The interface between task-modality and the use of previously known languages in young CLIL English learners

This article contributes to the scarcity of research on the interface between task-modality and the use of previously known languages (PKL) in young learners. It examines the use of Basque/Spanish by CLIL learners (aged 10-11) during oral interaction while completing two collaborative tasks in English: a speaking task and a speaking + writing task. Findings indicate that these learners are extensive users of their PKL. Task-modality is particularly evident in the case of amount of PKL use, as a higher number of PKL turns are obtained in the speaking + writing task. However, task-modality has a limited effect on the functions of PKL, which contrasts with previous studies with adults. Despite the extensive use of their PKL, these young and low-proficient learners employ them as cognitive tools that facilitate the organization of the tasks, the co-construction of meaning and the attention to formal aspects of language such as mechanics.


Introduction
The use of previously known languages (henceforth PKL) during task-based interaction has been the focus of attention of numerous English as a second language (ESL) and English as a foreign language (EFL) investigations, predominantly with adult learners. These studies have tackled the effect of factors such as proficiency, onset age, gender, task-type (Alegría de la Colina & García Mayo, 2009;Cenoz, 2001;Muñoz, 2007), and to a lesser extent, task-modality effects (Azkarai & García Mayo, 2015, 2017Payant & Kim, 2019). In this last respect, when comparing collaborative speaking tasks to tasks that include oral and written components, learners tend to resort to their PKL to a higher extent in the latter. In terms of functions, while learners have been found to use their PKL mainly for vocabulary searches in speaking tasks, in speaking + writing tasks, grammar talk is more common.
Recent years have witnessed the implementation of Content and Language Integrated Learning (CLIL) programs all over Europe, and with their inception, an explosion of research examining the effect of the more natural, meaningful and intense input provided in these learning contexts. Specifically, CLIL studies on the use of PKL have mainly concentrated on secondary-school learners (Arratibel-Irazusta & Martínez-Adrián, 2018, 2019Lázaro-Ibarrola, 2016;Lázaro-Ibarrola & García Mayo, 2012;Martínez-Adrián & Gutiérrez-Mangado, 2015), but a growing bulk of research exists with primary-school learners (Azkarai & Imaz Agirre, 2017;Gallardo-del-Puerto, 2015;García Mayo & Hidalgo, 2017;García Mayo & Lázaro-Ibarrola, 2015;Martínez-Adrián, 2020a). However, unlike research with adults in ESL and EFL settings, CLIL research has mainly analyzed the variable proficiency and there is a need to examine a wider range of variables such as task-type or task-modality, particularly with young learners. This paper will try to fill these gaps by examining the use of PKL (Basque and Spanish) during the oral interaction of young CLIL learners (aged 10-11) while performing two collaborative tasks: a speaking task and a speaking + writing task. Section 2 reviews the main empirical findings concerning the use of PKL during dyadic interaction and task-modality effects. In section 3, the research questions are formulated and the study is described. Section 4 presents the results. The paper finishes with the discussion of the results and the main conclusions drawn from the study.

