Peer corrective feedback as an opportunity for metalinguistic reflection in tandem telecollaboration 1

EN This paper focuses on peer written corrective feedback (PWCF), a pedagogic device whose potential appears still underexploited in second language teaching in Italian schools and universities. Specifically, we aim to contribute to the body of research on the benefits of PWCF as a learning activity for the development of metalinguistic reflection in peer-to-peer native/non-native online communication. Using a sample of tandem interactions between US learners of Italian and Italian learners of English, we conducted a qualitative analysis of the explanations of linguistic phenomena provided by native speakers when commenting on errors after giving corrective feedback on their non-native partners’ pieces of L2 writing. The data analysis confirmed that the feedback-discussing tasks pushed native/non-native peers to actively reflect on both source and target language, engaging in metalinguistic discussions and utilizing cross-linguistic knowledge.


Introduction
In line with the purpose of this Special Issue, our small-scale study stems from the idea that good (and bad) teaching practices implemented during the pandemic emergency could help us rethink and revise the way we do second language teaching from now on. We take this opportunity to reflect on a well-known pedagogic device-peer written corrective feedback-from a particular angle. Also referred to as "peer response," "peer review," and "peer editing," peer written corrective feedback (PWCF) is the reciprocal activity during which learners provide corrective feedback (CF), i.e., an indication that the partner's use of the target language is incorrect (Lightbown & Spada, 1999), on each other's drafts in pairs or small groups (Storch, 2019;Yu & Lee, 2016). It sometimes includes oral discussion in addition to the written comments on the draft. It can involve L2-learner peers, native-speaker peers, and native/non-native peers; also, it can occur in face-to-face or computer-mediated interactions. The focus of the present paper is on the metatalk that occurs in virtual exchanges when implementing PWCF followed by oral discussion between native (NS) and nonnative speakers (NNS). Specifically, we aim to explore the potential benefits of feedback-discussing activities in terms of metalinguistic reflection, regardless of the correctness and effectiveness of the feedback itself.
From a recent survey among foreign language teachers in Italian secondary schools and universities, it emerged that almost 50% of the respondents have never or rarely used PWCF during the emergency distance learning (Conti, 2021, this issue). This outcome is surprising since PWCF would have been particularly useful in a context where the most penalized aspects of learning were those related to interaction and learners' autonomy and where teachers have often emphasized the lack of contact with and among pupils, a decrease of participation and motivation, and difficulties in actively involving the class (Conti, 2021, this issue). The survey results revealed that PWCF was scarcely used even during in-person lessons before the pandemic, thus suggesting that it is an under-used instrument for second language teaching in Italian schools and universities.
The limited use of PWCF might be due to the fact that learners are considered not qualified to act as "substitutes for the teacher" (Rollinson, 2005). Nevertheless, several studies have shown the advantages of PWCF as a potentially valuable aid in L2 teaching for its social, cognitive, affective, and methodological benefits (Rollinson, 2005, p. 23). Not only can PWCF provide useful suggestions for the improvement of the author's text (Caulk, 1994), but it can also increase feedback givers' ability to critically revise their own writing, thus offering opportunities for students to develop autonomy in learning (Hyland, 2000).
The learning potential of peer feedback in L2 writing has been mainly investigated on nonnativespeaker peers and with a focus on the development of students' composition skills and global aspects of writing (Caulk, 1994;Hyland, 2000;Mendonça & Johnson, 1994). With the current study, we aim to contribute to the body of research on the benefits of PWCF by focusing on its potential for the development of metalinguistic reflection in peer-to-peer native/nonnative online communication. The context is that of telecollaboration, or virtual exchange (O'Dowd, 2021), a learning environment in which participants from different cultural backgrounds work together in online networks, expanding their opportunities to engage in meaningful and goal-oriented communication and developing their foreign language, intercultural, and digital competencies. In our particular case, the participants worked in pairs in tandem arrangement, that is, they were speakers of different L1s who were learning each other's language and alternated the use of the two languages (cf. Tardieu & Horgues, 2020).
In the first part of the paper, we will examine the theoretical underpinnings for PWCF from a languagelearning perspective (notably, the Interaction Hypothesis and the Sociocultural Theory) and provide an overview of the empirical studies that investigated the effects of PWCF. We will then examine the potential of metalinguistic reflection for language learning. In the second part of the article, we will present and discuss the findings of a small-scale observational study with tandem partnerships involved in feedback-providing-anddiscussing activities. Finally, some pedagogical implications will be suggested.

