Applied Linguistics 2012: 33/3: 229–256 ß Oxford University Press 2012 doi:10.1093/applin/ams002 Advance Access published on 14 March 2012 Collaborative Dialogue in Learner–Learner and Learner–Native Speaker Interaction ANA FERNÁNDEZ DOBAO Division of Spanish and Portuguese Studies, University of Washington, Padelford C-104, Box 354360, Seattle, WA 98195, USA E-mail: anadobao@u.washington.edu This study analyses intermediate and advanced learner–learner and learner– native speaker (NS) interaction looking for collaborative dialogue. It investigates how the presence of a NS interlocutor affects the frequency and nature of lexical language-related episodes (LREs) spontaneously generated during task-based interaction. Twenty-four learners of English as a foreign language and eight English NSs participated in the study. The results obtained confirm that lexical LREs tend to be more frequent and more likely to be successfully resolved in learner–NS than in learner–learner interaction. However, the analysis of the patterns of interaction of two selected dyads reveals that not all NSs can be expected to provide the same kind of linguistic assistance to the learner. The participants’ collaborative or non-collaborative orientation to the activity, shaped by their goals and level of involvement in the task, seems to have a stronger effect on the nature of the interaction and the opportunities this offers for LREs and learning than the overall proficiency of the dyad. Much research has been devoted in the past few years to the study of collaborative dialogue as an opportunity for second-language (L2) learning in learner–learner interaction (e.g. Swain and Lapkin 1998, 2002; Williams 2001; Lapkin et al. 2002; Watanabe and Swain 2007; Kim 2008). This research has provided evidence that when learners collaborate to solve linguistic problems, they engage in language-mediated cognitive activities that serve to build linguistic knowledge. By pooling their resources they are able to co-construct new language knowledge that can be internalized and transformed into an individual resource. Most of this research has been conducted in the classroom and with a clear pedagogical aim; therefore, it has focused almost exclusively on learner–learner interaction. In this study, I analyse collaborative dialogue in a different context by comparing intermediate and advanced learner–learner and learner– native speaker (NS) interaction. I examine lexical problems arising in interaction due to learners’ holes in L2 vocabulary knowledge and the dialogue they generate as learners collaborate with either native or non-native interlocutors in solving them. The focus of the study is therefore on lexis. The final aim is to identify to what extent collaborative dialogue may be spontaneously generated in learner–NS as well as learner–learner interaction, and how a NS 230 LEARNER–LEARNER AND LEARNER–NS INTERACTION interlocutor may affect the frequency and the nature of this collaborative knowledge-building activity. COLLABORATIVE DIALOGUE The role of collaboration in L2 learning has been emphasized by the study of interaction from a sociocultural theory of mind perspective (see, among others, Donato 1994, 2000; Lantolf and Appel 1994; Lantolf 2000; Ohta 2000, 2001; Swain 2000; Lantolf and Thorne 2006). Sociocultural researchers, building on Vygotsky (1978, 1987), claim that human cognitive development originates in social activity. ‘Psychological processes emerge first in collective behaviour, in co-operation with other people, and only subsequently become internalized as the individual’s own possessions’ (Stetsenko and Arievitch 1997: 161). From this perspective, linguistic knowledge can be collaboratively constructed through socially situated cognitive processes. Language learning takes place as the learner, through a gradual process of internalization, becomes able to make independent use of what is originally co-constructed language knowledge, that is, ‘able to use the language of others’ (Swain and Lapkin 1998: 321). From a sociocultural lens, language is a powerful semiotic tool that serves to mediate cognitive activity. During collaborative problem-solving, learners engage in knowledge-building language-mediated activities, such as formulating and testing hypotheses, or offering and assessing new input. These activities can be observed in their dialogue. Their talk is ‘an enactment of cognitive activity’ (Swain and Lapkin 1998: 322). This cognitive activity and the knowledge it builds represent language learning in progress (Donato 1994; Swain and Lapkin 1998). On this basis Swain concludes that collaborative dialogue, the dialogue that occurs between learners as they collaborate to solve linguistic problems, is ‘knowledge-building dialogue. In the case of our interests in second language learning, it is dialogue that constructs linguistic knowledge . . . It is where language use and language learning can co-occur’ (Swain 2000: 97). COLLABORATIVE DIALOGUE AND LEARNER–LEARNER INTERACTION Research on learner–learner interaction has operationalized the construct of collaborative dialogue through language-related episodes (LREs), defined as ‘any part of a dialogue where the students talk about the language they are producing, question their language use, or correct themselves or others’ (Swain and Lapkin 1998: 326). A distinction can be made between form-based LREs, when students’ attention is focused on grammar, and lexis-based LREs, that involve learners seeking L2 vocabulary or choosing among competing lexical items. A. FERNÁNDEZ DOBAO 231 Previous research has provided considerable evidence confirming that grammatical and lexical knowledge co-constructed in LREs tends to be retained by learners and therefore L2 learning occurs in collaborative dialogue (e.g. Swain and Lapkin 1998, 2002; Williams 2001; Lapkin et al. 2002; Tocalli-Beller and Swain 2005; Watanabe and Swain 2007; Kim 2008; Zeng and Takatsuka 2009). However, these previous studies have also documented considerable variation in the number of LREs produced by different pairs of learners while performing the same task (e.g. Storch 1998, 2002a,b; Swain and Lapkin 1998, 2001; Watanabe and Swain 2007). This has raised researchers’ interest in those learners’ characteristics that may have an influence on collaborative dialogue. The effect of the learner’s proficiency level has received much consideration. In 1999, looking at spontaneous attention to form in the classroom, Williams observed that the frequency of LREs produced by learners from four different courses increased with the level of the course (Williams 1999). Leeser (2004) compared high–high, high–low, and low–low proficiency level pairs, and found that the overall proficiency of the pair impacted not only the number of LREs produced, but also the focus of the LREs and the percentage of correctly resolved episodes. Watanabe and Swain (2007, 2008) examined the same learners interacting with both lower-level and higher-level learners, and Kim and McDonough (2008) compared the same learners working in dyads with same level and higher-level peers. Both studies revealed a positive relationship between learners’ proficiency level and LRE production. In a series of related studies on the nature of interaction and pair dynamics, Storch identified four different patterns of interaction: collaborative, expert/ novice, dominant/passive, and dominant/dominant (Storch 2001, 2002a,b, 2004). In the first two patterns there is frequent co-construction of knowledge and scaffolded assistance. Learners work collaboratively and create opportunities for L2 learning. However, Storch pointed out that ‘not all students work collaboratively when assigned to work on language tasks in pairs’ (Storch 2002a: 147). Learners in dominant/dominant pairs are less receptive to each other’s contributions and in dominant/passive dyads one learner takes control of the activity leaving little opportunity for the other learner to participate. This research has also shown that these non-collaborative patterns are less favourable for LRE occurrence and L2 learning. Some studies have tried to establish a relationship between differences in learners’ proficiency level and patterns of interaction (Watanabe and Swain 2007, 2008; Kim and McDonough 2008; Watanabe 2008). These studies have found that proficiency differences do not always account for the