The use of PKL when performing communicative tasks in English
The use of PKL when performing communicative tasks in English has been the focus of attention of numerous investigations in ESL and EFL contexts, particularly with adult learners. These investigations have tackled the effects of factors such as proficiency, onset age, gender and task-type (Alegría de la Colina & García Mayo, 2009;Cenoz, 2001;Muñoz, 2007).
As for proficiency, low-proficiency learners have been found to make use of PKL to a higher extent (Di Camilla & Antón, 2012;Storch & Aldosari, 2010;Swain & Lapkin, 2000). In terms of functions, low-proficient learners use their PKL for task management and high-proficient learners to discuss vocabulary searches (Di Camilla & Antón, 2012).
Regarding onset age, older learners have been reported to make a greater use of PKL during interaction (Cenoz, 2001(Cenoz, , 2003. With respect to gender effects, while Ross-Feldman (2005) did not observe differences between males and females in terms of amount of PKL use, Azkarai (2015) reported a higher use of the first language (L1) by females. When functions were explored, females used the L1 mainly for a phatic function (i.e., to overcome communication breakdowns), while males employed their L1 mainly for off-task and vocabulary issues.
Task-type has also been claimed to affect the use and functions of the L1 (Alegría de la Colina & García Mayo, 2009;Lasito & Storch, 2013;Rayati, Yaqubi, & Harsejsani, 2012;Storch & Aldosari, 2010;Storch & Wigglesworth, 2003;Swain & Lapkin, 2000). The participants in these investigations did not make an excessive use of PKL, which has led researchers to claim that PKL should not be banned during pair-work, as its use might aid in the language learning process. In terms of amount of PKL, tasks with a written component generate a higher use of PKL. In the case of functions, learners employ their PKL for a variety of purposes such as task management, vocabulary and grammar discussions, off-task and phatics, among others, and they differ from task to task. For example, while the use of PKL for vocabulary discussions is more common in tasks that focus on oral communication (Lasito & Storch, 2013), grammar talk and mechanics are more frequent in editing tasks (Rayati et al., 2012;Storch & Aldosari, 2010).
Recent years have witnessed a growing body of research with young learners (Enever, 2011, 2018Haselgreen, Drew, & Sørheim, 2012;Nikolov, 2009). However, studies examining the use of PKL by younger learners in ESL and EFL contexts (Pinter, 2007;Shintani, 2012;Tognini & Oliver, 2012) do not examine the functions of PKL in detail like studies conducted with adult learners (see Azkarai & García Mayo, 2017 for a review of these studies in this respect), and a call for more research with younger learners has been made. Children are unique and engage in the language learning process in ways that are distinct from their older counterparts, which reinforces the specificity of Second Language Acquisition (SLA) research with this population (Mackey & Gass, 2005 as cited in Oliver & Azkarai, 2017). Following Piaget's theory (1929), "cognitive development is gradual, initially localized (see, e.g., Gombert, 1992), and involves the construction of increasingly sophisticated mental representations (see, e.g., Bialystok, 1994Bialystok, , 2001Bialystok & Ryan, 1985)" (Roehr-Brackin, 2018, p. 10). Likewise, "language is an integrated part of general cognition that develops from localized, item-based to more abstract and general representations" (Roehr-Brackin, 2018, p. 10).
The need for more research with young learners has been particularly manifested in EFL contexts, usually considered low input contexts (Pinter, 2011), where learners are not exposed to large amounts of input as in ESL settings. The programs introducing foreign languages at an early age have proliferated in recent years, in particular CLIL programs in which in addition to English as a school subject, different subjects are taught through a foreign language. Despite the great diversity of CLIL programs, all of them are characterized by the provision of more natural, meaningful and intense input than in mainstream EFL classrooms (Coyle, 2007;Lázaro-Ibarrola & García Mayo, 2012;Marsh, 2002;Muñoz, 2007). Taking into account that CLIL programs are the norm more than the exception in the case of young learners in some European countries such as Spain, several researchers have pointed out the need to do research in these settings so as to provide learners with the best learning conditions (García Mayo, 2018).
Most research in CLIL contexts on the use of PKL during oral production has concentrated on secondary education (Arratibel-Irazusta & Martínez-Adrián, 2018, 2019Lázaro-Ibarrola, 2016;Lázaro-Ibarrola & García Mayo, 2012;Martínez-Adrián & Gutiérrez-Mangado, 2015), even though a growing interest in the study of primary-school learners has been observed in recent years (e.g., Azkarai & Imaz Agirre, 2017;Gallardo-del-Puerto, 2015;García Mayo & Hidalgo, 2017;García Mayo & Lázaro-Ibarrola, 2015;Martínez-Adrián, 2020a). The general finding that emerges from these studies is that CLIL learners do not rely so much on their PKL since the exposure to more intense and meaningful input provided in CLIL lessons leads them to a greater command of the target language (TL) and in turn to a decrease in the use of the PKL. These studies also confirm the facilitative role of PKL for task completion (García Mayo & Hidalgo, 2017;Martínez-Adrián, 2020a).
As for proficiency effects, results are somewhat contradictory. While Lázaro-Ibarrola and García  found that in secondary school learners the use of the PKL in appeals for assistance and discourse markers significantly decreased with proficiency growth, other studies with similar age learners such as Arratibel-Irazusta and Martínez-Adrián (2018, 2019) revealed not only the inexistence of a significant decrease of PKL use (interactional strategies, transfer lapses -that is, borrowings and foreignizings -code-switching and discourse markers), but also of TL-based strategies (i.e., analytic and holistic). Other investigations conducted with primary school learners that offer a comparison of CLIL and mainstream EFL learner groups (matched for hours of exposure) following a longitudinal perspective attested a decreasing tendency in the use of categories such as borrowings and code-switching in both groups (Pladevall-Ballester & Vraciu, 2017), a finding contrasting with other investigations in which a lower use of PKL was found on the part of CLIL learners. More recent studies conducted with young learners have examined the use of PKL together with TL use in appeals for assistance, clarification requests, metacomments, discourse markers and private speech in different age/proficiency groups (Martínez-Adrián, 2020b). While a greater use of PKL was revealed in the group of older and more proficient learners, especially in the case of those strategies that were less cooperative and more external to the task itself, both age/proficiency groups performed similarly in the case of TL use, except for metacomments, where a slightly higher use was observed in younger learners. Intragroup analyses also showed that learners in both age/proficiency groups resorted to their PKL more frequently than to English, except for metacomments. In terms of types, metacomments, followed by discourse markers and appeals for assistance, were the most common manifestations of PKL in both groups. Metacomments were also most commonly served by the use of TL English.
Other studies have investigated the effect of task-repetition and gender in young learners. Azkarai and García Mayo (2017) explored the effect of two types of task repetition (exact and procedural) at two testing times. L1 use decreased at the second testing time in both repetition conditions. Appeals, borrowings and metacognitive uses were the most common functions including L1 use of the learners at both testing times. Azkarai and Imaz Agirre (2017) examined the effect of gender on L1 use in two age groups of primary-school learners while completing a spot-the-difference task. Boys were reported to employ the L1 more often than girls, especially the older ones, and mainly to overcome communication breakdowns. In contrast, girls used the L1 only on occasions where they considered it necessary.
The overview of research carried out with CLIL learners has revealed that the vast majority of studies have explored the effect of proficiency and has uncovered the need to examine other variables, especially with young learners. The present study will try to add to the literature in this respect.