Theoretical underpinnings and empirical evidence on peer written corrective feedback
From a language-learning perspective, the theoretical underpinnings for PWCF rely on two major models, namely the Interaction Hypothesis (Long, 1983(Long, , 1996 and the Sociocultural Theory (Vygotsky, 1978). The interactionists (e.g., Long & Porter, 1985;Pica, 1991;Pica, Young, & Doughty, 1987) suggest that the negotiation of meaning in pair or group work may encourage language acquisition by making input comprehensible through explicit corrections, clarification requests, and confirmation checks. Peer corrective feedback might also lead learners to notice the gaps in their interlanguage, to test hypotheses, and modify their output. Sato (2017) points out that, based on the last 30 years of interactionist research (e.g., Oliver, 2002;Philp, Adams, and Iwashita, 2014;Pica, Lincoln-Porter, Paninos, & Linnell, 1996;Sato & Lyster, 2007;Varonis & Gass, 1985), peer interaction creates more learning opportunities than teacher-learner interaction since learners receive significantly more CF, engage more frequently in meaning negotiation, and tend to modify their initial non-target-like utterances to make them comprehensible for the interlocutor. Furthermore, peer interaction allows learners to freely experiment with language, reducing the anxiety of correction and increasing their autonomy. Peer interaction, therefore, seems to provide a rich learning context, creating "an environment in which learners are willing to take up the opportunity created by PCF and to engage in meaningful output practice" (Sato, 2017, p. 22).
From a sociocultural perspective, cognitive development, including language learning, occurs in social interaction between an expert and a novice. The role of the expert is to provide calibrated assistance (scaffolding) to the novice in order to respond to his or her needs. In the field of second language acquisition, scaffolding occurs not only in teacher-learner interaction but also in peer interaction. In this case, learners provide each other with bidirectional assistance and pool their linguistic resources (collective scaffolding, cf. Donato, 1988Donato, , 1994 to solve the problems they encounter and co-construct L2 knowledge. Peer feedback thus creates "a facilitative socio-interactive environment in which L2 learners receive social support and scaffolding from peers" (Hu & Lam, 2010, p. 373) and provides "a favorable instructional environment for readers and writers to work within their respective [...] ZPD 2 " (Villamil & Guerrero, 1998, p. 495). In this view, peer feedback is particularly valuable for language learning. Referring to peer feedback delivered during collaborative writing tasks, Storch (2019) explains that this kind of feedback is likely to be developmentally appropriate and aligned with learners' linguistic and cognitive capacities to process it. Peer feedback might also be more accessible to learners since it is often accompanied by brief and simple explanations of L2 rules but with little use of complex metalinguistic terms. Finally, this feedback modality is timely and contingently responsive to the learners' needs since feedback is provided when actually needed.
Empirical studies on PWCF have mainly followed the research orientation of Second Language Writing studies, analyzing PWCF effects on the improvement of students' general accuracy and writing skills. They have highlighted some concerns about this pedagogic tool, although its benefits seem to prevail.
According to certain studies (e.g., Guardado & Shi, 2007;Tsui & Ng, 2000;Yoshida, 2008), learners may lack confidence in their ability to provide CF and distrust the feedback provided by their peers; hence, they are reluctant to incorporate it in their drafts. Other researchers (e.g., Hyland, 2000;Leki, 1990;Lockhart & Ng, 1993;Mendonça & Johnson, 1994;Nelson & Murphy, 1992) stress students' difficulties in providing quality feedback. According to them, peer feedback would resort to formulaic comments; it would be overcritical (Amores, 1997) or over-focused on surface errors (McGroarty & Zhu, 1997) or on global issues such as content and organization. Learners would not focus on grammatical aspects unless the task is designed for this purpose.
In spite of these issues, several studies (e.g., Caulk, 1994;Hyland, 2000;Rollinson, 2005) have shown the advantages of PWCF for the development of students' revising skills and learning autonomy. It is generally agreed (e.g., De Guerrero & Villamil, 1996;Min, 2006;Peterson, 2003;Rahimi, 2009) that PWCF is beneficial for student writers, as peer readers can provide useful feedback (Caulk, 1994) and peer writers can revise effectively on the basis of the comments they received (Mendonça & Johnson, 1994). PWCF proved even more appreciated or effective than teacher feedback in some studies (Nelson & Murphy, 1992;Zhao, 2010).
Furthermore, it has been suggested that providing PWCF is beneficial for the provider as well as for the receiver. As Rollinson (2005, p. 24) points out, "becoming a critical reader of others' writing may make students more critical readers and revisers of their own writing" and reinforce their audience awareness (Breggen, 2015). A series of studies (e.g., Berg, 1999;Lundstrom and Baker, 2009;Min, 2005;Yang, Badger, & Yu, 2006) reports that giving feedback is particularly beneficial in improving global aspects of writing, e.g., organization, development, and cohesion.
The reviewed studies generally focused on improving learners' L2 writing skills. However, a learning activity involving PWCF can also deepen learners' understanding of form-meaning connections and promote acquisition of target-like structures (Storch, 2019), offering learners the opportunity to actively reflect on both source and target language, so as to develop metalinguistic awareness.

