1In face-to-face interactions, speakers and interlocutors engage in highly cooperative ways as they constantly seek and provide feedback responses which enable them to signal their understanding and agreement, both verbally and visually, hence ensuring that the interaction goes smoothly. Feedback responses, also known as ‘backchannels’ (Yngve, 1970) can be defined as “verbal and non-verbal listener feedback in spoken interaction that does not involve a speaker shift, but functions as a turn-continuer” (Bjørge, 2010: 193). While the term ‘feedback’ is broadly understood in terms of engagement, participation, or listening cues on the interlocutor’s part, whose aim is to support mutual understanding and cooperation, other types of feedback, which are more specific to communicative settings with a certain focus on form (language learning contexts especially), serve slightly different functions. Corrective feedback (CF), for instance, i.e. feedback typically provided by language instructors to correct L2 (i.e. second, foreign language or additional language) learners’ output (cf. negative evidence; Gass, 2003) has been the topic of numerous studies in the field of L2 acquisition and instruction (Nassaji & Kartchava, 2021). Conversational feedback (e.g. backchanneling, head nods, continuers), on the other hand, has mostly been the focus of first language (henceforth L1) studies (e.g. Allwood et al., 2007; Boudin et al., 2021). In addition, feedback is also essential when needing to re-establish common ground as part of miscommunication management (Dascal, 1999), whether in L1 or L2. The concept of grounding, put forward by Clark (1996), refers to the process of establishing mutual understanding, which requires active involvement on the speakers and interlocutors’ part. Therefore, the notions of common ground and feedback are intricately related.
2The aim of the present research is to bring together these different types of feedback in the context of tandem conversations (Calvert & Brammerts, 2003) where a high degree of cooperation is expected to address issues of (mis/non)understanding, to facilitate the overall communication flow, but also to promote the L2 partner’s language development. Building on previous work (Scheuer & Horgues, 2021; Kosmala, Horgues & Scheuer, 2023), this study explores the different types of feedback (corrective, conversational, miscommunication management, or a combination of the three) produced by L1 interlocutors in the course of interactions held in English, both in tandem (L1-L2) and control (L1-L1) settings. We adopt a multimodal approach to interaction (Goodwin, 2010; Mondada, 2019; Perniss, 2018), whereby interactants do not only make use of their voice to convey feedback (e.g. “mm mm”, “ok”) but their head, face, and hands, through facial expressions, head movements, and manual gestures. Both vocal-verbal and visual-gestural feedback cues will hence be taken into account for analysis. Our study is based on a selection of the SITAF Corpus (Horgues & Scheuer, 2015) with a focus on feedback provision by four L1 speakers in a task where their main role is to be the listeners of their conversation partner’s story. We carry out a comparative study of their feedback patterns in two interactional conditions for the same narrative task: when listening to the story told by i) a fellow L1 speaker or ii) their L2 tandem partner. In particular, we explore the different ways in which the L1 interlocutor provides a variety of feedback cues to maintain, establish or re-establish common ground, depending on who they interact with.
3Our paper is structured as follows: we begin with a review of the literature across L1 and L2 studies in tandem learning and common ground, conversational feedback, miscommunication, and corrective feedback (section 1), before presenting our data and methodology (section 2). Then, we describe our results (section 3), discuss our findings (section 4) and offer conclusions and perspectives.
4The communicative set-up of Tandem language learning lends itself particularly well to the study of common ground work. Indeed, Tandem learning originated in the post second world war era, when educators were striving to bring together language communities who had been antagonized by dramatic geopolitical events (see the creation of the Franco-German Youth Office promoted intercultural exchanges as of 1963). In its traditional form (Helmling, 2002; Calvert & Brammerts, 2003) tandem learning consisted of matching up tandem pairs of native speakers of different languages who met up on a regular basis in view of learning about each other’s language and culture. The founding principles for such tandem conversation exchanges were: autonomy, reciprocity, and intercultural learning.