Task-modality effects and the use of PKL
Research on dyadic task-based interaction has shown that more immediate tasks such as collaborative speaking tasks draw learners' attention to meaning while tasks that incorporate a written component promote more attention to formal linguistic aspects and demand higher levels of accuracy because of the greater planning time and more opportunities for editing they offer to the learner (Adams, 2006;Adams & Ross-Feldman, 2008;García Mayo & Azkarai, 2016). In other words, a modality effect has been obtained in several investigations as learners have been shown to produce and resolve more language related episodes (LREs) 1 when comparing speaking to speaking + writing tasks. Nevertheless, the vast majority of these studies have been conducted with adults in ESL contexts (Adams, 2006;Adams & Ross-Feldman, 2008;Ross-Feldman, 2007) and EFL settings (Azkarai & García Mayo, 2012;García Mayo & Azkarai, 2016;Niu, 2009;Payant & Kim, 2019), and research carried out with younger learners, particularly in EFL settings, is in its infancy (García Mayo & Imaz Agirre, 2019). In line with research conducted with adults, more LREs were produced by 6th year primary-school learners in the speaking + writing modality in García Mayo and Imaz Agirre (2019). As for the nature of LREs, lexical LREs were more frequent than form in both the speaking and the speaking + writing task. It seems as if children were in the need of producing more lexical LREs to move both tasks along. In terms of outcome, a higher percentage of resolved LREs was found in the speaking + writing task, supporting previous research with adults.
As regards the interface between task-modality and the use of PKL, to the knowledge of the authors, just two studies with adult learners in an EFL setting have been conducted in this respect (cf. Azkarai & García Mayo, 2015;Payant & Kim, 2019). Azkarai and García Mayo (2015) gathered data from 44 EFL Spanish learners at university. Participants paired up in same-proficiency dyads 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. All L1 turns were codified according to L1 or second language (L2) predominance 2 and the functions that the L1 served on the basis of the categorization adopted by Alegría de la Colina and García Mayo (2009), and Storch and Aldosari (2010): off-task, metacognitive talk, grammar talk, vocabulary, and phatics. The analysis of the results indicated that learners made minimal use of the L1, as also reported in other investigations with adult learners (Storch & Aldosari, 2010;Swain & Lapkin, 2000), which supports the claim that pair work provides learners with many opportunities to use their L2. In terms of the nature of turns, this study showed a greater production of minor L1 turns, which contrasts with Storch and Aldosari (2010). In particular, minor turns were more common in the functions of vocabulary and phatics, while predominant L1 turns were more frequent in off-task, metacognitive and grammar talk. As for task-modality effects in L1 use and its functions, more L1 turns were produced in speaking + writing tasks than in speaking tasks. When learners worked in the speaking + writing tasks, offtask, metacognitive talk, grammar talk and phatics were more common. In speaking tasks, learners employed the L1 for vocabulary searches. Thus, these results align with previous findings in which L1 use was task dependent (Alegría de la Colina & García Mayo, 2009;Storch & Aldosari, 2010;Storch & Wigglesworth, 2003). Azkarai and García Mayo (2015) also compared same-modality tasks. Greater reliance on the L1 was observed in text editing and picture placement than in their modality counterparts. The authors ascribed greater use of the L1 in these tasks to the complexity of these tasks. As regards the functions of L1 use, off-task and metacognitive talk were more frequent in dictogloss and picture placement tasks than in their modality counterparts, while grammar talk and vocabulary searches were more frequent in the text edition and picture differences than same modality tasks.
Payant and Kim (2019) examined the impact of task modality on learners' use of the L1 during the production of LREs. To this end, five dyads of L1 Spanish-L2 English bilingual learners of L3 French performed two decision-making tasks, both of them including oral and written components. Form-based and lexis-based LREs were codified for each language pattern and for each modality. As in Azkarai and García Mayo (2015), a greater number of LREs produced in the L1 was obtained in the written modality. However, learners preferred to use the TL French in the oral modality. The same language pattern applied to the types of LREs produced.
Given the scarcity of research along these lines, a call has been made for more studies with young learners and in other instructional settings such as CLIL (Azkarai & García Mayo, 2015;Azkarai & García Mayo, 2017). Thus, this article will try to fill these gaps by examining the use of PKL (Basque and Spanish) by young CLIL learners (aged 10-11) during oral interaction while completing two collaborative tasks: a speaking task and a speaking + writing task.