Metalinguistic reflection and language learning
Defining the term "metalinguistic" is not easy due to its adjectival status that requires supplements carrying different connotations, e.g., "awareness," "knowledge," "activity," "competence," or "reflection" (Watson & Newman, 2017). Its definition may also vary according to the discipline: in psychology, the focus is on cognition and on how one thinks about language; in socio-cultural studies, the focus is on how meaning is created in social contexts; in linguistics, the focus is on language and metalanguage (Myhill & Jones, 2015). Here, we define metalinguistic reflection as the action of using language reflexively to analyze language itself and the connections between form and meaning (Pinto, 2015). For this purpose, a certain awareness of the way language works is needed. Metalinguistic reflection is, thus, an activity relying more on the individual's declarative and explicit knowledge about language-"analyzed knowledge" in Bialystock & Ryan (1985)'s terms or "metalinguistic knowledge" for Gombert (1992)-than on his/hers implicit knowledge implying the procedural ability to use language-"control" for Bialystock & Ryan or "epilinguistic knowledge" for Gombert.
The relevance of metalinguistic instruction and knowledge has been a controversial issue in second language classes. In traditional teaching methods (e.g., the Grammar Translation Method, the Audiolingual Method, the Silent Way, etc.) metalinguistic knowledge was an indispensable part of second language teaching and learning (Richards & Rodgers, 2014). These "focus on forms" approaches (Long, 1991) generally resulted in good levels of accuracy in grammar tests but learners showed difficulties using the L2 in real-life situations. New approaches focusing on meaning and having communicative competence as the main goal have, therefore, been developed. With the spread of communicative language teaching (CLT), thus, metalinguistic instruction and knowledge have been increasingly marginalized. More recently, a certain attention to language structures in programs whose focus is on meaning has been claimed to be positive especially for adult learners.
In the field of second language acquisition, the theoretical debate on the relationship between explicit knowledge and L2 learning has also raised doubts about the role of metalinguistic knowledge. Some researchers (e.g., Krashen, 1981;Paradis, 1994) claim that no connection exists between explicit knowledge and L2 competence and performance; other researchers (e.g., DeKeyser, 2003;De Jong, 2005;Johnson, 1997) estimate that explicit knowledge about language can be proceduralized and automatized and, therefore, directly used in real-time language production; others (e.g., Ellis, 1994;Smith, 1991) affirm that explicit knowledge can make an indirect contribution "to the acquisition of implicit knowledge by facilitating attention to form in the input" (Ellis, 2004, p. 228).
Despite those different positions, the benefits of metalinguistic reflection are of no lesser importance for language learning and have been supported by many researchers. For example, Schmidt (1990Schmidt ( , 1994 and Skehan (1998) underline how, in metalinguistic activities, attentional resources are focused on language facilitating noticing, awareness, and, consequently, learning. According to R. Ellis (1994) explicit metalinguistic knowledge may play a facilitative role in L2 acquisition by accelerating the establishment of links between form and meaning. For N. Ellis (2008), reflecting on language may make certain grammatical features more salient and hence more noticeable to learners, and explicit knowledge can contribute to linguistic problem-solving and to conscious output production.
These claims are supported by many studies on adolescent and adult learners in classroom settings (e.g., Ellis, 2006;Elder & Manwaring, 2004;Renou, 2000Renou, , 2001Roehr, 2008), which show significant correlations between students' metalinguistic knowledge and their L2 proficiency. Investigating the interactions between learners working in pairs on a text reconstruction task, Storch (2008) showed that learners' metatalk over a range of grammatical and lexical items led to learning/consolidation of the structures they focused on for both members of the pair, especially when they showed elaborate engagement, pooling their linguistic resources, testing hypotheses, and providing each other with useful knowledge about meaning of words, grammatical rules, and conventions.

