The role of proficiency and pair formation method in language-related episodes: A study of young CLIL learners’ interaction

Abstract

Little is known about young CLIL (Content and Language Integrated Learning) learners’ attention to formal aspects of the target language when engaged in collaborative task-based interaction. Previous research on language-related episodes (LREs) with other populations indicates that certain variables (e.g. target language proficiency or pair formation method) may play a role in the production of LREs. This study investigates the amount, types and resolution of LREs produced by primary education CLIL learners in a collaborative picture-ordering + story-telling task depending on two variables – L2 English proficiency (grade 5 dyads vs. grade 6 dyads) and pairing method (proficiency-matched dyads vs. student self-selected dyads). Findings indicate that young CLIL learners’ interactive behaviour in L2 English, at least in terms of LRE production, does not differ as a consequence of target language proficiency, whereas pair formation method exerts some influence, self-selected pairs producing and resolving more meaning-based LREs. No differences were found for form-focused LREs.

Keywords

collaborative task content and language integrated learning (CLIL)language-related episodes (LREs)pair formation method target language proficiency

I Introduction

Research within the Interaction framework (Long, 1996) has examined the language learning opportunities that arise in task-based interaction (see Mackey and Goo, 2012 for an overview). A considerable body of research has revolved around the occurrence of learner-initiated attention to formal aspects of language while performing collaborative tasks, (e.g. Swain, 1998), namely the production of language-related episodes (LREs). LREs were originally defined by Swain (1998: 70) as ‘any part of the dialogue in which students talk about the language they are producing, question their language use, or other- or self-correct.’ A more recent definition stresses the fact that they constitute interactional feedback which generally involves a focus on meaning, and, to a lesser extent, a focus on morphosyntax (Mackey, 2007). LREs centre around meaning, spelling, pronunciation of a word or a grammatical form learners question about. One important factor related to the use of collaborative tasks in the foreign-language (FL) classroom is how learner characteristics exert an influence on task performance. Previous task-based research has established that learner proficiency (Basterrechea & Leeser, 2019; Benson, Pavitt & Jenkins, 2005; Kim & McDonough, 2008; Kowal & Swain, 1994; Leeser, 2004; Malmqvist, 2005; Williams, 1999, 2001), gender (Ross-Feldman, 2007), or personality traits (Storch & Aldosari, 2013; Watanabe & Swain, 2007) impact the nature of interactional feedback, as well as the quantity, quality and resolution of LREs.

This research has also examined the effect that factors such as pairing learners with similar proficiency levels or pair dynamics has on the production of LREs (e.g. García Mayo & Imaz Aguirre, 2019; Mozaffari, 2017; Storch, 2002; Storch & Aldosari, 2013; Watanabe & Swain, 2007). Some results suggest that pair dynamics and the relationships formed by dyad members may be a more important consideration than proficiency. Most research on LREs so far has considered adults in English as a second language (ESL; Benson et al., 2005), immersion (Kowal & Swain, 1994; Swain, 1998; Swain & Lapkin, 1998), content-based instruction (Leeser, 2004), or FL settings (Basterrechea & García Mayo, 2013; Basterrechea & Leeser, 2019; Kim & McDonough, 2008; García Mayo, 2002a, 2002b; García Mayo & Azkarai, 2016; Malmqvist, 2005; Storch & Aldosari, 2013), but little research has centred around young English as a foreign language (EFL) learners’ production of LREs. Several authors have recently claimed that, in general, little research exists that centres on the processes of foreign language learning by young learners at the elementary school level (Collins & Muñoz, 2016; García Mayo, 2017). However, it is worth investigating language learning during this period of life because language awareness is related to maturational factors and developed in parallel with the acquisition of literacy in those school years (Muñoz, 2017; Roehr-Brackin, 2018). The scarce research on child interaction in EFL contexts (e.g. García Mayo & Lázaro, 2015; Pinter, 2007) shows that young learners are concerned with language and they do negotiate about meaning. For instance, in a pioneering study, García Mayo and Lázaro (2015) compared interactional moves of mainstream EFL and CLIL (Content and Language Integrated Learning) learners. In CLIL, a learning context which has become widespread in Europe in the last decades, the FL (generally English) is the vehicular language of instruction in subjects other than English as a school subject, whereby learners are exposed to more hours of input than their mainstream EFL counterparts. The study analysed interactional moves of young EFL learners aged 8 to 11 while completing a picture placement task. Results showed that CLIL learners produce more interactional moves than EFL learners. More research on child interaction is needed in order to explore if young learners focus on language form while interacting, and consciously reflect on their language and produce LREs. This research should also inform about how we can form pairs that will best contribute to collaborative reflection on language use. With the exception of García Mayo and Imaz Aguirre (2019), no studies have investigated the impact that pairing learners with the same proficiency or having learners choose their partner has on the production of LREs by young EFL learners. Besides, a call for research on task-based interaction in CLIL has been made recently, as there is a dearth of this type of studies in that learning environment (Azkarai & García Mayo. 2015, 2017). Some authors have pointed out that teachers tend not to focus on form explicitly, in order not to deviate learners’ attention from content in CLIL lessons (Milla Melero & García Mayo, 2014). Hence, that type of research will shed light on how much CLIL learners focus on formal aspects of language while they accomplish certain tasks. Therefore, this study investigates how target language proficiency and the choice of pair formation method with young CLIL (English) learners (student-selected vs. proficiency-matched) may influence the quantity and quality of LREs produced in an oral narration task.