nature of pair interaction. More importantly, when considering the two factors together, pair dynamics seem to have a stronger impact on the occurrence of LREs than proficiency. This means that learners can benefit from working with both higher- and lower-level learners, as long as the two members of the dyad adopt a collaborative orientation. Building on Donato (1988, 1994), Storch (2004) examined variability using the framework of activity theory (Leont’ev 1981, Vygotsky 1981). Her study 232 LEARNER–LEARNER AND LEARNER–NS INTERACTION revealed that differences in patterns of behaviour were related to learners’ attitudes towards tasks and pair work, to their motives and goals in completing a given task, and to whether these goals converged or diverged. For instance, she found that the only learner adopting a passive role in her study had a strong preference for individual rather than group work, whereas the learners in the collaborative pattern emphasized the benefits of pair work and their intention to complete the task to the best of their ability. In line with Coughlan and Duff’s claim that a task is only ‘a kind of behavioural blueprint’ (Coughlan and Duff 1994: 175), Storch concluded that ‘it is the learners’ definition of the situation and, specifically, their perceived goals and roles that determine how an activity will be carried out (and not, for example, the task type)’ (Storch 2004: 473–4). RESEARCH ON LEARNER–NS INTERACTION From an interactionist perspective, a considerable amount of research has been devoted to compare learner–learner and learner–NS interaction (e.g. Varonis and Gass 1985; Pica et al. 1996; Shehadeh 1999; Mackey et al. 2003; Sato and Lyster 2007). Although opportunities for learners to interact with NSs tend to be highly valued, particularly in foreign-language contexts, the results of some of these studies suggest that learner–NS interaction is not always more beneficial for learning than learner–learner interaction. Varonis and Gass (1985) found that negotiation of meaning was more frequent in dyads involving two learners than in learner–NS dyads. Mackey et al. (2003) observed that, although NSs provided more feedback, the feedback offered by other learners created more opportunities for modified output. Similarly, Sato and Lyster (2007) found that native and non-native interlocutors provided different types of feedback, that learners modified their output more frequently in response to other learners’ feedback, and that they felt more comfortable asking questions about language to other learners than to NSs. In sum, interactionist research has shown that learner–learner and learner– NS interactions offer different opportunities for negotiation of meaning, feedback, and modified output. It has confirmed that learners interact differently with other learners than with NSs. However, as already noted, the study of collaborative dialogue has focused almost exclusively on learner–learner interaction. No study has compared the collaborative problem-solving activity that occurs in learner–learner dyads and learner–NS dyads. This is the aim of the present research. From a sociocultural perspective learning can occur in collaboration with more capable individuals, ‘experts’, who have knowledge or skills the learner has not yet acquired (Vygotsky 1987). Learner–learner interaction provides opportunities for L2 learning because it allows learners to act as both experts and novices. Because no two learners have the same weaknesses and strengths, they can help each other solve their language-related problems and, working together, achieve a level of performance that is above their A. FERNÁNDEZ DOBAO 233 individual level of competence (Donato 1994; Ohta 2000, 2001). NSs, because of their higher command of the language, can also act as the more knowledgeable interlocutor. They have the linguistic resources needed to act as experts and provide the feedback and assistance the learner needs for L2 learning to take place. This study investigates whether and how NSs collaborate with the learner in the solution of their linguistic problems and the co-construction of new language knowledge. To study this issue I compare learner–learner and learner–NS interaction. Looking for evidence of collaborative dialogue, I examine how learners and their native and non-native interlocutors deal with lexical difficulties encountered during the performance of a communicative oral task. Attention is therefore focused on lexical LREs spontaneously generated during a meaning-oriented task. Since previous research has regularly found that the learners’ proficiency level plays an important role in both the frequency and the nature of LREs, this factor is also taken into account. Both intermediate and advanced learner–learner and learner–NS interactions are analysed here looking for an answer to the following questions: 1 How does the presence of an NS interlocutor affect the frequency and the nature of lexical LREs? (a) Are lexical LREs more frequent in learner–NS than in learner–learner interaction? (b) Are lexical LREs more likely to be correctly resolved in learner–NS than in learner–learner interaction? 2 How does the learner’s proficiency level affect the frequency and the nature of lexical LREs? (b) Are lexical LREs more frequent in advanced than in intermediate-level learner dyads? (c) Are lexical LREs more likely to be correctly resolved in advanced than in intermediate-level learner dyads? METHOD Participants The participants in this study were 24 learners of English as a foreign language and 8 English NSs. The learners were undergraduate and graduate students in the English Department of a Spanish university. They were all Spanish NSs.1 There were 23 females and 1 male, and their ages ranged from 19 to 26 years. They were selected, among a larger number of volunteers, on the basis of their course level and their results on a placement test.2 The 12 intermediate-level learners were first- and second-year students of the English language and literature programme. The 12 advanced-level learners were selected among fourth-year undergraduate students and first-year graduate students. 234 LEARNER–LEARNER AND LEARNER–NS INTERACTION The eight English NSs, five males and three females, were international students at this same university, who freely volunteered to participate in the study. Five of them were British and three American. They had been in Spain for at least one semester and were all taking intermediate or advanced-level Spanish courses at the Modern Languages Centre. The 32 participants were paired in four dyads of intermediate-level learners, four dyads of advanced-level learners, four dyads of intermediate-level learners and NSs, and four dyads of advanced-level learners and NSs. The pairs were established at random. Since all the English language learners were enrolled in the same programme and taking similar level courses, some of them knew each other, but they were not friends or even classmates. Only one of the NSs and her partner were acquainted. Task Most previous studies have turned to collaborative tasks, such as jigsaw, dictogloss, or text reconstruction tasks, to elicit collaborative dialogue. Collaborative tasks require learners to work in pairs and produce one joint written text. This collaborative writing activity pushes learners to reflect on language while creating meaning, to collaborate in the solution of their linguistic problems, and to engage in collaborative dialogue (Swain 2001: 59–60). Yet, some studies have found that attention to language and LREs can be spontaneously generated by the learners themselves during meaning-oriented tasks (Williams 1999, 2001; Foster and Ohta 2005; Gass et al. 2005). Following this research, in the present study a meaning-oriented spot-the-difference task was designed to elicit both learner–learner and learner–NS interaction. In order to compare how native and non-native interlocutors collaborate with the learner in the solution of their linguistic problems, no pedagogical intervention was made to orient participants’ attention to language or to prompt LREs. Spot-the-difference tasks have been frequently used in previous research to create referential conflicts and encourage interaction. In the task designed for the purposes of this study, each member of the dyad was given a different version of the same picture strip story (see online supplementary material for appendix). The two participants were instructed to describe and compare their pictures in as much detail as possible in order to identify the differences existing between the two versions of the story without looking at each other’s images. The use of pictures provides well-defined and constant content across dyads. The need to identify the non-shared referents pushes the two members of the dyad to reach mutual agreement on meaning even when the lexical item desired to convey this meaning is not part of their vocabulary. Since the focus of the study is on lexis, the task was designed to create lexical problems. The differences, a total of 32, involved referents such as ‘badge’, ‘black eye’, ‘moustache’, ‘holding hands’, ‘to knock’, ‘wallpaper’, ‘to roll up’, or ‘freckles’. These were selected after piloting the materials to guarantee they A. FERNÁNDEZ DOBAO 235 would pose frequent problems to both intermediate and advanced-level learners. These 32 referents were present in version A of the picture strip and missing in version B (see online supplementary material for appendix). The participant who received version B had to write down the differences. The instructions specified that there was a minimum of two and a maximum of five differences in each picture, emphasizing that the goal of the task was to identify all the differences, that is, to identify and note all the missing elements in version B of the story. In learner–learner dyads, the two versions were distributed at random. In learner–NS interaction, the learner received version A and the NS version B. This condition was set to facilitate the comparative analysis of the data. In this way the learner had more information to convey than the NS and would therefore encounter more vocabulary problems. Furthermore, according to Yule and Macdonald (1990), this unequal distribution of information helps to prevent the more proficient speaker from dominating the conversation and, in this way, should facilitate a more balanced interaction between learners and NSs. Procedure The data were collected in a university classroom after school hours and in individual sessions. First, participants were given the task instructions and asked to complete a practice session, with a different set of pictures. Once familiarized with the procedure, they completed the main task. Learner–learner dyads spent an average of 28 minutes on the main task and learner–NS dyads 22 minutes. During the task the two members of the dyad sat at a table facing each other. A small screen was placed between them, so they could not see each other’s pictures, but could still see each other’s faces and make full use of nonverbal communication. Performance was audio and video-recorded in order to register gestures, facial expressions, and paralinguistic devices used by the two participants to deal with the communicative and lexical difficulties encountered. The spot-the-difference task was followed by a stimulated recall interview (Gass and Mackey 2000). This was conducted by the researcher immediately after the performance of the task and in the participants’ first language (L1). Learner–learner pairs completed the interview together, whereas learner–NS dyads worked separately. The NS did the interview in English and immediately afterwards the learner completed the procedure in Spanish. Before the actual stimulated recall session, the researcher explained the purpose of the study. Participants were then instructed to listen to their oral performance of the task and to stop the audio whenever a lexical difficulty had been experienced. Problems were discussed and participants were asked to identify the nature of the problem, the strategies used to compensate for it, and whether they considered mutual agreement on meaning had been 236 LEARNER–LEARNER AND LEARNER–NS INTERACTION successfully achieved or not. This particular procedure was followed in order to stimulate participants’ recall of their thoughts during task performance and to elicit their reflections on the specific problems encountered. DATA ANALYSIS Learner–learner and learner–NS interactions were transcribed and then analysed. First, I had to identify the lexical problems occurring in interaction due to learners’ holes in L2 vocabulary knowledge. Previous research has shown that when learners notice a lexical hole—that is, the lexical item they desire to use in order to communicate a given meaning is not yet part of their vocabulary or is not shared with their interlocutor—they resort to the use of communication strategies (CSs). They can choose between avoiding the originally intended message, or attempting an alternate means of expression, such as a synonym, a related term, a circumlocution, an L1 transfer, or a gesture (Tarone 1981; Færch and Kasper 1983; Dörnyei and Kormos 1998). In the present study, interactional data were analysed by searching for CSs as evidence of lexical difficulties. The total number of CS episodes identified in the data represents the total number of lexical problems encountered by the dyads during the performance of the spot-the-difference task. In order to compare how native and non-native interlocutors collaborate with the learner when lexical difficulties arise, these CS episodes were further analysed and the LREs they included were identified. These LREs were classified into two categories: successfully resolved and unsuccessfully resolved. Since the focus of the study is on lexis, only those LREs that occurred as a result of a lexical problem, and therefore within a CS episode, were taken into account. This analysis made it possible to consider the number of lexical LREs produced by each dyad in relation to the actual number of lexical difficulties encountered during their interaction, and in this way facilitated the comparative analysis of learner–learner and learner–NS interaction. The data analysis was conducted by the researcher following the participants’ retrospective comments during the stimulated recall interview. Only when the learners confirmed that they were aware that a lexical difficulty had been experienced was a CS episode identified. The participants’ retrospective comments were also taken into account for the identification and classification of LREs. I present some excerpts from both learner–learner and learner– NS interactions to illustrate how these analyses were conducted. Communication strategy episodes In Example 1, an advanced-level learner, Ángela, wants to communicate the meaning ‘rolled up’, but this lexical item is not yet part of her vocabulary. To compensate for this problem she uses a nonverbal strategy: she rolls up her own sleeve. As confirmed by the learners’ retrospective comments, this CS makes possible the successful communication of the meaning ‘rolled up’, A. FERNÁNDEZ DOBAO 237 even though this lexical item is not part of either Ángela’s or Vero’s vocabulary. The dialogue the two learners generate as they collaborate to establish mutual agreement on the meaning ‘rolled up’ constitutes a CS episode. Example 1 Advanced learner–learner interaction: 1 ÁNGELA: with: a: [. . . mm . . . I don’t know . . . how do you [rolling up her sleeve call this:, . . . you know? . . . he has:, . . . his shirt, [this way [pointing to her rolled up sleeve 2 VERO: mm: heh Learners’ stimulated recall interview:3 ANGELA: I wanted to say rolled up and I didn’t know how to say it RESEARCHER: and did you understand her? VERO: hm because of her gestures RESEARCHER: uhuh, because of the gesture, and do you know how to say rolled up? VERO: no, I don’t In the next example Carmen is dealing with the same problem. She uses ‘brought back’ as an alternate means of expression to compensate for her lack of knowledge of the lexical item ‘rolled up’. In the stimulated recall interview the NS interlocutor confirmed his correct understanding of Carmen’s message and explained he had not tried to correct her or offer the lexical item she was missing precisely because meaning had been successfully communicated. Example 2 Advanced learner–NS interaction: 1 CARMEN: and the sleeve of his shirt is e::h . . . tch is e::h eh it’s a long sleeve, but it is, e:h brought back because probably it is hot 2 SCOTT: hm 238 LEARNER–LEARNER AND LEARNER–NS INTERACTION Learner’s stimulated recall interview: CARMEN: here, rolled up, I didn’t know how to say it and I