Research questions
Based on previous findings regarding the use of PKL in young CLIL learners, the present study analyzes the nature of PKL turns and the existence of task-modality effects (speaking vs. speaking + writing task). More specifically, we address the following research questions: 1. Are there any differences in terms of amount between predominant and minor PKL-based turns? 2. Which functions do PKL-based predominant and minor turns serve? 3. Are there any differences between tasks (speaking vs. speaking + writing) with respect to the production of PKL turns and their functions? 4. What functions are more commonly served by PKL in each task?

Participants
The present study, which is part of a larger project, was conducted in a state school in the Basque Country in northern Spain. The participants in the study are 50 (15 females and 35 males) learners of third language (L3) English in the 5th and 6th year (ages 10-11) of primary education immersed in a CLIL program. The vast majority of the students come from Spanish-speaking families as this language is the majority language in the area where these learners live. Nevertheless, all of them are enrolled at school in Model D, in which the vast majority of subjects are taught in Basque except for the Spanish language and its literature, which is taught in Spanish, English as a language subject, and content subjects such as science, arts and crafts or physical education, which are taught in English. As Lasagabaster (2007) mentions, this model can be seen as an early total immersion program which allows students from Spanish-speaking families to obtain a high command in L2 Basque, so that they are considered Spanish/Basque bilinguals (Cenoz, 2009). The context in which the participants live has been defined as additive trilingualism where Basque, the language of instruction, is a minority language of Spain, Spanish is the majority language, and English is the main foreign language (Cenoz & Valencia, 1994).
All the participants started learning English in pre-primary education as a school subject. Since the 3rd year, English is also used as a vehicular language for some content subjects as mentioned above. In the 5th and 6th years, in which learners in the present study are enrolled, students receive 3 hours a week of instruction in EFL and 2 to 4 hours a week of CLIL instruction. Therefore, learners receive 5 to 7 weekly hours of instruction in the TL. At the moment of data gathering, 5th year learners had received from 777 hours of exposure to English, and 6th graders 962 hours. As far as their English proficiency is concerned, all the students were considered beginner (A1-A2) 3 learners according to the scores they obtained in the Key English Test (KET), administered at the outset of the study (UCLES, 2014). 4

Instruments and data analysis
Data were collected by means of four different instruments which are part of a wider project on the use of collaborative tasks among primary-school children. Participants were first tested on general proficiency by means of the KET, which was completed during classtime. Subsequently, they were matched in pairs on the basis of their proficiency scores so as to perform a speaking and a speaking + writing task. Previous studies analyzing task-modality effects both on the production of LREs and on PKL use have employed similar tasks (Adams & Ross-Feldman, 2008;Azkarai & García Mayo, 2015;García Mayo & Azkarai, 2016;García Mayo & Imaz Agirre, 2019;Niu, 2009;Payant & Kim, 2019). Nevertheless, taking into account that the framing of the task could overrule its inherent focus (Philp, Walter, & Basturkmen, 2010), we decided to ask participants in both tasks to pay special attention to accuracy. Likewise, they were asked to carry out the task naturally by employing all their resources at hand. If doubts or difficulties concerning vocabulary issues emerged during task completion, they could always seek their partner's help but not the investigator's help. In each of these tasks, and prior to their completion, participants were individually asked to grade their motivation using a motivation thermometer. Specifically, they were requested to tell from 0 to 10 how they felt before and after doing each of the tasks and to select one of the reasons provided by the investigator to justify their motivation grade.
With respect to the speaking task, each dyad had to order a set of 6 pictures taken from the book Sparks 1 (House & Scott, 2009) 5 and then, in turns, they had to narrate the episodes depicted in them in story mode. The vignettes showed a girl who was provided with the picture of a rag-doll she had to sew out of the set of pieces given. The girl starts sewing but the rag-doll does not match the one shown in the picture, as the legs were sewn where the arms should be and the arms where the legs should be. When she looks at the result, she starts crying. Her friend, the fairy, sees that the girl is very upset and decides to help her by sprinkling some magic dust on the doll. Suddenly, the arms and the legs are sewn in the right positions, which makes the girl really happy. This type of task has been widely employed in several investigations with both children 4 The fact that they are still beginner learners despite the additional exposure received since grade 3 aligns with the research findings reported in other CLIL studies that have claimed that a later start in CLIL could optimize resources (see Muñoz, 2015, for a review of these findings). 5 Note that this book is not used in this school. and adults for similar research purposes in both non-CLIL and CLIL settings (e.g., Alegría de la Colina & García Mayo, 2009;Arratibel-Irazusta & Martínez-Adrián, 2018, 2019García Mayo & Hidalgo, 2017;García Mayo & Lázaro-Ibarrola, 2015;Martínez-Adrián, 2020a;Storch & Aldosari, 2010).
For the speaking + writing task, learners had to examine some pictures provided by the investigator in which a boy in a park had found a lost dog. With the help of some clues, they had to decide who the owner of the dog was and why they thought so. Then, they had to write down a short note for the boy explaining who the owner of the dog was, why they thought so, and also giving instructions on how to take the dog back to its owner. Similar tasks have been administered in previous studies with EFL adults (Azkarai & García Mayo, 2015) and more recently with child CLIL learners (García Mayo & Imaz Agirre, 2019). 6 7 Even if similar tasks have been used already with child learners in other investigations, the researchers held several meetings with the school teachers so as to ensure that the task was appropriate for the learners tested and to discuss possible adaptations. It was also pilot-tested with similar-age children so as to detect potential problems that could emerge during the administration of the tasks.
All the tasks were audio and videotaped, transcribed and codified into CHILDES (McWhinney, 2000). All turns that included the use of PKL (i.e., Basque and Spanish) were identified. These turns were classified according to language predominance, that is, whether they were predominant (¿esto es una chica o un chico? -"Is this a girl or a boy") or minor (the owner is Jack Smith the vet punto -"the owner is Jack Smith the vet, full stop") PKL-based turns. Subsequently, they were subdivided according to the different functions they served on the basis of Azkarai and García Mayo (2015) and Storch and Aldosari (2010), namely off-task, metacognitive talk, grammar talk, vocabulary, phatics, and mechanics. The description provided below contains examples of our database: 6 As Azkarai and García  indicate, even if the interface between task-based language teaching and CLIL is still in its infancy, we can observe the use of similar tasks to the ones employed in EFL in recent studies with CLIL children (García Mayo & Hidalgo, 2017;García Mayo & Lázaro Ibarrola, 2015). 7 We are aware, as pointed out by one of the reviewers, that the tasks used in the speaking and in the speaking + writing modality also differ in type. The speaking task is a storytelling task, while the speaking + writing task is an opinion gap task. Ideally, both conditions should have kept constant the level of complexity between tasks. However, this study is part of a bigger project on task-modality in which we have extended prior research with adults to children by designing similar tasks (see García Mayo & Azkarai, 2016 with adults and García Mayo & Imaz Agirre, 2019 with children). But so as to solve the limitation of previous studies as regards the lack of control of accuracy that both modalities demand, we kept instructions constant by asking learners to attend to accuracy in both modalities.