5Technological advances then prompted some evolution in the tandem modalities, with the Internet allowing for the possibility for interactants to engage in distant, asynchronous written electronic interaction (in e-tandem, O’Rourke, 2005, 2007). About a decade later, videoconferencing tools expanded spoken communication possibilities to computer-mediated virtual synchronous exchanges (often called ‘Teletandem’, see Vassallo & Telles, 2006; Cappellini & Zhang, 2013; Cappellini & Macré, 2020. Helping to overcome the differences in language proficiency levels — the L1 speaker being deemed the “expert” and the L2 learner being considered the “novice” — and cultural identities existing between two tandem partners, tandem learning has been described as a difference-levelling kind of practice. Indeed, momentary asymmetries (linguistic and/or cultural gaps) would be lessened or levelled out, at the moment when partners switch languages (and their expert/novice roles therefore get reversed). Vassallo & Telles (2006) have referred to this as a process of “symmetrisation” through role reversal. In that sense, the reciprocity principle, at the heart of the tandem learning method, plays a key role in (re)establishing common ground between tandem partners through evening out linguistic, cultural, and interactional asymmetries between them.
6In the last 10 years, there has also been a shift in the linguistic paradigm of traditional tandem practices (originally based on the bi-lingual, bi-cultural model) to embrace speakers’ increasing multilingualism. This has been reflected in online intercultural projects (O’Dowd, 2018) recently labelled VE (Virtual Exchanges, O’Dowd, 2018) involving speakers of various language proficiency levels (more or less proficient users). In such telecollaborative learning set-ups, nativeness becomes irrelevant, since conversation partners interact in a lingua franca to ensure communicative common ground (target language development no longer being the prime learning objective) so as to develop other competences through task-based projects.
7Whether tandem learning is practiced in its traditional form or in its more recent adaptations, the fact that interactants engage in it intentionally and with explicit (collaborative) learning objectives — be they mainly linguistic or focused on other skills — make this communicative setting different from ‘simple’, informal, naturally-occurring conversations between L1 and L2 interactants. This difference is important to underline before embarking on the study of interlocutor feedback. Indeed, when interacting in a tandem set-up, L1 and L2 speakers agree on a co-learning ‘contract’ and are likely to resort to — and expect — feedback from their interlocutor more so than they would do if they were having unprompted spontaneous conversations with one another in their everyday life.
8Let us now review previous findings about some key functions of interlocutor feedback for our present study. The term ‘feedback’, depending on the theoretical framework adopted or area of study, has different meanings and implications. Based on a review of the literature in L1 and L2 studies, we distinguish between three types of feedback: (1) Conversational feedback, (2) Miscommunication feedback, and (3) Corrective feedback.
- 1 It should be noted that the terms “continuers” or “backchannels” are slightly more restrictive than (...)
- 2 Read Rasenberg et al., (2020) for a deeper understanding of ‘grounding’ versus ‘priming’ where alig (...)
9Conversational feedback (Prévot & Bertrand, 2012), also known as backchanneling (Ynge, 1970), listener responses (Bavelas et al., 2002), or continuers (Schegloff, 1982)1, includes audio-visual signals that interlocutors offer the main speakers in the course of the interaction to indicate attention, understanding, or agreement. Originally, the notion of backchanneling put forward by Yngve (1970) implied two channels of communication that operated simultaneously in the course of a conversation. While the predominant channel is mainly concerned with speakers’ turns, the secondary channel involves listeners who continuously provide listening cues. However, listening is far from being a passive or secondary activity, as the speaker-listener status is constantly being redefined moment by moment in the interaction, and thus involves dynamic turn shifts. Indeed, feedback cues do not merely signal that the interlocutors are listening, but can actively shape the main speakers’ storytelling: in this view, listeners can be considered as “co-narrators” (Bavelas et al., 2000). These cues, such as ‘uh-uh’, brief comments (“I see”), or other visible kinds of behaviors (smiles, frowns, head nods etc.) hence play a crucial role in ensuring a smooth, two-way communication. During a storytelling activity for instance, visual feedback cues, such as nodding, serve as relevant signals that inform the main speaker that the interlocutor is not only paying attention to the story, but is also affiliating with the teller’s stance (Stivers, 2008). In addition, the notion of feedback can also be related to the concept of alignment and more specifically common ground2: according to Pickering & Garrod (2004), listeners constantly align their linguistic representations and mental states through feedback markers in order to facilitate adjustment and ensure mutual understanding.