II Literature background

A number of studies has investigated the nature of LREs and the factors that may affect their occurrence in different educational contexts. Some studies have considered the relationship between LREs and L2 development (Adams, 2007; Kim, 2008; LaPierre, 1994; McDonough & Sunitham, 2009; Swain & Lapkin, 1998, 2001; Williams, 2001) and have shown that, in general, the LREs produced during collaborative tasks have a positive impact on L2 acquisition. Studies have also shown that task types and their features affect the amount and type of LREs produced (Adams & Ross-Feldman, 2008; García Mayo & Azkarai, 2016). Another strand of research on LREs has investigated how learner-internal factors, such as target language proficiency and personality traits, affect the quantity, quality and resolution of LREs (Basterrechea & Leeser, 2019; Benson et al., 2005; Kim & McDonough, 2008; Kowal & Swain, 1994; Leeser, 2004; Malmqvist, 2005; Mozaffari, 2017; Storch & Aldosari, 2013; Watanabe & Swain, 2007; Williams, 1999, 2001), motivated by the fact that, as LREs centre around gaps (Swain, 1995) in a learner’s interlanguage, proficiency may impact the type of LREs produced. For instance, Leeser (2004) examined whether L2 Spanish learners with different proficiency levels in a content-based university Spanish course would differ in terms of the amount, type and resolution of LREs. The learners were assigned to carry out a dictogloss task in pairs as follows: some dyads were proficiency-matched (where both members had a high or a low proficiency level – high-high or low-low) and others had a varying proficiency (high-low). It was found that high proficiency learners produced a greater number of form-focused LREs than meaning-focused LREs, and correctly resolved a higher amount of LREs than the high-low and low-low proficiency dyads did. The lower proficiency learners’ deliberations focused on the meaning they could extract from the text and generation of ideas, and less on formal aspects, and were more likely to leave LREs unresolved. Basterrechea and Leeser (2019) also found a positive correlation between the production of LREs and learner proficiency in a study that investigated how adolescent CLIL learners’ proficiency in English affected the production of LREs involving a specific grammatical feature (the English 3rd person singular form) during a dictogloss task. The study showed that the more advanced learners generated more LREs, had the ability to solve them correctly to a greater extent and used this information more effectively in writing. Positive correlations were also found between learner proficiency and correctly resolved form-focused LREs involving the target form, as well as the correct use of the grammatical feature in the dictogloss task.

Among the studies that examine the effect of proficiency on the production of LREs, one of the issues that has not been resolved yet seems to be the procedure for pairing the learners. For instance, Malmqvist (2005) matched Swedish learners of German as a FL with differing proficiency levels, assuming that they would benefit more from heterogeneous grouping. It was found that, as attested in Leeser (2004) and Basterrechea and Leeser (2019), less proficient learners attended to lexical items. The study also showed that personality traits exert a crucial role in group discussions, as lower proficiency learners took a leading role. The author thus concluded that interpersonal dynamics should be considered when pairing learners. Kim and McDonough (2008) similarly investigated the effect of proficiency on the occurrence and resolution of LREs, as well as pair dynamics in intermediate and advanced learners of L2 Korean with different L1 backgrounds. Results showed that learners took different roles when paired with advanced or with intermediate peers. Following Storch’s (2002) model of dyadic interaction,¹ it was found that learners who were collaborative with an intermediate-level interlocutor, were passive or novice when paired with an advanced-level peer. However, learners exhibiting a dominant role with an intermediate-level peer took a collaborative role with an advanced-level interlocutor. In addition, unlike other studies, learners produced more meaning-focused LREs with an advanced-level interlocutor than with an intermediate-level one, whereas no differences were attested in the production of form-focused LREs. Storch and Aldosari (2013) examined the nature of pair work in an EFL classroom in Saudi Arabia with pairs with similar and varying proficiency levels. High proficiency dyads produced, once again, more LREs. Similar proficiency pairs exhibited a tendency to form more collaborative contributions, whereas in mixed proficiency dyads, the more proficient learner tended to take a leading role, while the less proficient learner did not contribute to the task or focus on language use, especially if they formed non-collaborative relationships.

In sum, the findings from the aforementioned studies seem to show that pair dynamics and the relationship formed by the dyad members may influence the quantity, quality and resolution of LREs during collaboration. However, the relationship among LREs, proficiency and pair dynamics needs further consideration, as it is still unclear how we can best form high performance groups, and if learner proficiency or the relationship formed by the dyad members may be more influential in promoting attention to form. Previous studies have shown that what learners bring to the task in terms of relationships with one another may impact the incidence of focus on form (Philp, Walter and Basturkmen, 2010). Overall, studies that compare self-selected and teacher-assigned pairs have exhibited an advantage for the former in task accomplishment (e.g. Russell, 2010), although these have been conducted in learning contexts other than the SL/FL classroom. Mozaffari (2017) compared the production of LREs by student-selected and teacher-assigned pairs of adult (ages 20-26) EFL (Iranian L1) learners, as well as their patterns of interaction. For the self-selected grouping, it was assumed that whenever learners were free to choose, they would pair up with a friend, as attested in other studies (Russell, 2010). In the case of teacher-assigned pairs, learners were proficiency-matched, as all had an intermediate level. It was found that teacher-assigned pairs generated significantly more LREs than student-selected ones. Results also indicated that self-selected pairs talked about matters unrelated to the task more frequently than teacher-selected pairs. As for the patterns of interaction, no difference was found, both groups exhibiting a collaborative relationship. Recently, García Mayo and Imaz Aguirre (2019) have examined how pairing method affects the production of LREs by young EFL learners. Among other issues, they compared how proficiency-paired, teacher-selected (based on the participants’ teacher’s perception of the students’ personality) and self-selected groups produced LREs while completing an oral task (where they had to agree on an order for a set of cartoon strips) and an oral+written task (where they were asked to act as detectives and write an answer based on drawings with some clues). Results showed that proficiency-paired groups produced more LREs in both types of tasks, followed by the teacher-assigned pairs and finally by the self-selected pairs. In addition, the proficiency-paired group exhibited a collaborative type of dynamics to a larger extent in the oral+written task than the two other groups. To sum up, the scarce research on the effect that pairing method has on the production of LREs reported on above seems to show that pairing learners based on their proficiency is more beneficial in fostering attention to language form than self-selected pairing. Nonetheless, these studies solely focused on the amount of LREs produced by the comparison groups, and not on their quality (i.e. form-focused or meaning-focused) and resolution (i.e. resolved or unresolved).

Thus, this study attempts to expand the existing research on LREs by investigating the occurrence, type and resolution of LREs of young L2 English learners in a CLIL programme in the Basque Autonomous Community, as more research that explores how we can optimize collaborative work that fosters attention to formal aspects of language in this context with limited exposure to the target language (TL) is needed, since ‘the negotiation concept provides an excellent basis for a content-and-language approach, given that the school subjects are talked into being during lessons’ (Dalton-Puffer, 2011: 191). Besides, more research evidence is needed from elementary school learners’ behaviour in task-based interaction (Azkarai & García Mayo, 2015, 2017).

These are the research questions entertained in the study:

Does proficiency/grade affect the amount and types of LREs?

Does pair formation method affect the amount and types of LREs?

