially, no significant differences emerged in either category, except for resolution in favor of dyads performing the tasks sequentially. In sum, task-modality influenced the amount of LREs, while the order of administration affected the resolution of LREs.
In an English as a Foreign Language (EFL) setting, García Mayo and Azkarai (2016) explored the effect of task-modality on the incidence, nature, and outcome of LREs produced by Basque-Spanish bilingual learners of L3 English. Same-proficiency dyads were asked to perform four different tasks: picture placement and picture differences constituted the speaking modality tasks and dictogloss together with text editing the speaking+writing modality tasks. In terms of incidence, a greater number of LREs were observed in the writing tasks. As for the nature of LREs, those LREs dealing with form were more common in writing, while those dealing with meaning were more common in speaking. In the case of outcome, writing tasks yielded a higher number of resolved LREs. Niu (2009) also examined the production of LREs by upper intermediate EFL learners in a Chinese university. Eight same-gender pairs were asked to perform a textreconstruction task. These dyads were randomly arranged to complete the task either as a collaborative oral output task or as a collaborative written output task. Those performing the collaborative written output task were found to initiate more LREs related to lexis, grammar and discourse than those doing the oral output task. Likewise, learners in the written output task provided more justifications and explanations while discussing the language forms they focused on. Similarly, learners in both the oral output and the written output condition made correct decisions while resolving the LREs.

Payant and Kim (2019) tested L1 Spanish intermediate university learners in
Mexico while performing two decision making tasks with oral and written components in L3 French. Similarly, tailor-made post-tests based on the LREs produced during each component were administered to the learners. Modality effects were visible in the higher production of LREs and in the more target-like resolutions during the written modality. However, the focus of LREs was partially affected by task-modality, as even if meaning-based LREs were more common in the speaking component, the ratio of both lexis-and form-based LREs was comparable. Additionally, the incidence of resolutions facilitated language development in the post-test even if learners did not resolve in a target-like manner.
As aforementioned, research conducted with young learners along these lines is in its infancy. García Mayo and Imaz Agirre (2019) gathered data from sixty-two 6 th primary school children from three different intact classes. They were asked to complete two tasks: an oral task and an oral+written task. Both were decision-making tasks, but the second one required learners to submit a written product. As in the case of research conducted with adults, more LREs were produced by 6 th year primary school learners in the speaking+writing modality. As regards the nature of LREs, lexical LREs were more common than form in both the speaking and the speaking+writing task. The authors argue that it seems as if these young learners were needed to produce more lexical LREs to move both tasks forward. With respect to the resolution of LREs, as in research with adults, a higher percentage of resolved LREs was obtained in the speaking+writing task.
The review of studies on the impact of task-modality on collaborative dialogue indicates that in general terms bi-/multilingual learners have been reported to produce and resolve a greater number of LREs in those tasks that combine oral and written modalities (Adams & Ross-Feldman, 2008;García Mayo & Azkarai, 2016;Niu, 2009;Payant & Kim, 2019). This trend has also been very recently attested in young learners (García Mayo & Imaz Agirre, 2019), but this line of research is in its infancy and more investigations are needed in this respect. All these studies have ascribed the greater incidence and higher number of resolved LREs in speaking+writing tasks, as well as the existence of greater accuracy in these tasks, to the extra processing time learners have to reflect on their production that this type of tasks inherently offers. However, no studies exist that have actually controlled for the different levels of accuracy that speaking and speaking+writing tasks demand as a consequence of their on-line and off-line nature, respectively. Thus, this study will look into whether the framing of the task by asking learners to attend to accuracy in both modalities could overrule the inherent focus of the task (Philp, Walter, & Basturkmen, 2010).

Task motivation
Research has also examined learners' perception of the tasks they perform in order to learn a language. The relationship between tasks and motivation is suggested to be shaped by task engagement (Dörnyei & Kormos, 2000) rather than correlated with pretask value beliefs (Al Kahlil, 2016). In turn, task engagement has been found to be linked to factors related to task conditions such as topic choice and cognitive complexity (Poupore, 2013) or the opportunities for interaction and collaboration with peers. In general, collaborative tasks are perceived to be more engaging than individual tasks (Julkunen, 2001;Kopinska & Azkarai, 2020), whereas learner grouping modes (pairs vs. groups) in collaborative tasks do not seem to modify task motivation substantially (Calzada & García Mayo, 2020;Fernández Dobao & Blum, 2013).