Aims and methodology
In the present paper we aim to observe whether and how metalinguistic reflection occurs in peer-topeer native/non-native (i.e., L1 and L2 English) online interaction stimulated by a particular type of CF, namely indirect written feedback followed by oral discussion. The study is observational in nature and does not involve any experimental manipulation. A small sample of interactions will be qualitatively analyzed to explore one of the potential benefits of feedback-discussing tasks, regardless of the correctness and effectiveness of the feedback itself.

Participants and data collection
The data come from a telecollaborative (TC) program that was implemented in the fall semester of 2018/2019 between Italian and US (Californian) university students (cf. Nuzzo & Cortés Velásquez, 2021). Each Italian student was paired up with a US partner, and they met regularly on Zoom, videorecording their meetings. The participants were between 20 and 30 years old. They had an intermediate to advanced proficiency level in the target 3 language, and they often had more than two languages in their repertoire. Specifically, many students from the US university had a Latin-American background, with different levels of proficiency in Spanish. English may not have been the first language of some of them. However, for all of them English was the language of instruction and daily interaction with teachers and peers. Similarly, some students from the Italian university had an immigrant background, but they used Italian as their first language in the academic context. Four dyads were randomly selected for the qualitative analysis of this small-scale study (see table 1, where the first three letters of the participants' surnames were used to anonymize their identities).
The TC program aimed at giving learners opportunities for meaningful and goal-oriented communication through a set of macro tasks to be completed collaboratively, but entailing some individual work too (e.g., organizing a trip; reviewing a film; writing a short story). Each macro task included several subtasks: students had to interview each other, write a text in the L2, provide WCF on their partner's text, and then discuss errors and corrections in a videocall. For the present study, we focus on the feedback discussion phase, which was implemented to prevent the participants from misunderstanding or avoiding the correction, a problem emphasized by Hyland and Hyland (2006). Furthermore, this feedback discussion phase was added based on the assumption that the metatalk occurring when learners talk about the language they have produced may deepen their "knowledge about language use, about the relationship between meaning, form and function" (Storch, 2008, p. 96).
Participants received written instructions on how to provide and discuss feedback. As for the written phase, they were asked to signal the errors and explain what was wrong, without giving the correct forms (indirect WCF). During the feedback discussion phase, they were asked to help their partner find the correct form (oral prompt). On several occasions, however, the instructions for the feedback-providing task were not followed consistently in the discussion phase during the video-call sessions. The NSs sometimes did not help their partners find the correct form, but instead provided it themselves.
Language alternation during oral communication was dictated by instructions. The feedback discussion phase was held in the feedback providers' L1, the language in which the text was written (L2 for the author/feedback receiver). For the present study, four English feedback-discussion extracts have been analyzed (see table 1 for an overview). Three of them are taken from session 3, belonging to the macro task of organizing a 3-day trip for the partner; whereas one is taken from session 9, belonging to the macro task of writing a short story 4 . The extracts, transcribed according to a simplified version of the CHAT Transcription Format (MacWhinney, 2000), have different durations, as no time limitation was set for the completion of the task. About one hour of conversation in total was analyzed.

Data analysis and discussion
In each of the four extracts considered for this study, the Language Related Episodes (LREs) were identified (see table 2). According to Swain and Lapkin (1998), an LRE is "any part of a dialogue where the students talk about the language they are producing, question their language use, or correct themselves or others" (p. 326). In our case, each error commented by the NS feedback provider, and the relevant negotiation with the NNS partner, was counted as a new LRE. The number of LREs in each extract is, therefore, related to that of the mistakes signaled by the feedback provider in his/her partner's writing. Similarly, the topics discussed depend on the linguistic items addressed by WCF. As can be seen in the examples reported below (1-11), the participants focused on different aspects of English morphosyntax, spelling, vocabulary, and mechanics. The LREs varied in length and in the participants' contribution to the resolution of language-related problems, or the explanation of linguistic phenomena. Referring to the categorization proposed by García Mayo and Azkarai (2016), some showed limited engagement by one participant, like in Example (1), or by both, like in Example (2), whereas others exhibited elaborate engagement by both -see Example (3) 5 .

don't) understand what is the difference between
seat and the-the other verb that you (re-that you write it) GAR: ok ehm # basically the verb-the verbal phrase# eh to seat would be BE SEATED because it needs the auxiliary verb to <come along with it ## yeah> MIG: <ah ok # ok> GAR: yeah while sit it's just a simple verb so# <it's up to you> MIG: <what does it mean> in Italian? ## sedersi GAR: eeh sedersi it could be be seated or sit so it has two translations in English for some reason MIG: ok ## ah perfect now I understood Any of the analyzed LREs entailed a certain degree of metalinguistic negotiation. However, the more elaborate the LRE, the deeper the level of effort shown by the participants. Any time the feedback providers were trying to explain the reason(s) why they provided WCF on a non-target-like linguistic element, the effort they and their partners made to understand and/or fix the problem compelled them to stretch their metalinguistic resources. In the following, two exemplary LREs from each of the four extracts will be reported and commented on. The examples have been selected so as to give an idea of the different linguistic phenomena addressed by the students.