10Recent corpus-based studies have offered multimodal models of feedback, combining different levels of analysis, such as prosody, syntax, and gesture (Boudin et al., 2021; Ferré & Renaudier, 2017; Prévot & Bertrand, 2012). For instance, Ferré & Renaudier (2017) distinguished between unimodal feedback (either audible or visible) and bimodal feedback (combined modalities). Similarly, Boudin (2022) offered a multi-layered annotation system including a wide range of phenomena, along the lines of Bertrand & Goujon (2017) who annotated eyebrow movements, discourse markers, and change-of-state tokens. Similarly, our study integrates multiple levels, and does not prioritize one feedback modality over another.
11In the previous section, we emphasized the role of feedback in conversation to ensure continuity and mutual understanding between the interactants. However, feedback is not only used to provide evidence of understanding and non-problematic communication, as it can sometimes specifically serve a problem-signaling function. In this case, it plays an essential role to re-establish common ground destabilized by miscommunication (Dascal, 1999). Indeed, Clark & Schaefer reported that feedback may show ‘evidence of trouble in understanding’ (1989: 268, our emphasis) from which speakers will then engage in some trouble-shooting or management of this troubled communication, also referred to as repair work (Schegloff, Sacks & Jefferson, 1977) or negotiation of meaning (Varonis & Gass, 1985) in conversation analysis and Second Language Acquisition theories.
12We use ‘miscommunication’ as an umbrella term for any situation of problematic comprehension during an interaction comprising mis-understanding, non-understanding, or encountering difficulty to understand someone’s message. In this sense, we follow Mauranen’s definition of miscommunication as “a potential breakdown point in conversation or at least a kind of communicative turbulence” (Mauranen, 2006: 128). In such situations, the interlocutor’s feedback plays an essential role to ensure that miscommunication is noticed, taken into account, and potentially resolved. The absence of any such feedback could result in communication failure.
13In particular, interlocutor feedback is essential to ensure two essential steps in the miscommunication management sequence: the signal and reaction phases (see figure 1).
Fig. 1: The 4 phases of miscommunication management (adapted from Varonis & Gass, 1985; Mauranen, 2006)
- 3 CB instances are defined in this research as one listener’s demonstrable difficulty / incapability (...)
14While miscommunication is present in everyday L1-L1 conversations (read Robles, 2017 for example), one interactional context which has been described as increasing the potentiality for miscommunication is that of L1-L2 interactions. In particular, Varonis & Gass (1985) reported that L1-L2 interactions yielded more miscommunication opportunities than interactions between fellow L1 speakers. These authors attributed this to the limited common linguistic and socio-cultural ground (“belief space” in their terms) between L1 and L2 speakers whom they described as being “multiply handicapped in interactions with one another” (ibid: 340) especially due to “radically different customs, modes of interacting, notions of appropriateness and, of course, linguistic systems” (ibid: 337). This observation now seems a little excessive in the light of how, in the twenty-first century, global communication, global mobility and wide-spread multilingualism have allowed to bridge differences and reinforce common ground between L1 and L2 speakers on many levels (linguistic, socio-cultural and pragmatic). However, we can assume that L1-L2 conversations may still evidence regular instances of miscommunication where the interlocutor’s feedback plays a pivotal role. In particular, we have shown in previous research (Horgues & Scheuer, 2018; Scheuer & Horgues, 2021; Kosmala, Horgues & Scheuer, 2023) that i) Communication Breakdown (CB) instances3 were relatively rare (fewer than 100 over 5 hours of speech analysed for 21 tandem pairs), ii) miscommunication was a two-way street emerging both from the L2 speaker’s output (about two thirds of the time) but also from the L1 speaker’s output (about one third of the time) with CB detection, management and resolution being co-constructed and successful and highly multimodal.