III Methodology

1 Participants

The study was conducted in a public school in the Basque Country. Twenty-seven dyads of Basque-Spanish bilinguals (34% female, 66% male) learning English as their L3 in the 5th and 6th grades of primary education (age range 10-11) took part in the study. Participants belonged to a CLIL program, in which English is taught from pre-primary education (age 4) as a school subject and it becomes a vehicle of instruction from the 3rd grade of primary education in subjects such as science, arts and crafts or physical education. In 5th and 6th grades (where learners in the study belonged to) schoolchildren receive three hours a week of EFL instruction and two-to-four hours a week of CLIL instruction, which amounts to five-to-seven weekly hours of exposure to the TL. At the time of data gathering, 5th graders (n = 20) had received 740 hours of exposure in a classroom setting, whereas 6th graders (n = 34) had received 925 hours. As far as their overall English proficiency, a Mann U-Whitney test revealed that there was a statistically significant difference (U = 187.0, p = .001) in the average proficiency level of the two grades, with a mean of 28.16 (Median = 24.00) in 5th grade and a mean of 44.08 (Median = 41.50) in 6th grade, as measured by the Key English Test (KET; Cambridge University Press). Students in each of the grades were placed in pairs following two different criteria –some students (n = 20; 10 from grade 5 and 10 from grade 6) were asked to select a partner to work with, whereas the rest of the participants (n = 34; 10 from grade 5 and 24 from grade 6) were paired according to the scores they obtained in the KET they had been administered prior to the treatment. Previous studies (e.g. Mozaffari, 2017; Russell, 2010), which have surveyed the criterion upon which students select their working partners collectively, have shown that pre-existing friendship is behind their choice. Accordingly, it was assumed that the participants in the former group would select their partner on the basis of friendship, as well.

2 Research instruments and procedure

Before data collection, we obtained parental permission to video-record the participants. Subsequently, they completed a background questionnaire and the KET for their proficiency level to be assessed. During the treatment session, they completed an oral narration task in pairs. This activity was based on the story ‘Dotty’s doll’ from the Teachers’ book Sparks 1 (House and Scott, 2009: 74-75²), which participants had to order following these steps: first they were instructed to agree on an order for six vignettes, which depicted how a girl was crying because her doll was broken, and how a friend volunteered to put the different pieces of the doll together by sewing them, but she did not do it properly; subsequently, they were asked to tell the story in turns in such a way that it was coherent and grammatically correct.

Learners’ interaction was audiotaped, transcribed verbatim and codified using CHILDES (MacWhinney, 2000). All cases of LREs were identified and, drawing on García Mayo and Azkarai (2016), classified according to their nature and outcome. In terms of nature, they were coded as meaning-focused, which includes word meaning or word choice, and form-focused, which includes phonology, prepositions and morphosyntax. Finally, based on the work of Leeser (2004), LREs were further categorized for their resolution, as resolved (target-like and non target-like) and unresolved. Two different researchers accomplished this LRE codification independently and any discrepancies were solved on a case-by-case basis. Self-repair, defined as any instance in which a learner modified his/her own utterance in the turn or in an adjacent turn without input from his/her interlocutor (Adams and Ross-Feldman, 2008), was also included. The examples from our database provided below illustrate meaning-focused and form-focused LREs.

Excerpt (1) contains a meaning-focused LRE dealing with word choice. Child 1 suggested the use of do, but child 2 corrected child 2’s utterance with make instead, which was accepted by child 1 in the following turn, reaching a correct resolution.

(1) *CHI1: and later they start doing it.

*CHI2: making it.

*CHI1: making it later.

The second excerpt (2) illustrates a target-like form-focused LRE, resolved by self-repair, where child 2 corrected his/her own utterance, adding the English 3rd person singular marker -s in want, while child 1 was not involved in the correction.

(2) *CHI2: then the girl has a new idea and (.) it eh and he want she wants to eh.

The following example (3) displays an instance of a non target-like form-focused LRE referring to subject–verb agreement. Child 1 omits the verb in the first turn, which was subsequently corrected by child 2, an incorrect resolution accepted by the former, as shown in the last turn.

(3) *CHI1: children happy (.) eh.

*CHI2: children is happy.

*CHI1: children is happy.

The following excerpt (4) illustrates an example of an unresolved LRE, also found in our database, dealing with word choice. Child 1 inquired about the word sew in English. The meaning-focused LRE was left unresolved.

(4) *CHI2: como se decía coser? (How do we say to sew?)

*CHI1: eh (.) ni idea. (No idea)

3 Analyses

Means (x̄), medians (Md) and standard deviations (SD) were calculated both for the whole set of LREs and for each type of LRE according to the variables of the study, that is, school grade (grade 5 vs. grade 6) and learner pairing (proficiency-matched vs. self-selected). For each of these variables, data were analysed in two different ways – the first analysis looked into inter-group differences for all LREs overall and for each LRE type whereas the second analysis explored intra-group differences with regard to the production of the different LRE types.

Prior to any of these analyses, Kolmogorov-Smirnov tests were run so as to know whether the distribution of the samples was normal. Samples were not normally distributed and, consequently, non-parametric tests were used.

With regard to intergroup comparisons, as the study variables (proficiency and pairing method) have two values (grade 5 vs. grade 6; proficiency-matched vs. self-selected), Mann–Whitney U tests (also known as Wilcoxon Rank-Sum tests) for two independent samples were computed.

As for intragroup comparisons, non-parametric Friedman tests of differences among the various LRE categories were conducted and rendered significant (p < .05) Chi-square values of 22.22 and 33.08 for grade 5 and 6 samples, and of 25.52 and 31.95 for proficiency-matched and self-selected samples. Hence, post-hoc Wilcoxon Signed-Ranks tests for dependent samples were computed in order to compare the different LRE categories on a one-to-one basis.

Regarding statistical probability, an alpha level of .05 (*) was used. Besides, marginally significant p-values below .09 (#) were indicated. Additionally, in order to discover the strength of the differences, effect sizes were calculated by means of Cohen’s d. It is worth noting that this effect is considered ‘small’ if Cohen’s d is around .2, ‘medium’ if about .5, and ‘large’ if above .8.

IV Results

In this section, the results of the analyses performed on the data will be shown according to the two variables of the study. First, we will present the results in relation with the variable ‘proficiency/grade’. Next, the outcomes of the analyses conducted to explore the influence of the variable ‘pair formation method’ will be displayed.

1 Proficiency/grade

In order to examine the effect of proficiency/grade on the production of LREs, both intergroup and intragroup analyses were performed. As for the former, the results of the Mann–Whitney tests (see Table 1) indicated that there were no statistically significant differences between grade 5 and grade 6 students as far as the overall number of LREs. The very same lack of statistical significance was found for both meaning-focused LREs and form-focused LRES, as well as for any of the LRE subcategories examined (resolved, target-like resolved, non-target-like resolved, and unresolved).