Overall, learners report that collaborative tasks help them to construct knowledge, boost their self-confidence and even learn grammar (Lin & Maarof, 2013), more positive attitudes being developed as learners gain familiarity with the task (Kopinska & Azkarai, 2020;Shak, 2006;Shak & Gardner, 2008) or if the collaborative experience with the other task performers is optimal (Chen & Yu, 2019). Besides, learners seem to perceive the blend of the written and the oral mode positively (Calzada & García Mayo, 2020), even though experimental research comparing different modalities is lacking.
It is important to mention that most of the investigations on task motivation reported above have been conducted with adolescents, university students or adult learners. Research on task motivation with young learners is still scarce (Muñoz, 2017a). With the exception of Shak and Gardner (2008), task motivation studies with young learners have focused on the exploration of just one task, namely a dictogloss task (Calzada & García Mayo, 2020;Kopinska & Azkarai, 2020;Shak, 2006). Learners in Shak's (2006) study did not enjoy the writing stage of this task probably because they performed that part individually. Conversely, participants in Calzada and García Mayo (2020) and in Kopinska and Azkarai (2020) felt more motivated as a consequence of their reported positive attitudes towards the collaboration and the opportunities for peer assistance that the dictogloss task fostered as well as the blend of the oral and the written modality that they experienced. Taking into account the lack of research on the interface between motivation and the examination of other tasks as well as taskmodality, we are in the need of further studies that come to fill this gap. It is important to know if tasks actually appeal to young learners, as very little empirical evidence exists in this respect (Shak & Gardner, 2008). Looking into students' attitudes before and after performing the tasks will shed more light on the feasibility and effectiveness of this type of tasks in primary education classrooms.

Research questions
Given the scarcity of research along these lines with young learners, this study will analyse the amount and types of LREs produced by primary-school Basque/Spanish bilinguals learning L3 English in a CLIL setting when performing a speaking and a speaking+writing task. In particular, we address these research questions: 1. Are there any differences between task-modalities in terms of number, types, and outcome of LREs?
2. What are the most common types of LREs (nature and outcome) in each task?
3. How does the resolution of LREs affect the written product in the speaking+writing task?
4. Are there any differences between the two tasks in terms of student motivation?

Participants
Fifty primary school learners took part in the study.  (Cenoz & Valencia, 1994). The CLIL programme, in turn, was an attempt to further improve these learners' foreign language (English) proficiency.
As far as English exposure at school is concerned, all the participants had begun learning English as a school subject at about age 4 in pre-primary education. Some years later, when they were in primary education

Instruments
Data were collected by means of two instruments where students had to collaborate in same-proficiency pairs, namely a speaking task and speaking+writing task. The speaking task was adapted from an activity taken from the book Sparks 1 (House & Scott, 2009) and it consisted of two different phases. In the first phase, students had to arrange six disordered pictures into a meaningful story. The second phase asked students to tell the story in turns. Picture-ordering plus story-telling tasks have been profusely employed in prior investigations with adults and children for similar purposes The speaking+writing task was specifically designed for the purposes of this study and it also consisted of two different phases. In the first phase, students had to look at two different pictures -the first one shows a boy who has found a lost dog in a park and the second one shows the potential owners of the lost dog, their professions, and a city map with the places where they work. Students had to decide who the owner of the dog is on the basis of some clues, namely a picture on this person's shirt. They also had to guess where the owner of the dog worked on the map. In the second phase, students were asked to write a note for the boy, explaining who the owner of the dog is and why, and also giving directions from the park to the place where the owner works so that the boy knows how to take the dog back to its owner. Similar decision-making tasks have been administered in previous research for similar purposes (Azkarai & García Mayo, 2015; García Mayo & Imaz Agirre, 2019) 1 .
A third instrument was added to this study in order to obtain information about students' motivation towards each of the two collaborative tasks described above at two different times -immediately before and immediately after the completion of each task.
The instrument was designed according to the scale used in a previous study measuring learners' perceptions across different times (before, during or after task completion), namely Al-Kahlil's (2016) 'task-related motivation thermometer', where motivation was measured by means of a Likert scale where 1 point meant the lowest motivation and 10 points the highest motivation. The purpose of this instrument was to gain face validity in our research and to make sure that students were motivated at the outset of the study so that they could give their best when performing the collaborative tasks. We also wanted to know whether the tasks were motivating to them, which will eventually inform about the suitability of this type of tasks for primary education students, in addition to discovering any task-modality differences with regard to motivation.
Besides, as mentioned in the previous section, all participants were tested on English proficiency through their participation in the listening, reading, and writing sections of the Key English Test (KET; UCLES, 2014) at the outset of the study. This exam is proof of one's ability to communicate in English in simple situations.