15Another possible function of feedback pertains to educational contexts, not only in child-caretakers’ repair work (Chouinard & Clark, 2003; Morgenstern et al., 2013), but also in formal language instruction contexts. In the field of Second Language Acquisition (SLA) research in particular, the focus has often been given to the feedback language teachers provide in response to L2 learners’ errors, in view of rectifying these erroneous productions. What has been termed Corrective Feedback (henceforth CF) therefore serves a didactic function, as it is the remedial intervention traditionally performed by an English as a Second Language/English as a Foreign Language (ESL/EFL) teacher providing “negative evidence” (Gass, 2003), i.e. evidence of which language forms are not target-like. Li (2014) summarizes that teacher CF consists in “responses to learners’ production errors, the purpose of which is, or is perceived as, remedial, regardless of whether the errors cause communication problems” (Li, 2014: 374).
16Here again (Fig. 2), (teacher) feedback is central in the CF sequence.
Fig. 2: The 3 phases of the corrective feedback sequence (adapted from (Lyster & Ranta, 1997).
17At the end of the CF sequence may also follow another feedback move by the teacher to validate and/or acknowledge the learner’s uptake (although this phase is more rarely included in the SLA literature). The teacher may resort to different types of CF which are more or less implicit: recasts, clarification requests, repetition of learner error, elicitation, explicit correction. Recasts, which consists in “a teacher’s reformulation of all or part of a student’s utterance, minus the error” (Lyster & Ranta, 1997: 46), have been reported to be language teachers’ preferred CF strategy although its effectiveness for L2 learners’ linguistic development is often disputed due to its limited saliency (Brown, 2016; Mackey et al., 2000; Saito, 2021; Saito & Lyster, 2012).
18In recent years, the focus has shifted from teacher to peer feedback especially in language learning contexts (Iwashita & Dao, 2021; Sato, 2017), but corrective feedback in tandem L1-L2 exchanges is still understudied. Our previous analyses (Debras et al., 2015; Scheuer & Horgues, 2020, 2021) of L1 English speakers having face-to-face tandem conversations in English with French L2 learners (see SITAF corpus described below in 3.1.) have shown that i) the Native-English tandem partners tended to provide CF focusing preferably on vocabulary (52.5%) more than pronunciation (20%) or morphosyntax (12.5%) with some combined foci CF (15%), ii) their feedback was more often spontaneously offered (57%) than elicited by the L2 learner iii) recast was the largely predominant strategy used by NS tandem partners (84%), iv) their CF did not necessarily (only 52.6%) yield to modified output uptake moves by their partners , and v) and that NS feedback was highly multimodal.
19In sum, few studies have explored conversational feedback in L1-L2 interactions, with a few exceptions, see Uematsu (2001) on backchannels between L1 and L2 speakers in English and Japanese, Lee (2020) in ELF communications, or Sbranna et al. (2022) on L1 Italian, L1 German and L2 German. Conversely, a majority of studies in SLA have focused on CF, especially in language learning contexts. Having reviewed some characteristics of tandem L1-L2 interactions, and some key theoretical functions of interlocutor feedback in relation to common ground, we would now like to explore the following research questions:
(1) Does ‘main speaker’ condition (L2 vs L1 main speaker) have an effect on L1 interlocutor feedback?
(2) How does L1 interlocutor feedback serve to (re-)establish and maintain common ground in tandem compared to control interactions?
20More precisely, since tandem L1-L2 interactions are thought to induce more linguistic and cultural gaps to fill, i.e. a bigger asymmetry between tandem partners, we would like to test the hypothesis that common ground requires more work and attention, especially through a bigger role of interlocutor feedback, in tandem exchanges compared to L1-L1 interactions.