Off-task
Off-task refers to those situations in which participants use their PKL as casual talk that is not related to the task (Alegría de la Colina & García Mayo, 2009;Azkarai & García Mayo, 2015), as illustrated in (1): (1) CHI1: Que feo, mira que cara tiene el perro, parece un animatronic.
[How ugly, look at the dog's face, it looks like an 'animatronic']

Metacognitive talk
Metacognitive talk, as can be observed in (2), includes instances in which the students use their PKL to talk about the task itself in order to plan, organize and monitor the activity, among other functions (Alegría de la Colina & García Mayo, 2009;Azkarai & García Mayo, 2015). (

Grammar talk
Grammar talk involves cases in which learners use their PKL to discuss issues related to grammar such as morphosyntax and text structure (Azkarai & García Mayo, 2015;Storch & Aldosari, 2010), as showed in (3)

Phatics
Phatics, as illustrated in (5), refers to expressions such as ok, well, so produced in Basque and Spanish to establish social contact and to express sociability rather than specific meaning (Azkarai & García Mayo, 2015): (5) CHI1: and friend is very o sea girl is very happy because the doll is pretty.
[and friend is very I mean girl is very happy because the doll is pretty]

Mechanics
Mechanics involves instances in which participants use their PKL to discuss punctuation, spelling and pronunciation (Storch & Aldosari, 2010), as in (6) With respect to statistical analyses, results were analyzed by means of SPSS 24 (IMB Corp., 2010). Both descriptive and inferential analyses were computed. In the case of descriptive analyses, means, medians, standard deviations, minimum and maximum scores were calculated. Kolmogorov-Smirnov tests were run to verify the normality of distribution of the samples. As the data did not meet the criteria for normal distribution, Wilcoxon's Signed Rank Tests and Friedman Tests were performed for intertask and intratask comparisons.