Data collection
Once parental and school permission was issued, students were first tested on English proficiency by means of the KET. The test was administered during regular lessons in class and students were givenone hour and 40 minutes to complete it.
The two collaborative tasks were accomplished by the student pairs in a quiet room at school. Data were collected over a period of two weeks in two different sessions. They performed the speaking task inthe firstdata-gathering session and the speaking+writing task inthe next session. Before the completion of each task, students completed the motivation scale individually, for which they were given 1 minute, and were subsequenly reminded of the importance of their paying attention to accuracy in the second phase of each task, that is, in the story-telling and note-writing respectively.
Taking into account that writing allows for higher levels of accuracy due to the off-line nature of the task when compared to speaking (Adams & Ross-Feldman, 2008; Azkarai& García Mayo, 2015;García Mayo & Azkarai, 2016;Niu, 2009;Payant & Kim, 2019), by asking learners to attend to accuracy in both modalities, we could verify whether the framing of the task rules out the inherent focus of the task (Philp, Walter & Basturkmen, 2010). It is also important to note that a researcher was with the students in the room where the tasks were being carried out, but participants were asked to perform the task with their resources at hand and to avoid asking the researcher for help. On average, learners needed about 15 minutes to carry out each collaborative tasks. Once they had finished each task, they were given one minute to complete the motivation scale again, which they had to fill out individually.

Data analysis
The collaborative tasks were both audio and videotaped. Recorded productions were orthographically transcribed and later codified into CHILDES (MacWhinney, 2000) for the production of LREs with the help of CLAN protocols. All turns in which students engaged in language discussion or self-correction were identified as LREs in each task by two independent researchers, who jointly came to an agreement in those cases in which any controversy in their classification of the LREs was detected.
As for the classification of the LREs in eack task, we firstly followed Adams and Ross-Feldman (2008) and García Mayo and Azkarai's (2016) classification in terms of nature, that is, LREs were classified into two main categories, namely meaningfocused and form-focused. The former includes cases of word meaning or word choice whereas the later comprises those LREs involving spelling, phonology, morphosyntax,and prepositions. Secondly, the aforementioned authors' taxonomy of LREs according to outcome was considered and each of the main categories of LREs were secondly classified as resolved or unresolved. The former include those cases in which the LRE reached a resolution, regardless of whether it was resolved in a targetlike or non-target-like manner, categories which were added to our LRE classification according to Payant and Kim's (2019) distinction between correctly and incorrectly resolved LREs. The latter refered to those cases in which the language concern was left unresolved and no answer to the linguistic inquiry was provided by any of the members of the dyad. Thirdly, in the case of the task including a written product, we added an original classification in which resolved LREs were further classified into incorporated and non-incorporated, depending on whether the resolution that dyad members reached was incorporated (or not) in the eventual written outcome of the task.  (1) Unresolved meaning-focused LRE.
*CHI1: she's make eh another toy but eh (.) eh (.) mejor (Eng: 'better') how do you say mejor? *CHI2: I don't know. (whispering) In (1), the first child asks the second child how the Spanish word mejor is said in English, but the second child does not know, leaving the question unanswered.
*CHI1: is the same picture. *CHI2: is the same snake? *CHI1: is the same snake [Written output: 'is the seim snake in the two pictures'] In (2), the second child corrects the first child and changes the word picture for a more precise word defining what is in the picture, that is, a snake. The first child immediately incorporates the word snake in the next turn, which is also included in the final written output.
*CHI1: eh serpiente (Eng: snake) *CHI2: the snake *CHI1: the snake *CHI2: the snake *CHI1: is here In (3), the first child produces a word in Spanish (serpiente) and the second child gives him/her the English term (snake) in the next turn. The word is accepted by the first child immediately after, but then it will not appear in the eventual written output.
(4) Incorporated non-target-like resolved meaning-focused LRE. In (4), the first child wants to know how Spanish después is said in English. The second child provides an unintelligible word, so the first child decides first to adapt this Spanish word to English phonologically and attempts the foreignised term dispos, but then he decides to borrow the term from Basque and uses the term gero. The Basque word is finally accepted by the second child in the last utterance and it will be incorporated in the written outcome eventually.
In (5), the two children are discussing how to refer to the male character in the story, the terms child, children, and boy being entertained. They finally opt for the inaccurate phrase 'a children boy', which will not ultimately be incorporated in the written output. and front of (.) eh church mm ha eh is the (.) vet eh clinic.