21To test our hypotheses, we resort to a video-recorded, multimodal corpus of spoken face-to-face tandem interactions between French students learning English and English students learning French (the SITAF corpus, Horgues & Scheuer, 2015). The communicative setting is that of the traditional form of tandem learning (see 1.1.) since each participant is a native speaker of their partner’s target language and that the partnership has a clear learning objective revolving around L2-linguistic development and intercultural learning.
22The SITAF corpus consists of 21 such English/French tandem pairs who were video-recorded at the start and again at the end of their autonomous (optional) tandem practice through the course of one university semester (semester 2, 2012-2013). The 21 French participants (coded F01 to F21) were undergraduate students preparing for a degree in English studies at a French university (L2 English proficiency level ranging from upper-intermediate to advanced), and their 21 Native-English partners (coded A01 to A21) were incoming mobility undergraduate students at that same university, coming from various countries (England, Ireland, USA, Canada, Australia; L2 French proficiency level ranging from lower intermediate to advanced). The tandem pairs performed two main interactive semi-spontaneous game-like tasks on these two occasions: a narrative task called “Liar Liar” and a debating task called “Like Minds”. These two games were initially performed in the participants’ first language with a fellow native speaker (the L1-L1 control condition) for participants to get familiarized with the task. They then played these same games (in the two languages) in the L1-L2 tandem condition.
23Since the present study focuses on a thorough analysis of interlocutor feedback, especially feedback provided by the L1 interlocutor, we selected Game 1 (‘Liar, Liar’) for analysis as it places the L2 learner in the main “speaker role” and the L1 participant in the main “listener role/posture” thus counterbalancing the linguistic proficiency asymmetry between the interactants. This task gives ample opportunity to analyze the L1 interlocutor as a main “feedback giver”, whereas in the debating game both interactants share both speaking and listening roles more equally. Game 1 is also mostly a narrative/storytelling task, which serves as a good basis for comparison with earlier studies conducted on narrations (e.g. Bavelas et al., 2000; Stivers, 2008). In the Liar-Liar game the L2 learner narrated the story of their latest holiday making sure to include three discrete lies, which their L1 interlocutor had to identify by asking questions at the end of the L2 learner’s story. We selected 5 (out of the 21) SITAF corpus tandem pairs which reflected a variety of feedback giving styles: pairs A05-F05, A06-F06, A08-F08, A09-F09, A11-F11.
- 4 We excluded the Q/A section occurring after the L2 learner’s story from our analysis, since this po (...)
24We focused on observing the L1 interlocutor’s feedback signals (verbal, vocal and/or visual) in response to the main speaker’s “story4”. To reflect the quantity of feedback provided by one listener/interlocutor in relation to the quantity of speech produced by the main speaker, we calculated the number of words contained in each L2 learner’s main “story” (mean number of words, ranging from 205 words to 776). This allows us to calculate the rate or frequency of feedback tokens per interlocutor, and compare it between interlocutors, especially in the two conditions. Indeed, we systematically analyzed the feedback signals provided by these same L1 English-speaking interlocutors (i.e. A05, A06, A08, A09, A11) as they were listening to different main speakers depending on the interactional condition. We therefore ran a comparative study of interlocutor feedback patterns depending on who the main speaker was: either an L2 speaker (their tandem partner) in the tandem condition, or a fellow L1 speaker in the control condition. Both authors watched the video samples and coded feedback signals independently, and then discussed problematic tokens to make final decisions.
25A feedback token/move was defined as any reaction (verbal, vocal, and/or visual) produced by one of the five target L1 interlocutors as an immediate response/reaction to a speech turn produced by the main speaker they were listening to. When one main speaker’s speech turn triggered multiple sequential responses in the interlocutor, these were coded as separate feedback tokens. However, if several feedback markers appeared within the same response (e.g. “oh + you went where?”) it was treated as a single token, with a mixed function.