Table 1.

Means/medians (and standard deviations) for each LRE category in grade 5 vs. grade 6 dyads.

LRE types	Grade 5 (10 dyads)		Grade 6 (17 dyads)		Mann–Whitney		Size
LRE types	x̄ / Md	SD	x̄ / Md	SD	U	p.	d
All	4.55 / 5.00	3.86	3.47 / 3.00	2.50	69.5	.52	.332
Meaning:	4.36 / 4.50	3.61	3.18 / 3.00	2.27	63.5	.45	.391
Resolved meaning	3.27 / 3.00	3.16	2.47 / 2.00	2.09	66.5	.58	.299
Target-like resolved meaning	1.36 / 1.00	1.80	1.24 / 1.50	1.35	78.5	.92	.075
Non-target-like resolved meaning	1.91 / 1.50	2.43	1.24 / 1.50	1.35	77.0	.71	.341
Unresolved meaning	1.09 / 1.00	1.22	.71 / 1.00	.92	71.0	.45	.352
Form:	.18 / .00	.60	.29 / .00	.77	82.5	.56	−.159
Resolved form	.18 / .00	.60	.29 / .00	.77	82.5	.56	.159
Target-like resolved form	.00 / .00	.00	.29 / .00	.77	74.0	.15	−.533
Non-target-like resolved form	.18 / .00	.60	.00 / .00	.00	81.0	.21	.424
Unresolved form	.00 / .00	.00	.00 / .00	.00	85.0	1.00	–

Tables 2 and 3 present the intragroup analyses performed with grade 5 and grade 6 separately in order to discover intragroup differences among the six LRE types classified in this study: target-like resolved, non-target-like resolved and unresolved for both meaning-based and form-based LREs. As far as the grade 5 sample is concerned (see Table 2), Wilcoxon tests indicated that there were statistical differences when meaning-focused LREs were compared to form-focused LREs, meaning-focused LREs always being more frequent.

Table 2.

LRE type comparison in grade 5 dyads (10 dyads).

	Target-like resolved meaning	Non-target-like resolved meaning	Unresolved meaning	Target-like resolved form	Non-target-like resolved form	Unresolved form
Mean (SD)	1.36 (1.80)	1.91 (2.43)	1.09 (1.22)	.00 (.00)	.18 (.60)	.00 (.00)
Median	1.00	1.50	1.00	.00	.00	.00
Target-like resolved meaning
Non-target-like resolved meaning	z = 1.02p = .30d = -.257
Unresolved meaning	z = .43p = .07#d = .175	z = .64p = .53d = .426
Target-like resolved form	z = 2.41p = .02* d = 1.068	z = 2.21p = .03* d = 1.112	z = 2.22p = .03* d = 1.263
Non-target-like resolved form	z = 1.99p = .05* d = .879	z = 2.21p = .03* d = .977	z = 1.71p = .09#d = .947	z = 1.00p = .32d = -.424
Unresolved form	z = 2.41p = .02* d = 1.068	z = 2.21p = .03* d = 1.112	z = 2.22p = .03* d = 1.263	z = .00p = 1.00d = –	z = 1.00p = .32d = .424

Notes. *p < .05. #p < .09.

Table 3.

LRE type comparison in grade 6 dyads (17 dyads).

	Target-like resolved meaning	Non-target-like resolved meaning	Unresolved meaning	Target-like resolved form	Non-target-like resolved form	Unresolved form
Mean (SD)	1.24(1.35)	1.24 (1.35)	.71 (.92)	.29 (.60)	.00 (.00)	.00 (.00)
Median	1.50	1.50	1.00	.00	.00	.00
Target-like resolved meaning
Non-target-like resolved meaning	z = .24p = .81d = 0
Unresolved meaning	z = 1.54p = .12d = .459	z = 1.12p = .26d = .459
Target-like resolved form	z = 2.12p = .04* d = .909	z = 2.55p = .01* d = .458	z = 1.28p = .20d = .541
Non-target-like resolved form	z = 2.98p = .01* d = 1.299	z = 2.83p = .01* d = 1.299	z = 2.59p = .01* d = 1.091	z = 1.63p = .10d = .683
Unresolved form	z = 2.98p = .01* d = 1.299	z = 2.83p = .01* d = 1.299	z = 2.59p = .01* d = 1.091	z = 1.63p = .10d = .683	z = .00p = 1.00d = –

Note. * p < .05.

In fact, the overall comparison between all the meaning-focused LREs vs. all the form-focused LREs produced by grade 5 learners reached statistical significance (z = 2.06 p = .04* d = 1.615). Cohen’s d values were above .8 whenever a significant difference was found, indicating that the magnitude of the effect was large. When comparisons were established either within different types of form-focused LREs or within different types of meaning-focused LREs, no statistically significant differences were discovered, with the exception of a marginal difference between target-like resolved and unresolved meaning-focused LREs in favor of the former. It is also worth mentioning that, as shown by the means obtained, most of the meaning-focused LREs produced by the grade 5 learners were resolved (either in a target-like manner or not). The production of form-focused LREs was minimal though, with just one of the categories (non-target-like resolved) being rendered.

As for the grade 6 sample, (see Table 3), Wilcoxon tests revealed that meaning-focused LRE types were produced to a significantly larger extent than form-focused LREs, the only exception being the target-like resolved form-focused vs. unresolved meaning-focused LRE contrast, where differences in favor or meaning-focused LREs did not reach statistical significance. Consequently, grade 6 learners produced significantly more meaning-focused than form-focused LREs overall (z = 2.03 p = .04* d = 1.705). Besides, the effect size for these significant differences was above .8 on most occasions. As happened to grade 5 students, no significant differences were found for the grade 6 sample within the different meaning-focused LRE types or within the various form-focused LRE types. Overall, as shown by the means obtained, meaning-focused LREs were resolved (accurately or inaccurately) on most occasions, and the minimal incidence of form-based LREs was represented by a resolved (target-like) category.