In (6), the first child does not know how to pronounce the word 'church' and informs about it to his partner, who advised him to just write it and forget about its pronunciation.
(7) Incorporated target-like resolved form-focused LRE.  (7), the first child asks his partner how the English word 'dog' is spelt. The second child provides him/her with the right orthographical transcription of the word, which will also be part of the final written output.
(8) Non-incorporated target-like resolved form-focused LRE. *CHI1: we puts the name in your names? *CHI2: we puts we put our?
In (8), the first child incorrectly produces the verb form 'puts' for the first plural person 'we'. The second child corrects the first child and provides the accurate form 'put' without the 3 rd person singular present tense morpheme. Neither form will appear in the written outcome of the task.
*CHI1: then you have to go. *CHI2: you have to eh. *CHI1: xxx to go of eh at garden road [Written output: 'then you have to go at garden road'] In (9), the first child is not sure about the right preposition to be used with the verb to go. She first attempts the preposition 'of' inaccurately but immediately after she selfcorrects herself and opts for the wrong preposition 'at', which will be incorporated in the written outcome.
In (10), the first child suggests the preposition 'in' but the second child puts this suggestion into question although he later accepts it in his utterance 'is on the vet clinic'. However, the first child changes to 'on' in the following turn, a preposition which is finally accepted by the second child with an 'ok' in his last utterance. Neither preposition will be used in the written output eventually.
As far as statistical analyses are concerned, we computed both descriptive and inferential statistical procedures. As for descriptive statistics, the number of turns, the number of turns comprising LREs, the number of LRE types as well as their percentages, mean scores, and standard deviations were calculated. As far as inferential statistics is concerned, the LRE data were analysed to compare both the two tasks (intertask analyses) and the different LRE types within each task (intra-task analyses). As for student motivation analyses, means and standard deviations were obtained for each task at both the pre-task and the post-task phase. Motivation data were also explored through inferential statistics for both inter-task and inter-phase differences. All comparisons were made by means of non-parametric procedures (Wilcoxon signed-ranged tests), as Kolmogorov-Smirnov tests indicated that the distribution of the samples was not normal. An alpha level of .05 (*) was used for significant probability.

Results
In this section, we will show the results of the analyses performed to find answers to the four research questions. Tables 2 to 9 present the inter-task analyses conducted to explore the differences between the oral task and the oral+writing task in terms of number, types, and outcomes of LREs (RQ1). Intra-task analyses are also offered in Tables 3 to 9 so as to discover which types of LREs in terms of nature and outcome are the most common in each task (RQ2). The analyses carried out to discover whether the resolution of LREs appears in the written product (RQ3) are displayed in Tables 10 to   12. Finally, Table 13 offers the results pertaining to students' motivation.
As shown in Table 2, the data of our study was composed of 1197 turns in the oral task and 2236 in the oral+written task, of which 404 and 729 were turns comprising LREs respectively. As for the number of LREs, 110 episodes occurred in the oral task whereas 158 happened in the oral+writing task. In other words, learners' productions in general as well as the incidence of LREs in particular were more abundant in the oral+writing task. Mean scores revealed the very same pattern, since the oral task yielded a mean of 4.78 LREs per dyad and the oral+written task a mean of 6.83. The Wilcoxon test determined that the gap between the two mean scores was statistically significant.  Table 3), inter-task analyses indicated that there were no statistically significant differences between the meaning-focused LREs produced in the first and in the second task, as attested by their similar mean score figures (4.30 and 3.74). However, when the form-focused LREs produced were examined in each task, larger differences were found between the two means, the oral-written task leading to a significantly higher incidence of this type of LREs than the oral task (3.09 vs. 0.48). As for the intra-task comparisons, it is noteworthy to mention that the dyads did not behave alike in both tasks. In the oral task, there was a significantly higher occurrence of meaning-focused LREs than of form-focused LREs, with means of 4.30 and 0.48 representing 90% and 10% of the total number of LREs produced respectively. In the oral-written task, we found a more even distribution of meaning-focused and formfocused LREs, (55.06% and 44.94%) as well as a small gap between the mean scores for these two types of LREs (3.74 and 3.09). The Wilcoxon test did not uncover significant differences between the two types of LREs in the oral+written task. As for the outcome of the LREs, Table 4 displays the figures for all resolved and unresolved LREs in each of the two tasks. Inter-task comparisons demonstrated that there were significant differences between the mean scores of resolved LREs in the two tasks, the oral-writing task contributing to higher resolution means than the oral task (6.17 vs. 3.70). However, no statistically significant differences were attested between the two tasks as far as unresolved LREs are concerned. As for the intra-task contrasts, Wilcoxon tests pointed to significantly higher means of resolved LREs than of unresolved ones in both the first (3.70 vs. 1.09) and the second task (6.17 vs. 0.65).