26Each feedback token was coded for function and for modality:
– Function: we classified the feedback functions around 4 categories: conversational (i.e. response in non-problematic communication), miscommunication (manifestation of non-understanding / misunderstanding / processing difficulty), corrective (negative evidence expressed by the L1 speaker following L2 learner’s erroneous form production), mixed (one of complex feedback token combined several functions, namely Conv+Miscomm, or, Conv+Corrective, or Miscomm+Corrective, Conv+Miscomm+Corrective)
– Modality: we classified the feedback modalities around 3 categories: verbal-vocal only, visual only (hand, face, and/or body signals), multimodal (a combination of the verbal-vocal-visual signals). For multimodal cases involving different signals (e.g. a verbal signal accompanied by a gestural one), it was still counted as a single feedback token.
27The feedback sequences were annotated independently by the two authors on Excel. All the difficult and ambiguous cases were discussed one by one until the authors reached an agreement.
28A total number of 333 feedback tokens were found in the data: 106 in the control condition versus 227 in the tandem condition, amounting to 6.9 per hundred words (phw) in control and 10.1 phw in tandem. A log-likelihood ratio test confirmed that this difference in frequency between the two conditions was statistically significant, G²(1) = 44.99, p < .001. In addition, Figure 3 shows individual rates which further illustrate the differences between the two interacting conditions (main story-teller: L1 speaker or L2 speaker) across all listeners. However, it should be noted that the sample under study is very small (N =5), which limits how far these findings can be generalized.
Fig 3: Rate of feedback tokens (phw) provided by each L1 interlocutor in the two conditions
29When it comes to feedback types, we can also observe differences between these two conditions, as illustrated in Figure 4.
Fig 4: Proportion of feedback types across the two conditions
30A majority of the feedback tokens (90 %) are conversational in the control condition, while the distribution is more varied in the tandem condition with a higher proportion of corrective (18 %) and miscommunication (9 %) feedback, as well as mixed types (13 %). A chi-square test of independence found a significant association between condition and feedback type χ²(1, N = 333) = 27.93, p < 0.001. A Fisher’s exact test confirmed this result (OR = 5.67, p < 0.001) indicating a substantially higher proportion of corrective, miscommunication, and mixed feedback in the tandem condition.
31Figure 5 illustrates differences in modality across the two conditions, with verbal/vocal (only) visual (only), and multimodal (a combination of the three).
Fig. 5: Proportion of modality types across the two conditions
32A Chi-square of independence indicated a significant association between condition and modality χ²(2)= 14.76, p < .001, with a higher proportion of multimodal types in the tandem condition (69% vs 48%) as opposed to slightly more visual-only types (36% vs 26%) in the control one.
33Additionally, besides being multimodal, feedback tokens can be highly contiguous and/or intertwined in tandem interactions and thus form complex feedback sequences, as illustrated in the example below [1], taken from Pair 9 (A09 the L1 interlocutor and F09 the L2 speaker). This excerpt takes place at the very beginning of the task, where the L2 speaker starts her narrative.
34Different types of feedback are provided by A09 as part of a complex feedback sequence, the L1 interlocutor, as the story goes along. First, A09 is proactive and provides explicit corrective feedback to encourage his partner to seek help if needed (line 2), and then displays continuous conversational feedback with a series of nods and the verbal token ‘yeah’ (lines 3-5) to mark his understanding. As F09 introduces the word ‘Latvia’ (pronounced incorrectly), A09 experiences trouble in understanding the word, and therefore asks her to repeat it (line 7) while moving his trunk forward and squinting his eyes (miscommunication feedback). After F09 repeats the target word more clearly (line 8), A09 gazes up (line 9) and the misunderstanding is resolved. A09 finally provides a mixed feedback token which includes a combination of conversational feedback (with the discourse markers ‘oh’ and ‘ok’) and corrective feedback (with the recast ‘Latvia’).