Extract 1 -GAR-MIG
In the LRE reported in Example (4), the NS GAR tries to prompt his partner to replace I've been with I went in the sentence last summer I've been in Ischia. In the first part of the LRE he uses Italian, but then he shifts to English as MIG reminds him that they were expected to use English for this phase, according to the instructions. While explaining why I've been is not suitable in that context, he refers to the fact that the action is finished. He also uses some grammatical terminology (passato prossimo, imperfect), although inappropriately (his passato prossimo refers to English simple past, whereas imperfect refers to Italian trapassato remoto). MIG makes an attempt, referring to her experience as an instructed learner of English (I studied this thing), but she fails and eventually GAR provides the target-like form (I would personally say I went). Interestingly, GAR tries to help his partner by comparing the use of verbal tenses in Italian and English. His metalinguistic reflection involves cross-linguistic awareness.

Example (4)
GAR: ok eehm quindi mmh ## il tuo ok ### last summer I've been in Ischia also called the green island eemh # qui invece di usare I've been MIG: <yes> GAR: <puoi> usare altro tempo verbale ## che: # como hai già finito questa-questo viaggio è già finito # puoi usare il-# il passato prossimo MIG: ok ##### maybe we should eeeh speak in English this part of the meeting GAR: oh right yes yes I'm sorry # MIG: (no problem) GAR: basically by using I've been MIG: yes GAR: it-it does indicate that you've been there but it doesn't really indicate that the action is finished MIG: oh yeah yeah <I remember> GAR: <yeah> MIG: I studied this thing yes ## so:: eee I-I had been? GAR: you-e:hm I would personally say I WENT <meaning that> MIG: <I went> GAR: yes (that's what I would use personally) MIG: and I had been i::s an error o::r # is a mistake I have been o:r GAR: it's not really an error ehm ## but it-it would indicate like I had been there before something else kinda like # it's similar to the imperfect MIG: ok GAR: yeah ## so it's better just to use the past tense ### MIG: ok In the Example (5) from Extract 1, GAR comments on the spelling of the word sea, which MIG has confused with the homophone see. The NS uses both a technical term (it's an homophone) and a more simple explanation (it's just spelt differently) to help his partner understand her mistake. He also resorts to Italian to solve the ambiguity between the homophones (you used see … which means 'vedere').
Example (5) from Extract 1 GAR: eeh for the night I was (looking on Airb&b) ## ok # so:: basically # for the second sentence of-eh for the night paragraph (the apartment) is very close in fact to the sea # <you used see> MIG: <yes> GAR: which means vedere it's an homophone with the <other way to-> MIG: <yeah yeah yeah> GAR: ### yeah it's just spelt differently MIG: yeah sea with the a

Extract 2 -GEN-CIU
In Example (6), taken from Extract 2, the learners discuss some orthographic differences between UK and US English. The starting point is the feedback provided by the US participant GEN on the spelling of the word recognized. Interestingly, GEN seems to be aware of the issue but not really able to provide an explanation. She makes some effort to recall the rule, and eventually it is her linguistic partner who explains how the spelling works in the two varieties (I think that recognize should be with s in British English and probably yeah with the z in American English).
Example (6) GEN: ok so the first thing that I noticed was # uh the way you spelled recognized # but # the thing is that # I know in american english like ## uh it's different than english english sometimes # like <sometimes ## yeah> CIU: <yeah # the s> and z <often> GEN: <yeah> CIU: yeah GEN: I'm not sure if ## <in English English> CIU: <yeah I think tha::t> ## recognise should be with s in british english <and> probably yeah with the z # in american english yeah GEN: <yes> ### ok then # that's fine ## <um # xxx> if you're using english english but # that just stood out to me Again, in Example (7), the metalinguistic discussion is led by the Italian participant CIU, who refers to his experience as an instructed learner of L2 English. The two are commenting on the sentence she succeeded in focusing the whole nation attention on this event, and specifically on the use of the possessive 's'. GEN admits she is not able to provide a convincing reason for the need to use the possessive 's' with inanimate objects (I don't know), so her partner tries to find an explanation referring to the phenomenon of language change. The native speaker accepts her partner's explanation.
Whereas in Extract 1 the NS shows deep metalinguistic awareness of English and the ability to justify the feedback provided comparing English and Italian, in the second dyad the US participant appears far less aware or confident of how her L1 works. Nevertheless, the need to discuss with her partner after providing feedback on his writing "forces" her to reflect on the reasons behind her corrections.