Besides, the percentage of resolved LREs was much higher than that of unresolved ones, more so in the oral+writing (90.51% vs. 9.49%) than in the oral (77.27% vs. for meaning-focused and form-focused LREs independently. As for meaning-focused LREs (Table 5), inter-task comparisons revealed that there were no statistically significant differences between resolved LREs in the first and in the second task, with similarly high percentages (74.75% and 86.21%) and the same mean score (3.22) achieved. Nevertheless, the gap between unresolved meaning-focused LREs in the oral and in the oral+writing task did reach statistical significance. Unresolved LREs represented 25.25% of meaning-focused LREs in the oral task, with a mean of 1.09, whereas this percentage decreased to 13.79% in the oral-written task, with a mean of 0.52. In other words, unresolved meaning-focused LREs occurred to a significantly larger extent in the oral task. As far as the intra-task contrasts are concerned, both percentage and mean figures indicated that the production of resolved meaning-focused

22.73%) task.
LREs was more prominent than that of unresolved ones. As attested by the Wilcoxon tests, the gap between resolved and unresolved meaning-focused LREs was statistically significant in both tasks, even so more strikingly in the oral+writing task.  Table 6. In the inter-task analyses, it was observed that resolved form-focused LREs obtained a significantly higher mean in the oral+written task than in the oral task (2.96 vs. 0.48). However, the percentages and mean scores of unresolved form-focused LREs were extremely low, the difference between both tasks not reaching statistical significance. As for the intra-task comparisons, the pattern discovered in both tasks was that of a production of resolved form-focused LREs which was, by far, significantly more abundant than that of unresolved form-focused LREs, particularly in the case of the oral+writing task. The second type of analysis carried out to inquire the outcome of LREs involved whether these were resolved in a target-like manner. Table 7 displays the figures for all target-like and non-target-like resolved LREs in each task. When inter-task comparisons were made, it was discovered that the mean of target-like resolved LREs in the oral+writing task was significantly higher than the one in the oral task (4.65 vs. 2.04). However, as indicated by inferential statistics, there was no statistical support for the difference between the means of non-target-like resolved LREs in each task. Intratask contrasts, in turn, showed that the occurrence of target-like resolved LREs in the oral task was not statistically different from that of non-target-like resolved LREs. In the oral-writing task, however, Wilcoxon tests revealed that the target-like category mean was significantly superior to that of the non-target-like one (4.65 vs. 1.62), distribution percentages showing the very same tendency too (75.53% vs. 24.47%).  (Table 8), it was observed that inter-task comparisons reached statistical significance only in the case of LREs resolved in a non-target-like manner, with learner dyads producing significantly more meaning-focused resolved LREs of this type in the oral task (1.57) than in the oral+written task (0.74). No differences between the two tasks were found for the meaning-focused LREs which were resolved in a target-like manner. As regards intra-task comparisons, learners did not behave alike either. The distribution of target-like and non-target-like resolved meaning-focused LREs in the oral task was quite even, with similar percentages (51.35% and 48.65%) and mean scores (1.65 and 1.57), no statistically significant differences being found between target-like and non-target-like LREs. However, the distribution of these two types of meaning-focused resolved LREs was more dissimilar in the case of the oral+written task, with 77.33% and 22.67% percentages in target-like and non-targetlike categories respectively. Besides, the difference between the means found statistical support in the Wilcoxon test, which pointed out to a significantly higher mean in targetlike than in non-target-like resolved meaning-focused LREs (2.48 vs. 0.74). As for the accuracy analyses of resolved form-focused LREs (Table 9), Wilcoxon tests comparing the two tasks revealed that both target-like and non-target like categories were significantly more productive in the oral+writing task than in the oral task, this difference being statistically more marked in the case of target-like resolved formfocused LREs. As far as intra-task comparisons are concerned, learners behaved differently in each task. Even though in both tasks, the percentage of form-focused LREs which were resolved in a target-like manner was higher than that of those resolved in a non-target-like way, the accuracy difference was found to be statistically significant only in the case of the oral+written task, with more form-focused LREs being resolved in a target-like than in a non-target-like manner (2.17 vs. 0.78).