2 Pair formation method

Both intergroup and intragroup comparisons were made in order to discover whether the ‘pair formation method’ factor had an effect on the incidence, nature and resolution of the LREs produced by learners. Table 4 presents the results of the first type of analysis, that is, the intergroup analysis. As can be seen in the table, self-selected pairs globally produced a substantially larger number of LREs than proficiency-matched pairs. The Mann–Whitney test indicated that self-selected pairs’ overall production of LREs was significantly superior to that of proficiency-matched pairs. As indicated by Cohen’s d, the effect size between these two means was large. When we compared the two learner groups in their production of each LRE macro-category, that is, meaning-focused LREs and form-focused LREs separately, the Mann–Whitney tests indicated that there were significant differences depending on the pair formation method as far as the production of meaning-focused LRE was concerned, self-selected pairs obtaining a higher mean. However, the production of form-focused LREs was minimal in both learner groups, no differences emerging for this type of LREs.

Table 4.

Means/medians (and standard deviations) for each LRE category in proficiency-matched vs. self-selected dyads.

LRE types	Proficiency-matched (17 dyads)		Self-selected (10 dyads)		Mann–Whitney		Size
LRE types	x̄ / Md	SD	x̄ / Md	SD	U	p	d
All	2.72 / 2.00	2.35	6.00 / 6.50	3.23	26.5	.01*	−1.161
Meaning:	2.44 / 2.00	2.01	5.80 / 6.00	3.01	26.0	.01*	−1.313
Resolved meaning	1.61 / 1.00	1.68	4.90 / 5.00	2.51	23.0	.01*	−1.540
Target-like resolved meaning	.67 / 1.00	.91	2.40 / 2.00	1.78	38.5	.01*	−1.224
Non-target-like resolved meaning	.94 / 1.00	1.21	2.50 / 2.50	2.37	44.0	.07#	−.830
Unresolved meaning	.83 / 1.00	1.10	.90 / 1.00	.99	60.5	.70	−.067
Form:	.28 / .00	.75	.20 / .00	.63	56.5	.66	.155
Resolved form	.28 / .00	.75	.20 / .00	.63	56.5	.66	.155
Target-like resolved form	.28 / .00	.75	.00 / .00	.00	66.0	.18	.528
Non-target-like resolved form	.00 / .00	.00	.20 / .00	.63	68.0	.18	−.449
Unresolved form	.00 / .00	.00	.00 / .00	.00	85.0	1.00	–

Notes. *p < 0.5. #p < .09.

When we looked into the different types of meaning-focused LREs depending on whether they were resolved (or not) and if so, on whether the right form was targeted, the Mann–Whitney tests yielded significant intergroup differences for resolved meaning-focused LREs. Moreover, the production of both target-like and non-target-like meaning-focused LREs by self-selected pairs surpassed that of proficiency-matched pairs, this superiority being significant for the former and marginally significant for the latter. As shown by the Cohen’s d values, pairing method had a large effect on the incidence of these types of LREs. However, no inter-group differences were found for unresolved meaning-focused LREs, which yielded very similar means in both participant groups.

The very same analyses were performed with the various types of form-focused LREs. As mentioned above, the production of form-focused LREs was very low in both learner groups. The form-focused LREs produced were solved either in a target-like manner by the proficiency-matched group, or inaccurately by the self-selected pairs, and there was a lack of incidence of unresolved form-focused LREs in the data. The Mann–Whitney tests indicated that none of the intergroup differences were statistically supported in the case of the various types of form-focused LREs.

We will proceed to present the intragroup analyses next. As for the proficiency-matched sample, Table 5 displays the results of the intragroup analyses performed in order to discover differences among the six LRE types categorized in this study. Wilcoxon tests indicated that there were significant differences when the production of most of the different meaning-focused LRE types was compared to the incidence of the various form-focused LRE types. Meaning-focused LRE means were always higher than form-focused LRE means. Accordingly, a significant difference was also obtained when the global number of meaning-focused LREs was compared to the global number and form-focused LREs (z = 2.81 p = .01* d = 1.424). The magnitude of the effect was large, with Cohen’s d values above .8 on most occasions. When bidirectional comparisons were established either within different types of form-focused LREs or within different types of meaning-focused LREs, the Wilcoxon tests did not yield statistically significant differences. It should also be mentioned that means show that in the proficiency-matched pairs there were more resolved (target-like plus non-target-like) than unresolved meaning-based LREs. Besides, a resolved (target-like) LRE type was the one representing the production of form-based LREs in this sample.

Table 5.

LRE type comparison in proficiency-matched dyads (17 dyads).

	Target-like resolved meaning	Non-target-like resolved meaning	Unresolved meaning	Target-like resolved form	Non-target-like resolved form	Unresolved form
Mean (SD)	.67 (.91)	.94 (1.21)	.83 (1.10)	.28 (.75)	.00 (.00)	.00 (.00)
Median	1.00	1.00	1.00	.00	.00	.00
Target-like resolved meaning
Non-target-like resolved meaning	z = .97p = .33d = –.252
Unresolved meaning	z = .41p = .68d = .158	z = .20p = .84d = .095
Target-like resolved form	z = 1.35p = .18d = .468	z = 2.40p = .02* d = .656	z = 1.37p = .17d = .584
Non-target-like resolved form	z = 2.59p = .01* d = 1.041	z = 2.70p = .01* d = 1.099	z = 2.55p = .01* d = 1.067	z = 1.63p = .10d = .528
Unresolved form	z = 2.59p = .01* d = 1.041	z = 2.70p = .01* d = 1.099	z = 2.55p = .01* d = 1.067	z = 1.63p = .10d = .528	z = .00p = 1.00d = –

Note. *p < .05.

Table 6 shows the intra-group comparison results for the self-selected group. As the tendency discovered for the proficiency-matched sample, Wilcoxon tests revealed that most meaning-focused LRE types were found to be produced at a significantly higher frequency than the various types of form-focused LREs for self-selected pairs too, with large effect sizes being discovered. In agreement with this, self-selected pairs produced significantly more meaning- than form-focused LREs overall (z = 2.83 p = .01* d = 2.575). As for comparisons within meaning-focused LREs and within form-focused LREs separately, the latter did not yield statistically significant differences. However, some differences were found within meaning-focused LREs in the case of the self-selected sample, resolved meaning-focused LREs (both target-like and non-target-like) obtaining a higher production mean than unresolved meaning-focused LREs, with Cohen’s d values over .8 in these comparisons. In the self-selected dyads, as shown by the mean scores, the incidence of resolved meaning-based LREs was considerably higher than that of unresolved meaning-based LREs. Additionally, there was an absolute lack of unresolved types in the case of form-based LREs, a (non-target-like) resolved category being the only type of LREs rendered.

Table 6.

LRE type comparison in self-selected dyads (10 dyads).