Example (7)
CIU: ah ok because I used nation attention without the possessive s GEN: yeah exactly CIU: I used it without possessive s because long ago I was told that uh just animated things such as humans beasts animals would be able let's say this to use the possessive s GEN: ok CIU: so I mean the bottle mobile phone laptop # nation as I interpreted it # without being animated # would not use the possessive s so that's why I put nation GEN: <a:h ok but-> CIU: <but you say that> this is correct # I think maybe# I mean nation is composed of people # animated # so::: GEN: right CIU: it would probably work yes GEN: yeah CIU: I'm not sure about this actually because I mean I was told that u:h ## now for something like <like> GEN: <xxx> CIU: twelve years ago so ma-maybe the-the:: language evolved in thesein these years probably now it's more common to use possessive s GEN: <yeah because> CIU: <whenever> a sort of possession # so:: GEN: even if you were talking about a something that's not animated like # you said the cell phone like if you were talking about something about the cell phone like oh the cellphone's # <screen> CIU: <charger> <or screen> GEN: <yeah> cellphone's charger like you'd still put the possessive s CIU: mmh GEN: like it would sound kind of weird to say like this cellphone screen like CIU: uhm ok GEN: I don't know CIU: no-eh I-I mean mmm as all languages evolves that's probably:: normal because it was rather strange ((clearing throat)) and difficult to remember back way when I learned it # bu::t I think that should # yeah that should make sense I mean it's an evolution to:-towards simpler use of the:: language so GEN: right CIU: I think that would be fine because also it's not the first time I actually see an un-unanimated thing with the possessive s it's just that I'm not used to:: # to this use of the possessive s so:: # well ok so good to know that nowadays it's it's good to use possessive s # ok # a::nd ## let's see ## tutu

Extract 3 -GER-STU
In the first LRE reported from Extract 3 (Example 8), the US participant GER is explaining that the toponym Cinque Terre, although plural in Italian, is singular in English. In her document, the NNS STU used it as a plural, and she seems hard to convince that this choice is unacceptable in English. In her view, the toponym should be plural in English too, as it refers to a plural entity (Cinque Terre are five). GER needs to make some efforts to succeed. In order to further clarify the issue, he refers to another geographical name with plural form in Italian (Marche), thus showing a noticeable cross-linguistic awareness.

Example (8)
GER: and the:n oh # so for Cinque Terre # that's-<in English> that's not-it's a singular thing STU: <mmm mmm> ### ah okay # I-I didn't know that GER: ((smiling)) yeah it's # I wouldn't expect anyone too really but it's just like # just grammatically like # any <location> STU: <mmh mmh> aaa okay GER: <any location xx> STU: <mmm mmm ok> ok # ok GER: ((clearing throat)) the::-like if I wanted say # I am going to study in Recanati ## and then I would say Recanati # is IN le Marche or <it's just one-xxx and the::n compromise> STU: <okay but ## xx> because:: ehm # mm # from my perspective is just that Recanati is one village but Cinque Terre are five # <and I-I did in> GER: <ah yeah yeah> STU: in the plural for that reason GER: so I meant to say like I am going to study in le Marche ## STU: ah ok yeah GER: <which xx-> even though it's grammatically plural it's kinda # yeah <and the:n e:hm> STU: <yeah but it's> I don't know because if the name of the region it's not <something> GER: <right> right right I see what you mean STU: yeah okay <xxx anyway> GER: <it's still-> it's still for some reason in English singular ((laughing)) The second example from Extract 3, which is reported in Example (9), includes a discussion on the use of the comma before the last element of a list. Again, GER proposes a comparison between English and Italian (which I know is different from Italian) to appear more convincing to his partner. STU contributes to the discussion by confirming that Italian has a different rule (if you say 'e' you don't have to use the comma).