Results
So as to answer the first research question, all turns were tallied and those that included the use of PKL were identified. Subsequently, they were subdivided into predominant and minor turns. In addition, so as to enrich the analysis, the total number of words produced were calculated and the number of PKL words subtracted from this word count (see Table 1). This analysis indicated that from the total count of turns (3453 turns), 37.91% (1309 turns) included use of students' PKL. Focusing on the total number of words produced by the participants (13511 words), the same trend that was observed for turns was found as 37.63% of those words (5084 words) were uttered in Basque and/or in Spanish. When subdividing the turns that presented use of PKL into predominant or minor turns, 87.39% of those turns (1144 turns) were predominant whereas only 12.61% of the turns (165 turns) were minor. A Wilcoxon's Signed Rank Test revealed that the difference between predominant and minor PKL turns was statistically significant (Z = -4.061, p = .001). The second question inquired into the main functions that predominant and minor turns served. When examining the mean number of turns in Table 2, predominant turns were significantly more frequent in off-task, metacognitive talk, mechanics and vocabulary. No statistically significant differences emerged in grammar talk and in phatics. In order to analyze these results in more detail, the main functions that predominant and minor turns served in each of the tasks were examined. For the speaking task, as can be observed by looking at the means in Table 3, predominant turns were significantly more frequently used in off-task, metacognitive talk and vocabulary. No statistically significant differences were revealed between predominant and minor turns in grammar talk, phatics and mechanics. In the speaking + writing task, as can be observed in Table 4, predominant turns were significantly more common in off-task, metacognitive talk, vocabulary and mechanics. However, the comparison between predominant and minor turns did not yield statistically significant differences in grammar talk and phatics. With respect to the third research question that looked into task modality effects, differences between the speaking and the speaking + writing tasks regarding the production of PKL turns and their functions were examined. As in the case of the first research question, the total number of words produced as well as the words produced in their PKL have been considered. When analyzing the turns, as can be observed in Table 5, the speaking + writing task yielded a greater number of turns and words when compared to the speaking task. Participants also produced significantly more turns and words in their PKL in the speaking + writing task than in the speaking task. In addition, predominant PKL turns were significantly more frequent in the speaking + writing task, and minor PKL turns in the speaking task.
When comparing tasks in terms of amount of PKL in each function (see Table 6), metacognitive talk and mechanics were significantly more frequent in the speaking + writing task. In contrast, off-task, grammar talk, vocabulary and phatics were equally served by participants' PKL in both tasks, as no statistically significant differences were observed. When considering the nature of PKL turns and their functions between tasks, predominant PKL turns serving the function of metacognitive talk, grammar talk and mechanics yielded a statistically significant difference in favor of the speaking + writing task, as illustrated in Table 7. On the other hand, no differences were obtained in off-task, vocabulary and phatics predominant turns. In the case of minor PKL turns, as shown in Table 8, vocabulary was significantly more frequently used in the speaking task. In contrast, mechanics presented statistically significant differences in favor of the speaking + writing task. In the case of off-task, metacognitive talk, grammar talk, and phatics no differences were obtained. Finally, the last research question pertained to the functions more commonly served by participants' PKL in each task. Figure 1 depicts the descriptive comparison of the use of Spanish/Basque in the different categories in the speaking task. As shown in Figure 1, vocabulary and metacognitive talk were more commonly served by the use of Basque/Spanish, followed by phatics and off-task. In contrast, no instances of grammar and mechanics were found in this task. The Friedman test performed reported the existence of statistically significant differences (chi-square = 85.572, p = .001). Consequently, post-hoc analyses were conducted. The Wilcoxon's Signed Rank Test (see Table 9) confirmed the inexistence 10,21 9,88 0,88 0,29 0 0 0 2 4 6 8 10 of significant differences between vocabulary and metacognitive talk. But the differences between these two and the rest of the categories examined reached significance. Statistically significant differences also emerged among off-task, grammar talk, phatics, and mechanics, except for the contrast between mechanics and grammar. Figure 2 shows the results for the speaking + writing task. Metacognitive talk was the most common manifestation of Basque/Spanish use, followed by vocabulary and mechanics. Grammar talk, even if it is present in this task, is the least common category.

Figure 2
Functions more commonly served by participants' PKL in the speaking + writing task The Friedman test conducted revealed the existence of statistically significant differences among the categories (chi-square = 80.857, p = .001). Post-hoc analyses (see Table 10) indicated that metacognitive talk significantly differed from the rest of the categories. Similarly, vocabulary and mechanics presented significant differences with respect to off-task, grammar talk and phatics. No differences were reported for the contrast between grammar talk, phatics and off-task.