35Figure 6 shows individual differences for feedback types, exhibiting interlocutor-specific tendencies.
Fig 6: Feedback types per interlocutor
36These findings echo the ones depicted in Fig 3 where a great deal of interlocutor variability was found among the participants. In particular, the difference is rather striking for interlocutors A11 and A05 who exhibit two opposite patterns of behavior: while one seems maximally engaged, resorting to a variety of feedback types, the other is minimally engaged relying predominantly on conversational feedback only, regardless of the condition. These differences are illustrated in the following examples.
37In this example [2], the L2 speaker (F11) is in the middle of her narrative and is retelling an incident which took place in the winter.
38The L1 interlocutor (A11) first displays his active involvement and affiliation towards his partner’s narrative with raised eyebrows and a friendly and engaged tone (conversational feedback, lines 1-4) to express empathy towards the story. He then shifts to a different type of feedback, explicit corrective feedback, in reaction to the L2 speaker’s incorrect use of the noun phrase ‘leather interior’. He thus takes some time to explain the switch between the words with a series of switching hand gestures (repeated across lines 6-10). While the L2 speaker does not interpret the feedback received as corrective right away (as she resumes her turn, line 7), the L1 interlocutor provides conversational feedback (“yeah yeah”, line 8), and makes his corrective feedback more explicit (“just switch the words”, line 9, “it’s leather interior not interior leather”, line 10). The L2 learner then produces a similar switching hand gesture while repeating the target sequence “leather interior”, followed by the L1 interlocutor’s uptake validation (“there you go”) as a form of praise and support. In sum, A11 can be considered a maximally engaged feedback provider, as he was able to continuously display attention and empathy, while being an effective tandem partner through explicit form-focused feedback and assistance. F11’s repetition of the switching hand gesture further reinforced their mutual understanding and thus validates the reaching of some common ground.
39The next example [3] is in stark contrast with this one, as the L1 interlocutor displays minimal signs of engagement.
[3] Minimally engaged feedback provider (Pair 5)
1 F05: um and we went there in March for our birthday because we are all born in the beginning of March
2 F05: so we went there for four days from Sunday to Thu- Thursday I think Wednesday Thursday yeah
3 F05: it was very great very very sunny you know
4 F05: it was very cold in France but in was cold in Prague - Prague but it was very sunny you know with no clouds nothing just the sun and blue sky so it was very great
(slight smiling and nodding from A05)
40In this short excerpt, A05, the L1 interlocutor, barely moves her body in the course of F05’s narrative and does not produce any verbal-vocal feedback token (despite the main speakers’ discursive appeals through ‘you know’). Her gaze is maintained towards her interactant throughout the exchange, and she does smile and nod at some point (l. 4) but it is barely perceptible. We may wonder whether A05 adopted a ‘let it pass strategy’ or if she is simply a minimally engaged listener. Her similar behavior in the L1-L1 condition seems to point to an individual characteristic in her case.
- 5 Feedback forms and values were not reported in this study as they are not the scope of this paper.
41The first aim of this study was to investigate whether the ‘main speaker’ condition (L2 tandem partner vs L1 control speaker) would have an effect on the interlocutor’s feedback. We hypothesized that the tandem condition would lead to a higher rate of feedback tokens than the control condition, which was confirmed in our findings. Another effect on feedback was not only frequency, but complexity and modality, as we found a higher proportion of multifunctional and multimodal types in the tandem condition than in the control condition, but a slightly higher proportion of visual types in the control condition. This may suggest more limited presence of the interlocutor during L1-L1 interactions through short verbal responses (such as ‘ok’ or ‘yeah yeah’) along head and facial signals (e.g. smiles and nods)5, as opposed to the L1-L2 interactions where feedback strategies are more complex (discussed below). However, it is also important to consider individual differences, as different profiles of feedback providers emerged in the data: while some could be described maximally engaged, others are rather minimally engaged, both visually and verbally. For one speaker in particular (A05) very few differences in her feedback behavior were observed across the two conditions, so our findings need to be examined with caution. A larger sample of the data, along with statistical tests, would provide more generalizable results.