Example (9)
GER: the:n # local cuisine # wine tasting ((coughing)) so:: whenever there's a list of three or more things ## STU: <mmh mmh> GER: <(whether)> you are writing in English # before the AND you have to have a comma STU: ah okay GER: like between the second to last and last items in a list you have to have a comma STU: yeah # ah # okay ok GER: which I know is different from Italian of ## (xx) points <(of)> STU: <yeah> GER: (and the rest) before so <I am aware> STU: <yeah> ## yeah because if you say E # you # don't have to use the comma GER: right cause for English you do STU: it's not correct yeah # yeah

Extract 4 -TAP-PIE
In the LRE reported in Example (10), the NS TAP is trying to explain why she suggested replacing the expression a beautiful country to be visited with a beautiful country to visit. To justify the inappropriateness of her partner's sentence, TAP refers to a general quality of English writing, which she describes as more straightforward. The Italian participant contributes with a confirmation check in which she proposes a metalinguistic term for the inappropriate expression (so is not the passive form).

Example (10)
TAP: I:-I would use a different arrangement like for example I would suggest you to use Italy is a beautiful country to visit # PIE: ah ok TAP: like not-not-don't-don't say to be visited because it it's <more like> English is more straightforward # PIE: <ah> ### <mmh mmh> TAP: <language> so you can just say to visit PIE: ok TAP: ok? PIE: so is not the passive form TAP: no no passive # PIE: <ok> TAP: <it's-it's> I mean it's it's correct, it is ehm there's nothing wrong with that # e:h em except that we rather say to visit like more straight to the point In the last Example (11), TAP is suggesting using shorter sentences, which would be more appropriate in English writing. Interestingly, when her partner introduces the cross-linguistic reflection, namely a comparison between English and Italian about the length of sentences (we Italians have the problem to create big phrases), the US participant refers to a third language, Spanish, that they have in their repertoire.
Example (11) TAP: ehm mmmm ### the only thing I would suggest is to use # commas o::r separate sentences like # mmm PIE: <ok> TAP: <(probably)> two sentences divided by a comma and then you can use a period and start a new sentence:: # eh I get you <becaus-> PIE: <xx we italians> have the problem to::-to::-to create big phrases ((laughing)) TAP: <yeah I get you becaus-> PIE: <about three four> TAP: yeah that-<that> is the same with spanish PIE: <yeah> TAP: we tend to create like <huge sentences> without punctuation and it make sense but in English is a little <bit different> PIE: <yeah> #### <yeah> TAP: you have to have like a sentence and a period and a comma if you're going to-if you're going to join two sentences you need to have the comma:: or the proper connectio::n it's a little bit trickier yeah PIE: yeah TAP: but ehm-but ehm <xx> PIE: <I have> this problem I know it ((laughing)) TAP: if you want me to help you in that I can help you # for example # in that same paragraph # e::h PIE: <mmm> TAP: <you have> moreover a fantastic accommodation for the night is the Faro Bianco Gallo-Gallipoli ### you can use a <period> PIE: <yeah> TAP: and then # just start by <sayin-> it is located ### PIE: <yeah> ## it # yeah? TAP: it is located one kilometer-kilometer # from downtown <e:hm> PIE: <ok> TAP: ((smiling)) and then we have another period there PIE: ((laughing)) ok The examples reported show that the English NSs often engaged in interesting metalinguistic discussions with their NNS linguistic partners. The feedback-discussing task pushed them to actively reflect on their languages. Particularly, the need to explain the reason(s) why they had provided corrective feedback on a non-target-like linguistic element led the NS feedback providers to stretch their metalinguistic resources and make their explanations clear and convincing to the partner. In some cases, the NSs resorted to grammatical terminology, whereas in other cases they offered less technical explanations. Despite what has been claimed in some previous research on PWCF (e.g., Guardado & Shi, 2007;Hyland, 2000;Leki, 1990;Lockhart & Ng, 1993;Mendonça & Johnson, 1994;Nelson & Murphy, 1992Tsui & Ng, 2000;Yoshida, 2008), our study shows that learners are autonomous and able to provide feedback on (a variety of) language-related issues and to adopt different strategies to make sure that their partner understands the error. Hence, the provided feedback tends to be accessible and aligned with the peer's linguistic and cognitive capacities (Storch, 2019). In terms of the benefits for CF providers, the strategies that they have adopted to provide and discuss the feedbacknotably, cross-linguistic comparisons among the languages in their repertoires and meta-linguistic reflection through the recalling of grammar rules-can be seen as actual learning strategies. Not only might they be used to increase metalinguistic awareness in L1 and gain a better understanding of its formal and functional features, but these strategies might also be applied to L2 learning as self-correction strategies that will considerably increase learner autonomy.
On several occasions, comparisons between English and Italian were introduced in the negotiation, thus stimulating cross-linguistic awareness. Given that most of the participants in the online tandem program also had Spanish in their repertoire, this third language was occasionally involved in the cross-linguistic comparisons. Those comparisons might encourage the development of metalinguistic skills and a more conscious relationship with linguistic tools, considering that "esercitare la funzione metalinguistica riflessiva attraverso il confronto con un oggetto estraniante è già, direttamente, migliorare l'uso della lingua" (De Mauro & Boylan, 1995, p. 13).
In some cases, it was the NNS who led the meta-and cross-linguistic discussion. As instructed learners of L2 English, the Italian participants could refer to an explicit knowledge of some linguistic features that their NS partners might lack. The feedback-providing-and-discussing task gave the US NSs more opportunities to deepen their reflection on the use of English than they might usually have.