said pulled or brought back . . . I looked for something that could be understood NS’s stimulated recall interview: RESEARCHER: did you notice this difference? SCOTT: yeah, I think I wrote it down . . . yeah RESEARCHER: is brought back correct? how would you say it? SCOTT: tch, no, no, it would be rolled up RESEARCHER: but you didn’t try to correct her SCOTT: no, because I got it As illustrated in these two examples, a CS episode consists of a variable and unpredictable number of turns focused on one single lexical difficulty. It begins when the learner realizes that the lexical item desired to communicate an intended meaning is not available and a CS needs to be used. It ends when the two interlocutors reach mutual agreement on this meaning and/or decide to move on to the next topic in the conversation. When two different lexical difficulties were dealt with within the same segment of interaction, two different CS episodes were identified. LREs A CS illustrates what the learner is trying to say but cannot say with correct target-like vocabulary. It is a non-desired, usually erroneous or inaccurate expression that can also be seen as a signal of a problem. It serves for the learner to externalize the lexical problem encountered and, in this way, may act as a trigger for collaborative dialogue. This means that within a CS episode participants may choose to focus only on meaning and the successful communication of the message (as seen in Examples 1 and 2), or also on lexical accuracy and vocabulary knowledge (see Examples 3, 4, and 5). They may collaborate to search for a new lexical item, to choose among competing vocabulary items, or to clarify the meaning of an unknown word. This kind of collaborative problem-solving activity constitutes a lexical LRE, as defined by Swain and Lapkin (1998). A CS episode may therefore evolve into a lexical LRE. This is what happens in Examples 3, 4, and 5.4 In Example 3, Sara uses a gesture and a circumlocution strategy to compensate for her lack of knowledge of the lexical item ‘chin’. Olga understands the message and offers the word ‘chin’. In turn 3 we can see Sara using this word. With Olga’s help, Sara has become able to use a previously unavailable lexical A. FERNÁNDEZ DOBAO 239 item. The two learners collaborate not only to reach mutual agreement on the meaning ‘chin’, but also to build the lexical knowledge needed to convey this meaning with correct target-like vocabulary. The result is a CS episode that also constitutes a lexical LRE. Example 3 Advanced learner–learner interaction: 1 SARA: e::h the child [is:: . . . touching his::, mm mm [touching her chin . . . tch the place behind the mouth, heh heh 2 OLGA: chin? 3 SARA: his chin, sı́ (‘yes’), yes, Learners’ stimulated recall interview: SARA: I couldn’t remember how to say chin RESEARCHER: when she said ‘the place behind the mouth’, what did you understand? OLGA: yeah, I understood her, because she used her hands . . . yes, because you were with the hand like this, so I understood it was the chin In Example 4, we can see this kind of collaborative problem-solving activity between an intermediate-level learner and an English NS. The NS, Mike, offers two different lexical items to solve the learner’s problem: ‘braces’, the British term, and ‘suspenders’, the American one. He explains the different value of each word, they consider the two different options, and finally, the NS being British, they choose to use ‘braces’ instead of ‘suspenders’. Example 4 Intermediate learner–NS interaction: 1 PAULA: an::d [he’s wearing:: . . . these things that you [drawing braces on her chest can use when you:: . . . you you wear . . . e::h . . . short e::h trousers . . . [a::nd [drawing braces on her chest 2 MIKE: ah! so that your trousers don’t fall down? 240 LEARNER–LEARNER AND LEARNER–NS INTERACTION 3 PAULA: yes 4 MIKE: braces 5 PAULA: braces? 6 MIKE: right 7 PAULA: a::nd 8 MIKE: or if you’re American suspenders 9 PAULA: hm hm . . . well I’m not American 10 MIKE: no I am not either 11 PAULA: not English heh heh heh . . . braces Learner’s stimulated recall interview: PAULA: braces I didn’t know how to say it . . . he said . . . ‘braces?’ RESEARCHER: yeah NS’s stimulated recall interview: MIKE: yeah she was trying to say . . . yeah braces Successfully and unsuccessfully resolved LREs The LREs were classified into two categories: successfully resolved and unsuccessfully resolved. The LRE was successfully resolved when the two members of the dyad were able to build the lexical knowledge needed to convey the originally intended message with correct target-like vocabulary, in other words, when they found the L2 lexical item that was originally missing. Therefore the previous episodes, Examples 3 and 4, constitute instances of successfully resolved lexical LREs. The LRE was unsuccessfully resolved when participants were not able to find the word they were looking for or agreed on an incorrect word. This is what happens in Example 5. The learner is trying to communicate the meaning ‘badge’, but the NS understands ‘buttonhole’ and this is the lexical item she offers. Neither the learner nor the NS are able to recognize the misunderstanding. Example 5 Intermediate learner–NS interaction: 1 ANA: [in the:: jacket, he has . . . some kind of . . . e::h [pointing to her chest 2 EMILY: an emblem? a letter? A. FERNÁNDEZ DOBAO 3 241 ANA: [like a button, in a:: side, I don’t know how to [pointing to her chest explain it. [. . .] 8 EMILY: oh! [okay okay . . . so he has a, . . . a buttonhole, [nodding Learner’s stimulated recall interview: ANA: I had no idea how to say that [. . .] RESEARCHER: did you think she had understood you? she had understood there was a badge there? ANA: yeah, not at the beginning, but then yeah, I thought she had understood me NS’s stimulated recall interview: EMILY: yeah, I thought she meant a buttonhole, RESULTS I present now the results of the quantitative analyses conducted to answer the questions that prompted the study: (i) how does the presence of a NS interlocutor affect the frequency and the nature of lexical LREs? and (ii) how does the learner’s proficiency level affect the frequency and the nature of lexical LREs? The nature and patterns of interaction of two selected dyads, a learner– NS and a learner–learner dyad, are also examined. Learner–NS and learner–learner interaction Table 1 presents the total number of CS episodes and lexical LREs in learner– NS and learner–learner interaction. The number of CS episodes indicates the number of lexical difficulties encountered, which means that lexical problems were more frequent in learner–learner dyads (268 CS episodes) than in learner–NS dyads (213 CS episodes). LREs were, however, more frequent in learner–NS interaction (93 LREs) than in learner–learner interaction (63 LREs).5 Of the lexical problems encountered by learner–NS dyads, 43.66 per cent resulted in an LRE, whereas only 23.51 per cent of the problems encountered by learner–learner dyads served as a trigger for collaborative dialogue. The results of the chi-square test revealed that this difference was statistically significant (2 = 21.999, df = 1, p < .001). 