	Target-like resolved meaning	Non-target-like resolved meaning	Unresolved meaning	Target-like resolved form	Non-target-like resolved form	Unresolved form
Mean (SD)	2.40 (1.78)	2.50 (2.37)	.90 (.99)	.00 (.00)	.20 (.63)	.00 (.00)
Median	2.00	2.50	1.00	.00	.00	.00
Target-like resolved meaning
Non-target-like resolved meaning	z = .41p = .68d = –.048
Unresolved meaning	z = 2.36p = .02* d = 1.041	z = 1.78p = .07#d = .881
Target-like resolved form	z = 2.83p = .01* d = 1.041	z = 2.37p = .02* d = 1.492	z = 2.26p = .02* d = 1.286
Non-target-like resolved form	z = 2.57p = .01* d = 1.648	z = 2.37p = .02* d = 1.326	z = 1.47p = .14d = .844	z = 1.00p = .32d = -.449
Unresolved form	z = 2.83p = .01* d = 1.041	z = 2.37p = .02* d = 1.492	z = 2.26p = .02* d = 1.286	z = .00p = 1.00d = –	z = 1.00p = .32d = .449

Notes. *p < .05. #p < .09.

V Discussion and conclusions

The purpose of this study was to investigate how target language proficiency and pair formation method may influence the quantity and quality of LREs produced by young CLIL English learners in an oral narration task, as studies that examine young EFL learner attention to form in interaction are thin on the ground, particularly those in CLIL settings.

Regarding our first research question (Does proficiency/grade affect the amount and types of LREs?), intra-group analyses revealed that beginner learners in both grades (5th and 6th) significantly produced more meaning-focused than form-focused LREs. The production of form-focused LREs was minimal. This finding coincides with previous research findings (Leeser, 2004; Malmqvist, 2005) where low proficient learners have been found to focus on lexis rather than on grammar when they discuss language aspects while performing collaborative tasks. Particularly, the learners in our study mostly focused their interaction on retrieving key vocabulary, a finding which aligns with recent research that equally examined the effect of pairing method with young EFL learners (García Mayo and Imaz Aguirre, 2019). In this vein, whereas previous studies looked into older students (e.g. Basterrechea & Leeser, 2019; Kim & McDonough, 2008; Mozaffari, 2017) – where findings seem to show that LREs have both a form focus and a meaning focus – our study analysed very young students (aged 10-11). Hence, these divergent findings seem to indicate that there may be age-related differences in how learners attend to form and meaning in collaborative work. This lesser focus on form could also be ascribed to the fact that participants were learning English in a CLIL environment. Prior research has shown that corrective feedback episodes are less frequent in CLIL than in traditional EFL lessons (Milla Melero and García Mayo, 2014) and a call for more attention to form has been made in these contexts (Gutiérrez Mangado and Martínez Adrián, 2018; Lyster, 2015; Martínez Adrián & Gutiérrez Mangado, 2015a, 2015b, 2018).

Additionally, our study did not reveal any significant differences between more and less proficient learners. In previous LRE research the proficiency factor proved to be more influential, as it accounted for the number of LREs, as in Storch and Aldosari (2013), or it explained the production of form-focused or meaning–focused LREs, as in Basterrechea and Leeser (2019), Leeser (2004) or Malmqvist (2005), all of whom found that the higher the proficiency, the greater the number of form-focused LREs. Similarly, the results in our study are consistent with the findings in Kim and McDonough (2008), who discovered that most LREs were meaning-focused. In addition, contrary to other studies, our results revealed that the higher the proficiency of the interlocutors, the greater the number of meaning-focused LREs, findings which may be attributed to the fact that the present study focused narrowly on beginner learners. Moreover, task features may have also mediated the quantity and quality of the LREs produced. In our study, dyads performed a picture ordering plus an oral (story telling) task, whereas previous literature findings reported on above come from collaborative writing tasks. In studies that assess the effect of task modality (i.e. oral vs. written) in the production of LREs, the general finding is that written tasks elicit more LREs than oral tasks. In terms of quality, written tasks favour form-focused LREs to a larger extent than oral tasks, which tend to involve more meaning-focused LREs (e.g. Adams, 2006; Adams & Ross-Feldman, 2008; García Mayo & Azkarai, 2016; Niu, 2009). Finally, we would like to pinpoint the fact that future research should consider designs where the variable proficiency would be better operationalized by comparing dyads of members with varied matched proficiencies (e.g. low proficiency dyads, average proficiency dyads and high proficiency dyads) rather than by contrasting two groups of participants, as our 5th and 6th graders, whose global proficiency is statistically different.

In order to answer the second research question (Does pair formation method affect the amount and types of LREs?), the study examined the effect of pairing (i.e. proficiency-matching vs. self-selection) on the quantity and quality of LREs produced. It was found that self-selected pairs produced a significantly larger number of meaning-focused LREs, no inter-group differences being found for form-focused LREs. Besides, self-selected pairs significantly produce more resolved than unresolved meaning-focused LREs, a pattern which is not discovered for proficiency-matched students. These results contrast with the ones from collaborative writing (Mozaffari, 2017; García Mayo & Imaz Aguirre, 2019), where there was less focus on language use by self-selected pairs. These controversial findings clearly call for further research on the interplay between task modality, learner age and pair formation method in terms of LRE production.

The findings obtained from the comparison of the two pairing methods indicate that young CLIL learners’ interactive behaviour in the L2, at least in terms of LRE production, seems to be more dependent on social considerations (Philp et al., 2010) than on psycho-linguistic aspects such as the level of proficiency. These results are in line with previous studies that examine the effect of proficiency level and/or pair dynamics on the production of LREs, where pair dynamics and the relationship established in the task are found to overrule proficiency (Kim & McDonough, 2008; Malmqvist, 2005; Philp et al., 2010; Storch & Aldosari, 2013; Watanabe & Swain, 2007). This may indicate that the interpersonal relationship between learners has promoted attention to language form during task-based interaction to a greater extent than their proficiency level. In other words, based on our results, there is certainly value for young English learners in a CLIL context to be paired with peers with whom there is a pre-existing friendship. Previous studies have attested that the student-selected condition leads to better collaboration (Russell, 2010) or how interlocutor familiarity may exert an influence on learner attention to form and meaning (Philp et al., 2010). Our study suggests that learner-initiated attention to form during task-based interaction subsumes a wider range of factors that may moderate the potential of focus on language form, one of which – learners’ interpersonal relationship – has been presented in this article. Unlike in other studies, where student-selected pairs produced fewer LREs (Mozaffari, 2017; García Mayo & Imaz Aguirre, 2019 in EFL contexts), the present study has shown that the learners’ relationship with one another may be directly influencing how (and how much) learners attend to form. Additionally, learners exhibited a clear tendency to address linguistic issues and reach correct resolutions when paired with a self-selected peer to greater advantage. This, although largely limited to meaning-oriented LREs (presumably as a consequence of task characteristics and/or proficiency level), may indicate that interpersonal factors positively affected young CLIL learners’ attention to form.