Discussion and conclusion
In this section the four research questions of the study will be answered. As regards the first research question (Are there any differences in terms of amount between predominant and minor turns?), unlike other investigations with adults (Azkarai & García Mayo, 2015;Storch & Aldosari, 2010), the child learners of the present study made greater use of their PKL. According to Storch and Aldosari (2010), they could be considered extensive users of their PKL as PKL words accounted for 37.63% of the total number of words, and over 37% of the turns were in the PKL. Their low proficiency probably explains the widespread use of predominant PKL turns, even though other factors could also be explored. One of these potential factors is related to the instructions. Learners were asked to perform these tasks naturally by using all their resources at hand. Therefore, when they used their PKL during task performance, they were not instructed to switch into English. In addition, learners were not so familiar with this type of tasks, which could have prompted them to rely more on their PKL. Likewise, pair dynamics or personality traits, which are beyond the scope of the present paper, could also have impacted the results.
With respect to the second research question (Which functions do predominant and minor turns serve?), when considering all tasks, predominant PKL turns were significantly more frequent in off-task, metacognitive talk, mechanics, and vocabulary. This applies to each individual task, with the exception of mechanics, which was inexistent in the speaking task. Learners tend to perform these functions by means of turns entirely produced in their PKL as in (7-10). In off-task (7), learners feel free to talk in their PKL about issues unrelated to the task. In metacognitive talk (8), learners mainly employed their PKL when they did not understand the instructions of the tasks well, and also to meet task requirements or to organize themselves. Likewise, when discussing vocabulary (9) or mechanics issues (10), they employed much more predominant turns to clear up their doubts or to make their decisions as regards word meaning/choice more effectively so as to prevent communication breakdowns: (10) *CHI2: works (.) erre ka ese. [works, 'r', 'k', 's'] This result partially aligns with Azkarai and García Mayo (2015) as they also found that predominant turns were more frequent in off-task and metacognitive talk. However, in their study, vocabulary searches were more common in the case of minor turns. In those turns, according to Azkarai and García Mayo (2015), the university learners in their study just referred to the word they needed in English. In the current study, the fact that the participants were even lower proficiency learners than in Azkarai and García Mayo (2015) and that they were child learners could explain the use of predominant turns for vocabulary deliberations. These learners were primarily making use of appeals for assistance during vocabulary deliberations, which they were producing entirely in Spanish most of the time. Appeals for assistance are strategies widely used by young learners in recent investigations carried out with children and oral production (Azkarai & García Mayo, 2017;García Mayo & Hidalgo, 2017;Martínez-Adrián, 2020a), and also with children and self-reported opinions (Martínez-Adrián, Gallardo-del-Puerto, & Basterrechea, 2019).
As an answer to the third research question (Are there any differences between tasks [speaking vs. speaking + writing] with respect to the production of PKL turns and their functions?), task-modality effects were evident in the case of the production of PKL turns. A greater number of PKL turns were produced in the speaking + writing task, as well as a higher number of PKL words. These findings align with previous research on task-modality effects conducted with adults (Azkarai & García Mayo, 2015;Payant & Kim, 2019) as well as those studies analyzing the effect of task-type (Alegría de la Colina & García Mayo, 2009;Storch & Aldosari, 2010;Storch & Wigglesworth, 2003), according to which the addition of a writing component generates more use of PKL than oral communicative tasks. As for the effect of task-modality on the functions of PKL, the demands of PKL for the different functions are similar in both tasks, except for metacognitive talk and mechanics, which yielded significant differences in favor of the speaking + writing task. This could be explained by the demands of the speaking + writing task. The fact that in this task learners talked to each other during task completion and also had to submit a written product could entail a greater investment of time in talking about task procedures. In addition, this finding could be due to the differences between tasks in terms of task-complexity reported in the methodology. Unlike the speaking task, in which they simply had to order the pictures, in the speaking + writing task they had to decide not only the owner of the dog but also the best way to take it to its owner. These very low-proficient learners, when confronted with a more complex task, could have spent more time talking about the focus of the task and reaching a joint understanding of task requirements. We acknowledge that this limitation in task design could have impacted the results and that future interventions should consider keeping the same level of complexity between tasks constant. Similarly, mechanics was inexistent in the speaking task as no episodes emerged in the speaking task during which learners could have talked about pronunciation issues. However, this category was productive in the speaking + writing task when discussing spelling and punctuation. That is, unlike in the speaking + writing task, in the more immediate task, they did not spend time on employing the PKL to discuss more language related issues in spite of having been requested to attend to accuracy as well as to content. They were more interested in meeting the communicative demands of the speaking task (Niu, 2009;Payant & Kim, 2019). This result is in line with Storch and Aldosari (2010), who found that mechanics was more common in the low-low proficiency pairs, particularly in the editing task.