Conclusion
In this paper, we started from observing that PWCF is a pedagogical resource whose potential is still underexploited in second language teaching, at least in Italian schools and universities. This is rather surprising, as the benefits of peer feedback have been largely outlined in the literature on L2 writing, second language instruction, and education in general. To contribute further to the body of research on PWCF and encourage language instructors to use it more with their students, we decided to explore the potential benefits of this pedagogical tool from a particular perspective. Through the qualitative analysis of a small sample of peer native/nonnative telecollaborative interactions, we observed how feedback-providing-and-discussing activities can engage learners in metalinguistic negotiation, regardless of the feedback's correctness or effectiveness. In order to collaborate in the resolution of a linguistic problem, the participants resorted to different strategies, showing autonomy and creativity.
Based on our small-scale observation, we can suggest that PWCF and its subsequent discussion in dyads during telecollaborative exchanges is a valuable activity for language learning in terms of the meta-and cross-linguistic reflection that it encourages. After the meaning has been communicated in the written text, PWCF leads learners to focus and reflect on the linguistic forms that are particularly relevant due to the presence of the error. Considering that the overriding focus of telecollaborative programs is on meaning, with little or no intervention by the language instructor, PWCF activities can represent a fruitful way to integrate metalinguistic reflection into these low-structured learning environments.
As previously announced, a rather problematic issue that emerged in our study is that, despite having received written instruction on how to provide indirect WCF accompanied by oral prompts, learners did not always follow the guidelines. In some occasions they directly provided the correction, which generally resulted in limited engagement by both learners and a poor or no metalinguistic reflection. In order to avoid this issue and encourage learners' metatalk, it would, therefore, be helpful to organize more explicit and exhaustive feedback training sessions before PWCF activities. In those sessions, teachers may give learners more structured feedback rubrics (Sato, 2013) or, at least, provide them with some instruction on what errors they should or should not correct and how. As a matter of fact, although PWCF activities are mainly autonomously led by learners, their benefits might be maximized by teacher guidance.
Francesca La Russa, Freelance Researcher franlarussa3@gmail.com EN Francesca La Russa holds a PhD in Applied Linguistics. Her research interests include second language acquisition and teaching with a focus on Italian as a second language, corrective feedback, and task based language teaching. ES Francesca La Russa es Doctora de Lingüística Aplicada. Su investigación se centra en la adquisición y enseñanza de segundas lenguas, con especial atención al italiano como segunda lengua, al feedback correctivo y a la enseñanza de lenguas basada en proyectos.
IT Francesca La Russa è dottore di ricerca in Didattica delle lingue. I suoi interessi di ricerca vertono sull'acquisizione e l'insegnamento delle lingue straniere con focus sull'italiano lingua seconda, il feedback correttivo e il Task Based Language Teaching.
Elena Nuzzo, Università Roma Tre elena.nuzzo@uniroma3.it EN Elena Nuzzo is Associate Professor of Applied Linguistics at Università Roma Tre. She teaches and researches in the fields of second language acquisition and teaching, with a focus on Italian as a second language. Her main research interests are the applications of speech act theory to second language learning and cross-cultural communication, and task based language teaching. ES Elena Nuzzo es profesora titular de Lingüística Aplicada en la Universidad de Roma Tre. Imparte docencia y realiza investigación en las áreas de adquisición y enseñanza de segundas lenguas, con especial atención al italiano como segunda lengua. Su investigación se centra principalmente en la aplicación de la teoría de los actos de habla al aprendizaje de segundas lenguas y a la comunicación intercultural, y el enfoque por tareas en la enseñanza de lenguas.