242 LEARNER–LEARNER AND LEARNER–NS INTERACTION Table 1: CS episodes and LREs in learner–NS and learner–learner dyads Learner–NS dyads (n = 8) CS episodes LREs Successfully resolved LREs Unsuccessfully resolved LREs a Learner–learner dyads (n = 8) Sum Median Per centa Sum Median Per centa 213 93 86 7 26.50 10.50 10.00 1.00 43.66 92.47 7.53 268 63 51 12 32.50 8.00 6.50 1.00 23.51 80.95 19.05 Percentage of LREs per CS episodes. In Table 1, we can also see the number and percentage of successfully and unsuccessfully resolved lexical LREs. Learner–NS dyads were able to reach a correct solution for 92.47 per cent of their LREs and learner–learner dyads for 80.95 per cent. This difference was also statistically significant (2 = 4.661, df = 1, p = .031). In sum, the findings related to the first question of the study revealed that lexical LREs were both more frequent and more likely to be correctly resolved in learner–NS than in learner–learner interaction. The presence of a NS interlocutor had a positive effect on both the frequency and the nature of collaborative dialogue. As illustrated in Examples 1 and 2, the spot-the-difference task used for the purposes of this study could be successfully completed without the kind of collaborative knowledge-building activity that takes place in LREs. Since no pedagogical intervention was made to orient participants’ attention to language, when lexical LREs occurred, they were spontaneously generated by the participants. Although only 23.51 per cent of the CS episodes produced in learner–learner interaction evolved into an LRE, learner–NS dyads engaged in collaborative dialogue in almost half of their CS episodes, 43.66 per cent. This means that, when lexical difficulties arose, NSs quite often collaborated with the learner not only to achieve the successful communication of the message, but also to build the lexical knowledge needed to convey this message with accuracy and precision. In other words, they used their linguistic expertise to help learners enhance their use and knowledge of the language, even though this was not a specific requirement of the communicative task at hand. Learners were not always able to provide each other with this kind of linguistic assistance. As seen in Example 1, when a lexical hole was noticed and neither of the two learners had the knowledge to fill it, they focused on the successful communication of the message, without explicitly looking for the missing L2 lexical item. This explains the lower frequency of lexical LREs in learner–learner interaction and also the relatively low percentage of A. FERNÁNDEZ DOBAO 243 unsuccessfully resolved LREs even for this condition. It seems that lexical LREs were spontaneously generated by the learners only when they believed they had the necessary knowledge to solve them successfully. Advanced and intermediate learner interaction The results of the comparative analysis of advanced and intermediate-level learner dyads are presented in Table 2. Intermediate learner–learner and learner–NS dyads produced a total of 259 CS episodes whereas advanced learner–learner and learner–NS dyads produced 222 CS episodes. Lexical LREs were also more frequent when intermediate-level learners participated in the interaction: 37.45 per cent of their CS episodes evolved into an LRE, whereas only 26.58 per cent of the CS episodes produced by advanced learner dyads involved an LRE. This difference was statistically significant (2 = 6.452, df = 1, p = .011). The percentage of successfully resolved LREs was, however, higher for dyads with advanced level learners: 91.53 per cent versus 85.57 per cent. But the results of the chi-square test revealed that this difference was not statistically significant (2 = 1.218, df = 1, p = .270). In sum, lexical LREs were neither more frequent nor more likely to be correctly resolved in advanced than in intermediate learner–NS and learner–learner interaction. In order to understand these results, a more detailed analysis of each individual dyad was conducted. In Table 3, we can see the number of CS episodes, the number and percentage of CS episodes evolving into lexical LREs, and the proportion of successfully and unsuccessfully resolved LREs for each individual dyad. These data revealed a considerable range of variation between dyads. As already noticed, individual variation has been frequently observed in previous research and explained in relation to dyads’ patterns of interaction (Storch 2001, 2002a,b, 2004; Watanabe and Swain 2007; Watanabe 2008). Following this research, Table 2: CS episodes and LREs in intermediate and advanced learner dyads Advanced learner dyads (n = 8) CS episodes LREs Successfully resolved LREs Unsuccessfully resolved LREs a Intermediate learner dyads (n = 8) Sum Median Per centa Sum Median Per centa 222 59 54 5 26.50 8.50 7.50 0.50 26.58 91.53 8.47 259 97 83 14 28.00 10.50 9.50 1.00 37.45 85.57 14.43 Percentage of LREs per CS episodes. 244 LEARNER–LEARNER AND LEARNER–NS INTERACTION Table 3: CS episodes and LREs by dyad CS episodes LREs Advanced learner– NS dyads Bárbara and Sean Carmen and Scott Isabel and Mary Raquel and Anne Intermediate learner– NS dyads Ana and Emily Carla and Simon Paula and Mike Sara and Larry Advanced learner– learner dyads Ángela and Vero Luisa and Toñi Marı́a and Mónica Sara and Olga Intermediate learner– learner dyads Carolina and Lola Fran and Celia Sofı́a and Clara Victoria and Marta a Successfully resolved LRES Unsuccessfully resolved LREs n n Per centa n Per cent n Per cent 14 34 20 23 8 4 10 12 57.14 11.76 50.00 52.17 8 3 10 12 100 75.00 100 100 0 1 0 0 0.00 25.00 0.00 0.00 28 25 41 28 16 10 22 11 57.14 40.00 53.66 39.29 13 9 21 10 81.25 90.00 95.45 90.91 3 1 1 1 18.75 10.00 4.55 9.09 28 37 41 25 3 4 9 9 10.71 10.81 21.95 36.00 3 3 8 7 100 75.00 88.89 77.78 0 1 1 2 0.00 25.00 11.11 22.22 40 45 25 27 10 18 7 3 25.00 40.00 28.00 11.11 9 12 6 3 90.00 66.67 85.71 100 1 6 1 0 10.00 33.33 14.29 0.00 Percentage of LREs per CS episodes. I decided to analyse the patterns of interaction of the two pairs with the highest deviation: Carmen and Scott, the learner–NS dyad with the lowest frequency of LREs, and Fran and Celia, the learner–learner dyad with the highest frequency of LREs. Carmen and Scott: non-collaborative pattern of interaction Carmen and Scott spent 21 minutes on the task and produced a total of 3,634 words, higher than the average for their condition (2,901 words). This seems to A. FERNÁNDEZ DOBAO 245 suggest they were highly involved in the task. However, the analysis of their interaction showed that the learner, Carmen, did most of the talking. She produced 3,191 words, 87.81 per cent of the total, and her ratio of words per turn was 27.27. As illustrated in Example 6, the interaction was characterized by Carmen’s long monologues, which previous research has identified as a trait of a non-collaborative pattern of interaction (Storch 2002a; Watanabe and Swain 2007; Watanabe 2008). Example 6 Advanced learner–NS interaction: 1 CARMEN: the father e::h seems to be sleeping in the:: . . . in the couch . . . with well because I think that he has the:: he’s is is that his mouth is:: . . . is open even snoring probably and he has a:: a:: . . . newspaper is in his in his right hand but he’s not reading it just . . . lying in the sofa and . . . e::h in relation to the clothes he has a:: trousers and one of the::se e::h . . . tch e::h [jerseys but without sleeves a heh [pointing to her shoulders 2 SCOTT: ah! [hm hm [nodding 3 CARMEN: and the probably he has a:: a shirt . . . and and shoes and he has the: . . . one of his hands is closed and the other one is open with the with the:: . . . tch newspaper and he has e:h a big very big very big moumoustache and the:: . . . tch the eyes the eye that I can see is closed and he has also very big . . . ear and the:: the hair is is:: is black 4 SCOTT: very good . . . beautiful 5 CARMEN: okay so [the next one now? [pointing to the pictures 6 SCOTT: mhm 246 LEARNER–LEARNER AND LEARNER–NS INTERACTION Learner’s stimulated recall interview: RESEARCHER: when you say ‘jersey without sleeves’? CARMEN: yeah waistcoat hm RESEARCHER: do you think he understood you? CARMEN: yeah, In most of the dyads the two participants collaborated to identify the differences by asking each other questions about their pictures, completing each other’s descriptions, and reaching joint decisions about the missing referents. Carmen and Scott approached the task differently. Carmen took responsibility for describing the pictures without asking Scott to comment on his own images or the differences he was noticing. Scott hardly ever tried to complete the learner’s account of the story or overtly compare the two sets of pictures. As illustrated in Example 6, when he noticed a difference he just wrote it down without discussing it with Carmen. Two differences were spotted in this segment of interaction, the father’s moustache and his waistcoat—Carmen used ‘one of these jerseys but without sleeves’ as a CS to compensate for her lack of knowledge of the word ‘waistcoat’ (American English, ‘vest’). Instead of collaborating to reach one common goal, each participant assumed a different role and a separate set of responsibilities. In Examples 6, 7, and 8, we can see that Scott’s contributions were not only short and scarce, but they also reveal little engagement with the learner’s contributions. Overlapping talk, other-repetition, or co-construction of sentences, which according to previous research characterize a collaborative orientation (Storch 2002b; Watanabe 2008; Watanabe and Swain 2008), were extremely rare in this interaction. Scott’s contributions were mostly short acknowledgements of what the learner had just said: ‘mhm’, ‘hm hm’, ‘very good’, and nodding gestures. In Storch’s model, this is considered as a passive trait (Storch 2002a,b). Even when Carmen checked for comprehension uttering a CS with rising intonation, such as ‘a jersey without sleeves?’ in Example 7, Scott made no open attempt to collaborate with her, neither offering the missing L2 vocabulary item nor an affirmative answer to the comprehension check. In Example 8, Carmen described the wall as ‘decorated with flowers’ because she couldn’t remember the word ‘wallpaper’. Again, the rising intonation indicates she is checking for comprehension and serves as a signal of a problem. Scott understood the learner’s message and this time provided a signal to confirm his understanding, but without trying to go beyond the minimum demands of the task to offer the missing lexical item. A. FERNÁNDEZ DOBAO 247 Example 7 Advanced learner–NS interaction: 1 CARMEN: well I don’t know if it is a:: [a jacket of a jer- or [pointing to her chest a jersey . . . without sleeves? . . . and yes, he has two pockets. 2 SCOTT: [mm [nodding Example 8 Advanced learner–NS interaction: 1 CARMEN: I can see only:, a part of the house, very funny because the: the walls are . . . painted or: decorated wi:th e::h flowers? 2 SCOTT: mm:: Learner’s stimulated recall interview: CARMEN: when I said ‘painted’ ‘painted’, because it could be painted, but when I said ‘decorated’, I wanted to say wallpapered wallpapered . . . I couldn’t remember the word ‘wallpaper’ . . . but I was thinking about wallpaper NS’s stimulated recall interview: RESEARCHER: did you imagine something like this? SCOTT: yeah yeah [. . .] RESEARCHER: how would you describe this? SCOTT: yeah, flowery wallpaper The low levels of equality and mutuality displayed in this interaction suggest a non-collaborative orientation, which previous research has also found to correlate with low frequency of LREs (Storch 2001, 2002a,b; Watanabe and Swain 2007; Watanabe 2008).6 During the stimulated recall interview, Scott confirmed he had noticed Carmen had experienced frequent lexical 248 LEARNER–LEARNER AND LEARNER–NS INTERACTION difficulties. He also explained he had not tried to help or correct her because it had not been necessary, since ‘her English was very good’ and he ‘could understand everything she was saying’ (see also Example 2). During the performance of the task Scott offered constant positive support to Carmen with comments such as ‘good’, ‘right’, ‘very good’, or ‘brilliant’. At the end of the session, he even congratulated her on her English proficiency level. This may also help to understand why LREs were less frequent in advanced learner dyads than in intermediate learner dyads. Given the communicative nature of the task, it may be that advanced-level learners received less assistance from their interlocutors precisely because they were perceived as more competent and effective communicators than their intermediate counterparts. Carmen’s interpretation of the task seems to match Scott’s. As illustrated in the following extract from the stimulated recall interview, Carmen’s main goal was to make herself understood, so that Scott could identify the differences existing between the two sets of pictures. Since communication of meaning could be easily achieved without the need of correct target-like vocabulary, Carmen and Scott tended to focus on meaning ignoring linguistic accuracy. Carmen did not ask Scott to rephrase her strategic utterances and Scott did not try to correct Carmen’s non-target-like use of the language. Learner’s stimulated recall interview: CARMEN: yeah, you are not sure, but you try, I think to myself, I am going to try and see if he can understand me, if he understands me I feel more confident, even though, yes, I know there is a better word, but, I mean, since I don’t need to . . . because he understood me after all . . . and that’s what it is about . . . so:: Fran and Celia: collaborative pattern of interaction Fran and Celia spent more time completing the task than most of the other dyads, 35 minutes, producing a total of 4,168 words, which illustrates their high involvement in the task (the average for their condition was 28 minutes and 3,622 words). They described the pictures in great detail and said explicitly that they wanted to do their best and identify as many differences as possible. They identified a total of 30 differences, the highest number among all the dyads.7 In their interaction the amount of talk was quite balanced. Fran, who had version A of the story, produced 2,166 words and Celia 2,002. The ratio of words per turn was 5.04 for Fran and 4.67 for Celia. Unlike Carmen and Scott, Fran and Celia shared responsibility for both the description of the pictures and A. FERNÁNDEZ DOBAO 249 the identification of the differences. Celia gave detailed accounts of the pictures on her strip, completing Fran’s descriptions whenever necessary. As illustrated in Examples 9 and 10, this facilitated the comparison of the two strips, making it possible for the two learners to reach joint decisions on the differences existing between their two versions of the story. Example 9 Intermediate learner–learner interaction: 1 FRAN: the newspaper a- appears o::n on the:: . . . on the floor 2 CELIA: on the floor! . . . no eh he:: . . . mine has still the:: 3 FRAN: the newspaper in the hand? 4 CELIA: the newspaper in the hand. 5 FRAN: ah! Learners’ stimulated recall interview: FRAN: that he dropped the newspaper, or that the newspaper had been dropped, but I don’t know how to say that in English Example 10 Intermediate learner–learner interaction: 1 FRAN: her right eye is black, 2 CELIA: mm what? 3 FRAN: her:: her right eh eye is black 4 CELIA: his right eye is . . . black? 5 FRAN: his right eye 6 CELIA: [no, mine not, mine is normal . . . eh a black eye? no? [shaking her head 7 FRAN: [yeah black eye [nodding Learners’ stimulated recall interview: RESEARCHER: did you know that a black eye is ‘black eye’ in English? FRAN: yeah, because I couldn’t remember that, but then she told me 250 LEARNER–LEARNER AND LEARNER–NS INTERACTION CELIA: yeah RESEARCHER: so when you said ‘his right eye is black’? FRAN: at the beginning I was describing it, because it’s black on the picture These excerpts also illustrate the two learners’ engagement with each other’s contributions. They did not only co-construct their descriptions of the pictures, but also their sentences: ‘mine has still the::’/‘the newspaper in the hand?’ (Example 9). Learners’ repetitions of each other’s words were particularly frequent throughout the interaction. Other-repetition helps to construct a shared perspective and is considered a signal of collaboration among learners (DiCamilla and Antón 1997). Fran and Celia used repetition to check for comprehension: ‘right eh eye is black’/‘right eye is . . . black?’ (Example 10), to confirm understanding of their messages: ‘the newspaper in the hand?’/‘the newspaper in the hand.’ (Example 9), and to acknowledge and accept new input: ‘a black eye? no?’/‘yeah black eye’ (Example 10). Self- and other-corrections were also frequent: ‘her right eh eye is black’/‘his right eye’ (Example 10), as well as explicit requests for assistance: ‘cómo se dice anterior?’ (Example 11). Example 11 Intermediate learner–learner interaction: 1 CELIA: the same things as the:: as . . . as the:: eh cómo se dice anterior? (‘how do you say the last one?’) ... 