These results are in contrast to Azkarai and García Mayo (2015) in that greater differences were observed in their study. For example, in their study, learners resorted to the L1 to a higher extent to discuss grammar issues in the speaking + writing task. In the present study, grammar talk in the PKL was practically non-existent in the data. However, the same child learners have been found to produce more grammar LREs in the speaking + writing task in other studies that are being currently conducted (Martínez-Adrián, Gallardo-del-Puerto, & Arratibel-Irazusta, 2019). However, when examining these grammar episodes in detail, the tendency in these learners is to resolve them by simply providing the relevant form without further elaborations in the form of justifications for the choice made (see Niu, 2009, for the classification of LREs in terms of elaborations). The fact that they are young and beginner learners could explain the low number of metalinguistic explanations either in the PKL or in the TL. Cognitive and personal traits are already fully developed in the case of adults. In contrast, young learners' metalanguage awareness skills are developing, and scaffolding each other's language might be challenging and costly, which could account for their difficulties in verbalizing rules (Muñoz, 2017;Tellier & Roehr-Bracking, 2017). Likewise, primary education in Spain is marked by a strong oral component and a special emphasis on vocabulary (Muñoz, 2017).
In addition, unlike Azkarai and García Mayo (2015), who found a greater reliance on the L1 for vocabulary purposes in the speaking task, in the present study, the younger learners employed the PKL to a similar extent for this function in both tasks. Note that in studies on LREs currently conducted with the same learners as well as in previous studies with similar-age learners (García Mayo & Imaz Agirre, 2019), vocabulary is key in both speaking and speaking + writing tasks. These young learners need vocabulary to move the tasks along and, the use of PKL to avoid communication breakdowns. More specifically, PKL mediated the discussion of vocabulary issues in the form of predominant turns in similar proportions in both tasks, while minor turns were more frequently used in the speaking task than in the speaking + writing task.
As regards the fourth research question (What functions are more commonly served by PKL in each task?), vocabulary and metacognitive talk were the most common functions, followed by phatics in the speaking task. These are the functions with greater use of the PKL in previous investigations with both children (Azkarai & García Mayo, 2017;García Mayo & Hidalgo, 2017;Martínez-Adrián, 2020a, 2020b and adults (Alegría de la Colina & García Mayo, 2009;Arratibel-Irazusta & Martínez-Adrián, 2018;Azkarai & García Mayo, 2015;Storch & Aldosari, 2010). Grammar and mechanics were inexistent in this task, as also attested in other studies that have examined oral tasks (Lasito & Storch, 2013). In the case of the speaking + writing task, metacognitive talk was the most common function followed by vocabulary and mechanics. These findings contrast with studies conducted with adults in which PKL mediated grammar talk to a greater extent (Alegría de la Colina & García Mayo, 2009;Azkarai & García Mayo, 2015;Rayati et al., 2012;Payant & Kim, 2019). But even if PKL did not intervene in grammar talk, these young learners do employ them when attending to other formal aspects such as mechanics. In fact, previous studies carried out with adolescents on collaborative writing have concluded that vocabulary and spelling attract a lot of attention from young learners (Kim & McDonough, 2011). This paper has contributed to the scarcity of studies examining the interface between task-modality and the use of PKL (Azkarai & García Mayo, 2015;Payant & Kim, 2019), particularly in the case of young learners. The analysis of the data has revealed that these learners are extensive users of PKL, most of the times in the form of predominant turns. However, we cannot dismiss the fact that they are young and low-proficient learners, and the use of PKL in this case acts as a cognitive tool, facilitating the organization of the tasks, the co-construction of meaning and the attention to formal aspects of language such as mechanics. Given their young age and their low proficiency, had we banned the use of their PKL during interaction, these learners could not have moved the tasks forward.
As regards task-modality, it had an effect on the amount of PKL use, as a higher number of PKL turns and PKL words in the speaking + writing task were obtained. Nevertheless, task-modality had a limited effect on functions of PKL use, which contrasts with previous studies conducted with adults. In the present study, higher use of PKL was observed in metacognitive talk and mechanics in the speaking + writing task, while no differences emerged between tasks in the other categories. Unlike previous research with adults, vocabulary was key in both tasks to move them forward and grammar talk mediated by PKL was nearly inexistent in the database, a finding that was ascribed to the low number of elaborations in the form of metalinguistic explanations. If these learners were capable of discussing grammar issues more extensively, they could have resorted to their PKL to verbalize their thoughts. Perhaps a more structured task such as an editing task could have prompted the provision of further grammar explanations and in turn a higher use of PKL.
Overall, the findings indicate the benefits of these tasks in contributing to the promotion of learning opportunities for young CLIL learners. These collaborative tasks offer young learners opportunities to question their language use. In particular, these tasks lead learners to notice holes in their interlanguage systems and to receive assistance by their peers, with their PKL playing a major role as mediators. However, even if the speaking + writing task is associated with more grammar discussions, they are not so elaborated, which seems to indicate that formal aspects of language should be particularly enhanced in CLIL contexts. Perhaps more focused tasks could bring about more extensive discussions about form, and in turn, greater noticing of gaps. In addition, by examining the use of their PKL in these discussions, we will be especially contributing to the effect of translanguaging practices in the learning of grammar (see Pawlak, 2019) and to the call made by some researchers for the exploration of the status of learners' PKL as regards the interface Task-based Language Teaching and Content-based Instruction (Ortega, 2015).
Future research should shed more light on the effect of more and less structured tasks as regards the functions of both PKL use and TL use. Similarly, future studies should consider a larger number of participants and different types of pairings (pairs vs. small groups; teacher-selected vs. self-selected pairs). Likewise, the study of task-modality effects across age, proficiency and gender both in CLIL and non-CLIL settings, would be advisable. More specifically, by comparing CLIL to non-CLIL learners, we could investigate the effect of the input provided in CLIL programs or if similar benefits emerge in both settings in young and older learners.