Other studies have alternatively shown that the proficiency-matched pairing condition elicited more attention to form in adult EFL (Mozaffari, 2017) and young EFL (García Mayo and Imaz Aguirre, 2019) learners. This contrast may indicate that our learners’ educational context – CLIL – may be playing a role, where more learner-learner interaction and dialogic activity are promoted compared to mainstream EFL settings (Coyle, 2007). We may speculate that collaborative group work in these contexts is more likely to be based on pre-existing friendship than on other factors. While we await for future research to address this issue, we may conclude that learners’ interpersonal relationship was a crucial moderating factor in young CLIL learners’ attention to form.

Some important pedagogical implications can be drawn from the findings for pairing method with young CLIL learners. Teachers are often concerned about how to form best performance groups, especially with young learners with a limited command of the target language. Special attention needs to be paid to interpersonal relationships, namely the pre-existing friendship prior to assigning learners to work in pairs, since, as the study seems to show, this factor may favourably contribute to collaborative reflection on language use. The study has also provided support for the importance of collaborative interaction, showing that young CLIL learners with a beginner proficiency level can engage in task-based interaction and, to some extent, focus on form. Although mainly focused on lexicon, the type of task used has shown to be amenable to promote focus on language use in a CLIL classroom. Thus, as previously attested (see García Mayo & Hidalgo Gordo, 2017; García Mayo & Lázaro, 2015), the use of tasks that promote learner-learner interaction and enhance the integration of subject-matter content and language with young CLIL learners would be highly recommended. Most importantly, unlike in other studies, participants’ personal relationship has shown to be a more important consideration than proficiency. Research into task-based interaction in CLIL contexts would benefit at the very least from examining how pair formation method influences interactional patterns and collaborative reflection on language use in a learning/teaching context that, by definition, encompasses a dual focus on form and content (Dalton-Puffer, Llinares, Lorenzo and Nikula, 2014). In this vein, CLIL teachers need to acknowledge the importance of designing tasks that elicit attention to form and meaning in order to maximize their potential with pairing learners based on their interpersonal relationships.

Further research would be advisable in order to analyse the effect of the LREs in a collaborative writing task, in order to determine how group discussion in tasks affects written language output. Besides, the grammatical accuracy of the stories that students produce could be analysed. The ecological validity of these studies would be enhanced if teachers were involved in the design and implementation of such tasks. In addition, the ‘pairing method’ variable could be further analysed by matching up, on the one hand, students with same proficiency levels and, on the other hand, students with different proficiency levels, as previous studies that examine the effect of mixed-proficiency dyads (e.g. Kim & McDonough, 2008; Storch & Aldosari, 2013) have rendered conflicting results. Apart from that, triangulation among pair formation method, quantity and quality of LREs and pair dynamics (Storch, 2002) would be advisable in order to test if the latter exerts some influence on collaborative reflection on language. Future research should also shed more light on the effect of CLIL on pairing method by comparing learners immersed in this methodological option to mainstream EFL learners so that the potential effects that CLIL might have had on the results of this study are tested experimentally.

Footnotes

Acknowledgements

We would like to express our gratitude to the school and the teachers involved in this project. Our special thanks go to the young students who took part in the study.

Funding

This work was supported by grants IT904-16 (Basque Government) and FFI2016-74950-P (Spanish Ministry of Economy and Competitiveness; National Research Agency and European Regional Development Fund - AEI/FEDER/EU).

ORCID iD

Francisco Gallardo del Puerto

Notes

References

Adams

(2007). Do second language learners benefit from interacting with each other? In Mackey

(Ed.), Conversational interaction in second language acquisition (pp. 29–52). Oxford: Oxford University Press.

Adams

(2006). L2 tasks and orientation to form a role for modality? ITL: International Journal of Applied Linguistics, 152, 7–34.

Adams

Ross-Feldman

(2008). Does writing influence learner attention to form? In Belcher

Hirvela

(Eds.), The oral/literate connection: Perspectives on L2 speaking/writing connections (pp. 243–267). Ann Arbor, MI: University of Michigan Press.

Azkarai

García Mayo

M.P.

(2015). Task-modality and L1 use in EFL oral interaction. Language Teaching Research, 19, 550–571.

Azkarai

García Mayo

M.P.

(2017). Task repetition effects on L1 use in EFL child task-based interaction. Language Teaching Research, 21(4), 480–495.

Basterrechea

García Mayo

M.P.

(2013). Language-related episodes (LREs) during collaborative tasks: A comparison of CLIL and EFL learners. In McDonough

Mackey

(Eds.), Second language interaction in diverse educational contexts (pp. 25–43). Amsterdam: John Benjamins.

Basterrechea

Leeser

(2019). Language-related episodes and learner proficiency during collaborative dialogue in CLIL. Language Awareness, 28, 97–113.

Benson

Pavitt

Jenkins

(2005). The use of dictogloss to encourage discussion of language use. Edinburgh Working Papers, 14, 1–17.

Collins

Muñoz

(2016). The foreign language classroom: Current perspectives and future considerations. The Modern Language Journal, 100, 133–147.

10.

Coyle

(2007). Content and language integrated learning: Toward a connected research agenda for CLIL pedagogies. The International Journal of Bilingual Education and Bilingualism, 10, 543–562.

11.

Dalton-Puffer

(2011). Content-and-language integrated learning: From practice to principles? Annual Review of Applied Linguistics 31, 182–204.

12.

Dalton-Puffer

Llinares

Lorenzo

Nikula

(2014). ‘You can stand under my umbrella’: Immersion, CLIL and bilingual education. Applied Linguistics, 35, 213–218.

13.

García Mayo

M.P.

(2002a). The effectiveness of two form-focused tasks in advanced EFL pedagogy. International Journal of Applied Linguistics, 12, 156–175.

14.

García Mayo

M.P.

(2002b). Interaction in advanced EFL pedagogy: A comparison of form-focused activities. International Journal of Educational Research, 37, 323–341.

15.

García Mayo

M.P.