Inferential statistics did not find accuracy differences for resolved form-focused LREs in the oral task, though. To sum up the analysis of the results as regards incidence, nature, resolution and outcome of the resolution, the speaking+writing task yielded a greater number of LREs, as well as more resolved LREs and accurate resolutions. In terms of the nature of LREs, while meaning-focused episodes were more common in the oral task, a similar rate of meaning-and form-focused LREs was obtained in the speaking+writing task.
Apart from the data on LRE accuracy previously reported, we looked into the rate of target-likeness in each task by calculating the rate of target-like meaning-focused and form-focused LREs over the total number of LREs. In the oral task, target-like meaning-focused resolved LREs represented 34.55% of the total number of LREs. In the oral-writing task, meaning-focused LREs resolved in a target-like manner accounted for 36.71% of the total number of LREs. As for the rate of target-likeness of formfocused LREs, the oral task contributed to 8.18% of LREs resolved in a target-like fashion over the total number of LREs. In the oral+writing task, target-like resolved form-focused LREs stood for 31.65% of the total number of LREs.
The last type of LRE analysis carried out involved the oral+writing task only.
We inspected whether the outcome of the LREs resolved in a target-like and non-targetlike manner by the learner dyads in their oral interaction appeared in the final written product of this task. In this regard, we distinguished between two categories of resolved LREs, namely incorporated and non-incorporated LREs. We will show the results for all resolved LREs as well as for meaning-focused LREs and form-focused LREs independently. As for all resolved LREs (Table 10), the comparison between the incorporated and non-incorporated categories turned out to be statistically significant, with a higher mean and percentage in the former than in the latter (4.87 vs. 1.30; 79.02% vs. 20.98%). When the analyses were carried out for target-like and non-targetlike resolved LREs separately, the Wilcoxon tests revealed that the superiority of the incorporated category over the non-incorporated one was statistically significant in the case of target-like resolved LREs only (3.96 vs. 0.70; 85.18% vs. 14.82%). Non-targetlike resolved LREs were more similarly distributed in terms of incorporation, with approaching means and percentages in the incorporated and non-incorporated categories. Regarding the analyses of incorporation of resolved meaning-focused LREs, the data in Table 11 shows that resolved meaning-focused LREs are incorporatedmore often than non-incorporated, with a higher mean and percentage in the former than in the latter category (2.48 vs. 0.74; 77.33% vs. 22.67%). These differences were significantly supported by the Wilcoxon test. The very same statistically supported pattern was found for target-like resolved meaning-focused LREs, with a mean of 2.00 incorporated LREs representing 81.03% of all target-like resolved meaning-focused LREs as opposed to a mean of 0.48 non-incorporated LREs representing 18.97%. Nonetheless, no differences were found between incorporated and non-incorporated categories in the non-target-like resolved meaning-focused LREs. As far as the incorporation of resolved form-focused LREs (Table 12), it was found that incorporated LREs were much more frequent than non-incorporated ones, as attested by the differences in the means (2.39 vs. 0.57) and percentages (80.88% vs. 19.12%), which the Wilcoxon tests found to be significant. When the incorporated vs. nonincorporated comparisons were made for target-like and non-target like resolved formfocused LREs independently, some differences were attested. Target-like resolved form-focused LREs were incorporated significantly more often than not (1.96 vs. 0.22; 90% vs. 10%), whereas no differences were found for non-target-like form-focused LREs in terms of incorporation. Finally, the analyses carried out to find out whether there are any differences as far as the motivation which students felt when accomplishing each task are offered in Table   13. Since motivation data were gathered at two different times, that is, immediately before and immediately after the completion of each task, a comparison of student motivation at both times was made. Results indicated that at the pre-task stage students were quite motivated to take part in the study, as evidenced by mean scores of 7.80 and 8.07 (out of a maximum of 10) in task 1 and 2, respectively. When these means were compared to the ones students gave at the post-task stage, Wilcoxon tests indicated that there were significant differences between both times in both tasks, students showing a higher motivation at the post-task than at the pre-task stage. In other words, students' engagement in the tasks resulted in their increased motivation, from 7.80 to 9.36 in the oral task and from 8.07 to 9.54 in the oral+writing task. Additionally, inferential statistics comparing the motivation means of the two tasks was performed. No statistically significant differences were found between students' motivation levels for each task either before (7.80 vs 8.07) or after (9.36 vs. 9.54) their accomplishment.