2 FRAN: as the last? 3 CELIA: as the last Fran and Celia explained that they had spent so much time on the task not because they had problems to understand each other, but because they wanted to identify all the differences and describe the pictures with the highest possible degree of accuracy. These two learners established linguistic accuracy as a subgoal of the task. They consciously and purposefully shaped their interaction into an opportunity to pool their linguistic resources and enhance each other’s use and knowledge of the language. Carmen and Scott’s non-collaborative orientation and their low level of engagement with each other’s contributions resulted in the NS ignoring the learner’s lexical holes. Fran and Celia, on the other hand, collaborated on all parts of the task. They co-constructed their descriptions, their sentences, and A. FERNÁNDEZ DOBAO 251 the lexical knowledge needed to solve each other’s lexical difficulties. In line with previous research in the field (Storch 2001, 2002a,b; Watanabe and Swain 2007; Watanabe 2008), the analysis of these two dyads suggests that a collaborative pattern of interaction facilitates collaborative dialogue, whereas a non-collaborative orientation reduces the opportunities for LREs and language learning. CONCLUSIONS The findings of the present study confirm that collaborative dialogue can occur in both learner–NS and learner–learner interaction during a meaning-oriented spot-the-difference task. When the goal of the task is to achieve successful communication of meaning, problems arising in interaction due to learners’ holes in L2 vocabulary knowledge are solved by means of CSs. Learners use alternate means of expressions, such as circumlocutions, L1 transfers or gestures to compensate for the lack of knowledge of the lexical items they desire to use in order to convey their messages. But these lexical problems can also act as a trigger for LREs. Learners and NSs may collaborate to fill these lexical holes, pooling their linguistic resources and engaging in language-mediated cognitive activities such as offering and assessing new input, or formulating and testing hypotheses about the meaning and function of the vocabulary they are using. As seen in previous research, this collaborative problem-solving activity serves to build new lexical knowledge; the lexical knowledge learners need to convey their messages with correct, previously unavailable, L2 vocabulary. The comparative analysis of advanced and intermediate learner–NS and learner–learner interaction suggests that the presence of an NS interlocutor affects both the amount and the nature of this collaborative dialogue. In the present study, lexical LREs were more frequent and more likely to be successfully resolved in learner–NS than in learner–learner interaction. The learner’s proficiency level also had an influence on the frequency of LREs, but not necessarily on their outcome. NSs provided more frequent assistance to the learners because they had the lexical knowledge needed for this purpose. Same level learners sometimes avoided collaborative dialogue and focused exclusively on the successful communication of the message because they lacked the linguistic resources needed to collaborate with their peers in the building of new lexical knowledge. The NSs’ higher level of expertise in the language explains also why learner–NS dyads were able to resolve successfully a higher percentage of their LREs than learner–learner dyads. In this study, most NSs tended to assist the learners by providing linguistic help to enhance their use and knowledge of the language, but others were more likely to stick to the communicative demands of the task and to focus on the successful communication of the message ignoring linguistic accuracy. Individual variation was also observed among learner–learner dyads. 252 LEARNER–LEARNER AND LEARNER–NS INTERACTION Following previous research that suggests that collaborative patterns of interaction tend to result in frequent linguistic assistance and co-construction of knowledge, whereas non-collaborative patterns correlate with low frequency of LREs (Storch 2001, 2002a,b; Watanabe and Swain 2007; Watanabe 2008), the interaction of two selected dyads was analysed more closely. Fran and Celia’s collaborative behaviour, their high level of involvement in the task, and their engagement with each other’s contributions resulted in a high frequency of LREs. On the other hand, Carmen and Scott’s less collaborative orientation towards the task, and in particular Scott’s limited involvement in the interaction, as illustrated in Examples 6, 7, and 8, were found to correlate with a remarkably low number of LREs. Storch (2004) explained learners’ different patterns of interaction within the framework of activity theory, as a result of their different motives and goals. As pointed out by Coughlan and Duff (1994), the ‘same task’ is not always ‘the same activity’. Learners approach tasks with different motives and beliefs that guide their behaviour, leading them to set their own goals and to maximize and select some actions over others. The analysis of Carmen and Scott’s, and Fran and Celia’s interactions, supported by their comments during the stimulated recall interview, illustrates their different goals and interpretations of the task. For Carmen and Scott the goal of the task was to identify the differences existing between their two sets of pictures, linguistic accuracy was not an issue and therefore the task could be successfully accomplished with little engagement from the NS. Fran and Celia, however, collaborated not only to identify the differences, but also to solve their lexical problems and enhance each other’s use and knowledge of the language. They went beyond the demands of the task, establishing linguistic accuracy as a joint goal that guided their actions and shaped their interaction into a knowledge-building activity. In sum, although small scale in nature, the present study reveals that the presence of a NS interlocutor, who has a level of expertise in the language that intermediate and advanced-level learners have not yet acquired, tends to facilitate the occurrence of LREs, even when no pedagogical intervention is made to promote this form of knowledge-building activity. However, the analysis of Carmen and Scott’s interaction also shows that the ‘expert’ in the L2, in this case the NS, cannot always be expected to provide the level of assistance that the ‘novice’, the learner, needs for new language knowledge to be built. Storch pointed out that ‘not all students work collaboratively when assigned to work on language tasks in pairs’ (Storch 2002a: 147). The present study suggests that this may also be true for learner–NS dyads. As illustrated by the comparative analysis of Carmen and Scott’s, and Fran and Celia’s interactions, the participants’ orientation to the task, their joint definition of the situation, their roles, and the goal of the activity may ultimately have a stronger effect on the nature of the interaction and the opportunities this offers for collaborative dialogue and language learning than their overall command of the language. A. FERNÁNDEZ DOBAO 253 Although the spot-the-difference task used for the purposes of this investigation was not designed with a pedagogical aim and the study was not conducted in a classroom context, the results obtained bear implications for teachers as well as researchers. The differences observed not only between learner–NS and learner–learner interaction, but also between individual dyads support the claim that, when asking learners to work in pairs, attention needs to be paid not only to task design and instructions, but also to the participants’ orientation to the activity and its impact on how the task is actually carried out. As explained by Storch, ‘simply assigning students to work in groups or pairs will not necessarily create conditions conducive to learning’ (Storch 2002a: 122). The present study was designed as a small-scale study examining interaction between a limited number of learners and NSs. Together with the reduced number of participants and amount of data collected it has two other main limitations. The conclusions are based on the analysis of different learners interacting with different interlocutors. Future research will need to investigate how the same learner interacts with both native and non-native interlocutors. Learning has not been traced beyond the limits of the LREs. The nature of the study did not allow for a pre- and post-test design, since it would have been impossible to establish whether learning had occurred during the performance of the task or during the stimulated recall interview with the researcher. Although previous research has provided plenty of evidence of learning occurring in LREs (e.g. Swain and Lapkin 1998, 2002; Williams 2001; Lapkin et al. 2002; Tocalli-Beller and Swain 2005; Watanabe and Swain 2007; Kim 2008; Zeng and Takatsuka 2009), Sato and Lyster (2007) found that learners respond differently to feedback provided by other learners than by NSs. Further research is therefore needed in order to establish the influence that NS interlocutors may have not only on the amount and nature of lexical LREs, but also on the internalization of the language knowledge socially constructed in these LREs. SUPPLEMENTARY DATA Supplementary material is available at Applied Linguistics online. NOTES 1 The English language learners were bilingual NSs of Spanish and Galician. All the English NSs had, at least, an intermediate proficiency level of Spanish, but a very limited knowledge of Galician. 