42A second related question was whether interlocutor feedback aimed at (re-)establishing or maintaining common ground across the two conditions. Indeed, our results from the data point towards several interpretations.
43First, the higher rate of feedback tokens found in the tandem condition than in the control one indicates more active presence from the interlocutor, whose role is not only to display attention and understanding (conversational feedback) but also to assist the L2 tandem partner facing language difficulties (corrective feedback). Some of these difficulties may lead to miscommunication (e.g. example [1]), where common ground is temporarily threatened, but where feedback strategies are deployed to reach mutual understanding. In the case of example [1], common ground was initially secured and maintained (through the repetitive nodding) but then had to be restored, following the miscommunication.
44Secondly, the fact that a majority of the feedback tokens in the control condition were conversational further suggests that the L1-L1 interactions were mostly non-problematic, with only a limited number of feedback related to miscommunication. Conversely, in the tandem condition, we found a higher proportion of multifunctional and complex feedback which illustrates the higher degree of co-construction found in tandem interactions.
45Establishing common ground is hence shared responsibility on the L1 and L2 speakers’ parts: given the co-learning contract and language proficiency gap at the heart of the tandem learning setting, the L2 learners are proactive in eliciting, anticipating, and receiving L1 interlocutors’ feedback. Being ready to address and secure common ground in tandem interactions is, in fact, a prerequisite for a friendly and safe environment for linguistic and intercultural learning. Combining conversational with corrective feedback (example [2]) or with miscommunication feedback (example [1]) into complex feedback sequences hence enables the L1 interlocutors to provide useful form-focused feedback without jeopardizing mutual friendliness nor violating socio-cultural conventions. In the L1-L1 condition, such strategies are not necessarily needed, since the participants’ interactional roles and language proficiency are not so asymmetric and interactants are not bound by a co-learning contract.
46This study has shown the importance of interlocutor feedback in the construction and preservation of common ground during tandem (L1-L2) and control (L1-L1) face-to-face interactions. While previous work in L1 and L2 studies has mostly focused on a single type of feedback, our study brings together different functions (conversational, miscommunication, and corrective) to illustrate how these interact and complement one another, adapting to the specific needs of a particular communicative context. Our findings suggest a distinction in feedback patterns between the two conditions observed: tandem interactions exhibit a higher frequency and complexity of feedback, with a co-presence and combination of corrective, miscommunication and conversational feedback, whereas control interactions predominantly rely on the last type. These results underscore the highly cooperative nature of tandem interactions, where mutual understanding and co-construction are essential for promoting a friendly environment, which is instrumental in language learning.
47The multimodal nature of feedback, integrating verbal, vocal, and visual signals, emerged as a key element in both (re-)establishing and maintaining common ground. Face-to-face tandem interactions, in particular, benefit from the dynamic use of feedback to address linguistic challenges and alleviate asymmetries between the tandem partners. In addition, individual differences in feedback styles emphasize the need for further exploration of speaker and interlocutor variability to better understand its impact on communication and learning outcomes.
48Future studies could expand the dataset to include a larger sample of participants from the corpus (in both English and French), and investigate how feedback may be initiated and provided by the L2 learners as well (which would further document the co-construction and asymmetry reduction characteristics of tandem exchanges between speakers of different L1 backgrounds). More work could also be done on the subfunctions of the feedback tokens within the different types, as well as the forms. In addition, it would be interesting to investigate recurrent multimodal patterns of combination, to see whether certain visual-gestural forms (e.g. head nods) tend to be associated with a certain type of feedback (e.g. conversational feedback).
49In conclusion, we hope our study contributes to a growing body of research in L1 and L2 feedback, opening pedagogical applications for raising tandem learners’ awareness of its role, both in common ground work and in linguistic/intercultural developments.
Flux RSS