Discussion
In this section the four research questions of the study will be answered. As for the first research question (Are there any differences between tasks in terms of number, types, and outcomes of LREs?), the oral+written mode promoted a greater number of LREs, and in particular, more form-focused LREs, as well as more resolved LREs and more correctly resolved LRES, a finding in line with previous research with both adults (Adams & Ross-Feldman, 2008;García Mayo & Azkarai, 2016;Niu, 2009;Payant & Kim, 2019) and young learners (García Mayo & Imaz Agirre, 2019). Speaking+writing tasks offer learners more time to think and to reflect on their written products, so more learning opportunities operationalized as LREs emerge. Likewise, both unresolved meaning-focused LREs and non-target-like resolved meaning-focused LREs were more frequent in the speaking task (see also Payant & Kim, 2019 in this respect). The oral task is a more immediate task, with very little time for planning and editing and implies greater cognitive load than writing (see Grabowski, 2007;Granfeldt, 2007 in this respect) as: (i) in speaking, the information produced must be maintained exclusively in memory, while in writing, the already written text can be re-read; (ii) speaking is faster than writing; (iii) cognitive resources can be used for a longer period of time in writing; (iv) speaking requires continuous progress, whereas language production in writing is self-determined, the writer being able to stop the grapho-motoric process and to concentrate only on retrieval or on the planning process (Kuiken & Vedder, 2012, p. 365-366). The oral task, thus, may pose more lexical difficulties that need to be overcome by these learners in order to move the task forward. Besides, these learners' low proficiency and their young age might prevent them from resolving the meaningfocused LREs in a more target-like way. Unlike adults, these young learners might benefit even less from the availability of cognitive resources during a shorter period of time in the oral mode, which could help them solve their lexical gaps. This claim is also reinforced by the evidence available from previous studies on negotiation of meaning carried out in an ESL setting that have compared children and adults (Oliver, 1998(Oliver, , 2009).Children are able to negotiate but at different rates. Likewise, studies on the provision of feedback with young learners have reported that children interact but not in a way that promotes accuracy (Lyster, 2001).
With respect to the second research question (What are the most common types of LREs (nature and outcome) in each task?), in the speaking task, meaning-related LREs were more common, even though learners had been requested to attend to accuracy. Thus, the inherent focus of the task overrules the framing of the task (Philp, Walter, & Basturkmen, 2010). To satisfy the demands of this more immediate and communication-oriented task, they need key vocabulary to move it along (García Mayo & Azkarai, 2016;Payant & Kim, 2019;Swain & Watanabe, 2013). In terms of resolution, there were more resolved than unresolved LREs both in the case of meaningand form-related episodes. However, this task did not yield more target-like resolutions in either condition. The on-line nature of this speaking task together with its primary focus (communication of meaning) may limit opportunities for greater accuracy. In addition, when compared to adults and adolescents from other investigations (Lasito & Storch, 2013;Niu, 2009), the rate of target-likeness seems to be even lower in this task, particularly in the case of form-focused LREs. This finding could be related to the young age of the participants of the present study. In this immediate and ephemeral task under communicative pressure (Adams, 2006), these young learners have even more difficulties to rapidly retrieve and verbalize their explicit knowledge which could help them solve their linguistic gaps in a more efficient way. Negotiating successfully could entail greater effort and might need more time to develop in this age range. We cannot forget these learners are immersed in a foreign language context in which cognitive maturity is key to succeed (Tellier & Roehr-Brackin, 2017). As learners gain cognitive maturity, metalinguistic awareness increases in foreign language contexts, which could aid them in negotiating more successfully.
As for the speaking+writing task, the ratio of both lexis and form LREs was comparable, a finding which could be related to the low proficiency of the learners (García Mayo & Imaz Agirre, 2019; Leeser, 2004;Payant & Kim, 2019;Williams, 2001). These low-proficient learners seem to be in the need of key vocabulary to move the task forward (García Mayo & Imaz Agirre, 2019). But as in previous research with adult learners, they also attend to formal aspects in this task (Adams, 2006;Adams & Ross-Feldman, 2008;García Mayo & Azkarai, 2016;Niu, 2009;Payant & Kim, 2019).
In other words, the addition of a written component increases learner opportunities to focus more on grammatical aspects (Adams, 2006). Likewise, this task yielded more target-like episodes both in the case of meaning-and form-related episodes, supporting previous research with adult learners (Adams, 2006;Adams & Ross-Feldman, 2008;García Mayo & Azkarai, 2016;Niu, 2009;Payant & Kim, 2019). Thanks to the extra processing time, learners can notice form and increase accuracy (Payant & Kim, 2019).