2 The test used to assess the learners’ proficiency level was the Oxford Placement Test (Allan, D. 1992. Oxford Placement Test. Oxford: Oxford University Press). This test is based on a scale of 0–190. A score of 135–149 is considered mid-intermediate, 150–159 upperintermediate, and 160–180 advanced. The scores of the intermediate-level learners who participated in the study 254 LEARNER–LEARNER AND LEARNER–NS INTERACTION ranged from 135 to 148 and those of the advanced-level learners from 161 to 176. 3 Since the stimulated recall interview was always conducted in the participants’ L1, all learners’ retrospective comments have been translated to English. When the learners used English to repeat the actual words used in the performance of the task, I resort to quotation marks. 4 Each of the examples here presented was counted as one CS episode. Examples 3–5 were also counted as LREs. They are simultaneously CS episodes and lexical LREs. 5 The number of lexical LREs may seem relatively low, but these results are in line with those of previous research examining LREs in meaning-oriented tasks. This research has shown that learners’ generated attention to language can occur during meaningfocused tasks, although it is ‘neither frequent nor extensive’ (Williams 2001: 328). 6 Equality refers to the degree of control or authority over the task and mutuality to the participants’ level of engagement with each other’s contributions (Storch 2002a: 127). 7 The number of differences between the two versions of the story that each dyad was able to identify ranged from 21 to 30, with an average of 26. Fran and Celia identified 30 differences, whereas Carmen and Scott identified only 24. REFERENCES Coughlan, P. and P. A. Duff. 1994. ‘Same task, different activities: Analysis of SLA task from an activity theory perspective in J. P. Lantolfand G. Appel (eds): Vygotskian Approaches to Second Language Research. Norwood, NJ: Ablex, pp. 173–93. DiCamilla, F. J. and M. Antón. 1997. ‘Repetition in the collaborative discourse of L2 learners: A Vygotskian perspective,’ The Canadian Modern Language Review 53/4: 609–33. Donato, R. 1988. ‘Beyond group: A psycholinguistic rationale for collectivity in second language learning,’ Unpublished Doctoral Dissertation, University of Delaware. Donato, R. 1994. ‘Collective scaffolding in second language learning in J. P. Lantolf and G. Appel (eds): Vygotskian Approaches to Second Language Research. Norwood, NJ: Ablex, pp. 33–56. Donato, R. 2000. ‘Sociocultural contributions to understanding the foreign and second language classroom in J. P. Lantolf (ed.): Sociocultural Theory and Second Language Learning. Oxford: Oxford University Press, pp. 27–50. Dörnyei, Z. and J. Kormos. 1998. ‘Problem-solving mechanisms in L2 communication: A psycholinguistic perspective,’ Studies in Second Language Acquisition 20/ 3: 349–85. Færch, C. and G. Kasper. 1983. ‘Plans and strategies in foreign language communication’ in C. Færch and G. Kasper (eds): Strategies in Interlanguage Communication. New York, NY: Longman, pp. 20–60. Foster, P. and A. Ohta. 2005. ‘Negotiation for meaning and peer assistance in second language classrooms,’ Applied Linguistics 26/3: 402–30. Gass, S. M. and A. Mackey. 2000. Stimulated Recall Methodology in Second Language Research. Mahwah, NJ: Lawrence Erlbaum Associates. Gass, S. M., A. Mackey, and L. Ross-Feldman. 2005. ‘Task-based interactions in classroom and laboratory settings,’ Language Learning 55/4: 575–611. Kim, Y 2008. ‘The contribution of collaborative and individual tasks to the acquisition of L2 vocabulary,’ The Modern Language Journal 92/ 1: 114–30. Kim, Y. and K. McDonough. 2008. ‘The effect of interlocutor proficiency on the collaborative dialogue between Korean as a second language A. FERNÁNDEZ DOBAO learners,’ Language Teaching Research 12/2: 211–34. Lantolf J. P. (ed.). 2000. Sociocultural Theory and Second Language Learning. Oxford: Oxford University Press. Lantolf J. P. and G. Appel (eds). 1994. Vygotskian Approaches to Second Language Research. Norwood, NJ: Ablex. Lantolf, J. P. and S. L. Thorne. 2006. Sociocultural Theory and the Genesis of Second Language Development. Oxford: Oxford University Press. Lapkin, S., M. Swain, and M. Smith. 2002. ‘Reformulation and the learning of French pronominal verbs in a Canadian French immersion context,’ The Modern Language Journal 86/4: 485–507. Leeser, M. J 2004. ‘Learner proficiency and focus on form during collaborative dialogue,’ Language Teaching Research 8/1: 55–81. Leont’ev, A. N. 1981. ‘The problem of activity in psychology’ in J. V. Wertsch (ed.): The Concept of Activity in Soviet Psychology. Armonk, NY: M. F. Sharpe, pp. 37–71. Mackey, A., R. Oliver, and J. Leeman. 2003. ‘Interactional input and the incorporation of feedback: An exploration of NS-NNS and NNS-NNS adult and child dyads,’ Language Learning 53/1: 35–66. Ohta, A. S. 2000. ‘Rethinking interaction in SLA: Developmentally appropriate assistance in the zone of proximal development and the acquisition of L2 grammar’ in J. P. Lantolf, , pp. 51–78 (ed.): Sociocultural Theory and Second Language Learning. Oxford: Oxford University Press, pp. 51–78. Ohta, A. S. 2001. Second Language Acquisition Processes in the Classroom: Learning Japanese. Mahwah, NJ: Lawrence Erlbaum. Pica, T., F. Lincoln-Porter, D. Paninos, and J. Linnell. 1996. ‘Language learners’ interaction: how does it address the input, output, and feedback needs of L2 learners?,’ TESOL Quarterly 30/1: 59–84. Sato, M. and R. Lyster. 2007. ‘Modified output of Japanese EFL learners: Variable effects of interlocutor versus feedback types’ in A. Mackey (ed.): Conversational Interaction in Second Language Acquisition: A Series of Empirical Studies. Oxford: Oxford University Press, pp. 123–42. Shehadeh, A. 1999. ‘Non-native speakers’ production of modified comprehensible output 255 and second language learning,’ Language Learning 49/4: 627–75. Stetsenko, A. and I. Arievitch. 1997. ‘Constructing and deconstructing the self: Comparing post-Vygotskian and discoursebased versions of social constructivism,’ Mind, Culture, and Activity 4/3: 159–72. Storch, N. 1998. ‘A classroom-base study: Insights from a collaborative text reconstruction task,’ ELT Journal 52/4: 291–300. Storch, N. 2001. ‘How collaborative is pair work? ESL tertiary students composing in pairs,’ Language Teaching Research 5/1: 29–53. Storch, N. 2002a. ‘Patterns of interaction in ESL pair work,’ Language Learning 52/1: 119–58. Storch, N. 2002b. ‘Relationships formed in dyadic interaction and opportunity for learning,’ International Journal of Educational Research 37/3–4: 305–22. Storch, N. 2004. ‘Using activity theory to explain differences in patterns of dyadic interactions in an ESL class,’ The Canadian Modern Language Review 60/4: 457–80. Swain, M. 2000. ‘The output hypothesis and beyond’ in J. P. Lantolf (ed.): Sociocultural Theory and Second Language Learning. Oxford: Oxford University Press, pp. 97–114. Swain, M. 2001. ‘Integrating language and content teaching through collaborative tasks,’ The Canadian Modern Language Review 58/1: 44–63. Swain, M. and S. Lapkin. 1998. ‘Interaction and second language learning: Two adolescent French immersion students working together,’ The Modern Language Journal 82/3: 320–37. Swain, M. and S. Lapkin. 2001. ‘Focus on form through collaborative dialogue: Exploring task effects’ in M. Bygate, P. Skehan, and M. Swain (eds): Researching Pedagogic Tasks: Second Language Learning, Teaching and Testing. London: Longman, pp. 99–118. Swain, M. and S. Lapkin. 2002. ‘Talking it through: Two French immersion learners’ response to reformulation,’ International Journal of Educational Research 37/3–4: 285–304. Tarone, E. 1981. ‘Some thoughts on the notion of communication strategy,’ TESOL Quarterly 15/3: 285–95. Tocalli-Beller, A. and M. Swain. 2005. ‘Reformulation: The cognitive conflict and L2 learning it generates,’ International Journal of Applied Linguistics 15/1: 5–28. Varonis, E. and S. M. Gass. 1985. ‘Non-native/ non-native conversations: A model for