(2017). Introduction. In García Mayo

M.P.

(Ed.), Learning foreign languages in primary school: Research insights (pp. xiii–xxiii). Bristol: Multilingual Matters.

16.

García Mayo

M.P.

Azkarai

(2016). EFL task-based interaction: Does task modality impact language-related episodes? In Sato

Ballinger

(Eds.), Peer interaction in second language learning (pp. 241–266). Amsterdam: John Benjamins.

17.

García Mayo

M.P.

Hidalgo Gordo

M.A.

(2017). L1 use among young EFL mainstream and CLIL learners in task-supported interaction. System, 67, 132–145.

18.

García Mayo

M.P.

Imaz Aguirre

(2019). Task modality and pair formation method: Their impact on patterns of interaction and LREs among EFL primary school children. System, 80, 165–175.

19.

García Mayo

M.P.

Lázaro

(2015). Do children negotiate for meaning in task-based interaction? Evidence from CLIL and EFL settings. System, 54, 40–54.

20.

Gutiérrez Mangado

Martínez Adrián

(2018). CLIL at the linguistic interfaces. International Journal of Immersion and Content-Based Education, 6, 85–112.

21.

House

Scott

(2009). Sparks 1. Teachers’ book. London: Richmond.

22.

Kim

(2008). The contribution of collaborative and individual tasks to the acquisition of L2 vocabulary. The Modern Language Journal, 92, 114–130.

23.

Kim

McDonough

(2008). The effect of interlocutor proficiency on the collaborative dialogue between Korean as a second language learners. Language Teaching Research, 12, 211–234.

24.

Kowal

Swain

(1994). Using collaborative language production tasks to promote students’ language awareness. Language Awareness, 3, 73–93.

25.

LaPierre

(1994). Language output in a cooperative learning setting: Determining its effects on second language learning. Unpublished PhD thesis, University of Toronto, Toronto, ON, Canada.

26.

Leeser

M.J.

(2004). Learner proficiency and focus on form during collaborative Dialogue. Language Teaching Research, 8, 55–81.

27.

Long

M.H.

(1996). The role of the linguistic environment in second language acquisition. In Ritchie

W.C.

Bhatia

T.K.

(Eds.), Handbook of language acquisition: Volume 2: Second language acquisition (pp. 413–468). New York: Academic Press.

28.

Lyster

(2015). Using form-focused tasks to investigate language across the immersion curriculum. System, 54, 4–13.

29.

Mackey

(Ed.). (2007). Conversational interaction in second language acquisition. Oxford: Oxford University Press.

30.

Mackey

Goo

(2012). Interaction approach in second language acquisition. In Chappelle

(Ed.), The encyclopedia of applied linguistics (pp. 2748–2758). Chichester: Wiley.

31.

MacWhinney

(2000). The CHILDES project: Tools for analyzing talk. 3rd edition. Mahwah, NJ: Lawrence Erlbaum.

32.

Malmqvist

(2005). How does group discussion in reconstruction tasks affect written language output? Language Awareness, 14, 128–141.

33.

Martínez Adrián

Gutiérrez Mangado

(2015a). Is CLIL instruction beneficial in terms of general proficiency and specific areas of grammar? International Journal of Immersion and Content-Based Education, 3, 51–76.

34.

Martínez Adrián

Gutiérrez Mangado

(2015b). L1 use, lexical richness, accuracy and complexity in CLIL and non-CLIL learners. Atlantis, 37, 175–200.

35.

McDonough

Sunitham

(2009). Collaborative dialogue between Thai EFL learners during self-access computer activities. TESOL Quarterly, 43, 231–254.

36.

Milla Melero

García Mayo

M.P.

(2014). Corrective feedback episodes in oral interaction: A comparison of a CLIL and an EFL classroom. International Journal of English Studies, 14, 1–20.

37.

Mozaffari

S.H.

(2017). Comparing student-selected and teacher-assigned pairs on collaborative writing. Language Teaching Research, 21, 496–516.

38.

Muñoz

(2017). The development of language awareness at the transition from primary to secondary school. In García Mayo

M.P.

(Ed.), Learning foreign languages in primary school: Research insights (pp. 49–68). Bristol: Multilingual Matters.

39.

Niu

(2009). Effect of task-inherent production modes on EFL learners’ focus on form. Language Awareness, 18, 384–402.

40.

Philp

Walter

Basturkmen

(2010). Peer interaction in the foreign language classroom: What factors foster a focus on form? Language Awareness, 19, 261–279.

41.

Pinter

(2007). Some benefits of peer-peer interaction: 10-year-old children practicing with a communication task. Language Teaching Research, 11, 189–207.

42.

Roehr-Brackin

(2018). Metalinguistic awareness and second language acquisition. New York: Routledge.

43.

Ross-Feldman

(2007). Interaction in the L2 classroom: Does gender influence learning opportunities? In Mackey

(Ed.), Conversational interaction in second language acquisition: A collection of empirical studies (pp. 52–77). Oxford: Oxford University Press.

44.

Russell

(2010). The formation of effective work groups within a FE classroom. Research in Post-Compulsory Education, 20, 286–305.

45.

Storch

(2002). Patterns of interaction in ESL pairwork. Language Learning, 52, 119–158.

46.

Storch

Aldosari

(2013). Pairing learners in pair work activity. Language Teaching Research, 17, 31–48.

47.

Swain

(1995). Three functions of output in second language learning. In Gook

Seidlhofer

(Eds.), Principle and practice in applied linguistics (pp. 125–144). Oxford: Oxford University Press.

48.

Swain

(1998). Focus on form through conscious reflection. Doughty

Williams

(Eds.), Focus on form in classroom second language acquisition (pp. 64–81). Cambridge: Cambridge University Press.

49.

Swain

Lapkin

(1998). Interaction and second language learning: Two adolescent French immersion students working together. The Modern Language Journal, 82, 320–337.

50.

Swain

Lapkin

(2001). Focus on form through collaborative dialogue: Exploring task effects. In Bygate

Skehan

Swain

(Eds.), Researching pedagogic tasks: Second language learning, teaching, and testing (pp. 99–118). New York: Longman.

51.

Watanabe

Swain

(2007). Effects of proficiency differences and patterns of pair interaction on second language learning: Collaborative dialogue between adult ESL learners. Language Teaching Research, 11, 121–142.

52.

Williams

(1999). Learner-generated attention to form. Language Learning, 51, 303–346.

53.

Williams

(2001). The effectiveness of spontaneous attention to form. System, 29, 325–340.