In these conditions, young learners have more time to draw upon their explicit knowledge (Ellis, 2003) as adults do. Nevertheless, even if the rate of target-likeness in form-focused LREs is higher in the speaking+writing task, LREs are not so elaborated since learners usually resolve them by providing the relevant form without further explanations or justifications as observed in (11) and (12) (see Niu, 2009 for the classification of LREs in terms of elaboration in adult learners): (11) *CHI2: the dog go. *CHI1: goes. *CHI2: goes. %sit: CHI2 continues writing. *CHI2: eh first to the dental (.) clinic.
The fact that the resolutions of LREs were not so elaborated as adults' could be explained by child learners' still developing metalinguistic awareness (Muñoz, 2017b).
Additionally, we cannot dismiss the fact that primary education in Spain is characterized by a strong oral component and a marked emphasis on vocabulary (Muñoz, 2017 b).
As an answer to the third research question (How does the resolution of LREs affect the written product in the speaking+writing task?), those episodes which were resolved in a target-like way were incorporated in the written product. Thus, language discussions geared towards accuracy do have a reflection in the final written product.
This result is not fully comparable to previous findings, as to our knowledge, no investigations have been conducted analysing the relationship between the resolution of LREs and the written output of collaborative tasks. However, the finding obtained in the present study is in line with prior research that has established a relationship between accurate resolutions and language development in the posttests administered (Basterrechea & García Mayo, 2013;Payant & Kim, 2019). Even if more research along the same lines is timely, the evidence reported in the present study confirms the great potential of this speaking+writing task as far asthe contribution of LREs to language learning.
As regards the last research question (Are there any differences between the two tasks in terms of student motivation?), the students were quite motivated before the completion of the tasks and their degree of motivation increased even further as a result of their engagement in the tasks. Likewise, no motivation differences were found between the two tasks neither at the pre-nor the post-task stage. These findings indicate that the type of tasks designed in our research were taken seriously by the children who participated in the study, perhaps because they perceived a connection between them and other educational activities they could be engaged in while in class (Mackey & Gass, 2005, p. 107).Both tasks turned out to be clearly attractive to young English learners, so teachers may be amenable to include them in primary education classrooms, more particularly the oral+written task given its superiority in terms of language learning opportunities, as attested by the LRE findings both in our study and in previous literature. As claimed by Hunt et al. (2005; as cited in Shak & Gardner, 2008), '[c]hildren will only persist in learning tasks if they see them as worthwhile' (p. 374).

Conclusions
This paper has contributed to the literature regarding the effect of task-modality on the production of LREs among young learners, a line of research still in its infancy.
Although the two task modalities examined in this study were equally motivating to students, the examination of the LRE data indicated that task-modality had a strong effect on the occurrence and resolution of LREs, as a greater number of LREs and more resolved LREs were obtained in the speaking+writing task, a finding in line with previous research with adults. Although learners had been encouraged to focus on accuracy in the speaking task, the inherent focus of this task overruled the framing of the task. However, the nature of LREs was partially mediated by task-modality, as even if the speaking task promoted more meaning-focused LREs, the speaking+writing task yielded a similar rate of meaning-and form-focused LREs, a finding which could be ascribed to the low-proficiency of these young learners. In addition, unlike adults, the rate of target-like episodes was lower in the speaking task. Likewise, the type of resolutions of LREs attested in these young learners does not match the ones observed in the literature with adult learners (Niu, 2009), which are more elaborated. Thus, the enhancement of metalinguistic awareness through appropriate form-focused training conditions could lead to extended negotiation of form among young learners (see Bouffard & Sakar, 2008 for an example along these lines), and by implication, greater development of language accuracy in these minimal input contexts (Tellier & Roehr-Brackin, 2017). All in all, the present study supports the superiority of the speaking+writing task in the promotion of learning opportunities for young learners, a result reinforced by the incorporation of target-like resolved episodes in the written product.
For future research, it would be convenient to investigate the effect of a wider range of tasks that could draw learners' attention to form more extensively by promoting the use of metalanguage and the verbalization of rules among learners. We also acknowledge thatthe type of tasks used in the speaking and in the speaking+writing modality also differ in type. The speaking task is a storytelling task, whereas the speaking+writing task is an opinion gap task. Follow-up studies should control for the nature of both tasks. Similarly, the investigation of the combination of different variables (i.e., age, proficiency, gender) and type of pairings would be advisable so as to offer learners the best learning conditions in this age period. Future studies should also consider the inclusion of tailor-made tests to measure actual learning gains.

Credit author statement
Both authors have participated in the conceptualization, design, analysis, writing, and revision of the manuscript.