Language use and intercomprehension in telecollaboration among Italian mentors and heritage Spanish speaker mentees

This study addresses multilingualism in the paradidactic setting of telecollaboration. During the Fall and Spring semesters of the 2018/2019 academic year, a telecollaborative program was implemented with students from Roma Tre University (R3) who served as native Italian speaker mentors and students enrolled in Italian classes at California State University, Long Beach (CSULB). CSULB offers targeted Italian learning courses for the substantial population of heritage speakers of Spanish to exploit the typological proximity of Italian and Spanish through intercomprehension. The telecollaborative program employed two different modalities: mentoring and partnership. In this study, we focus on the mentoring, to which


INTRODUCTION
Language distance has been identified as a crucial factor in the acquisition of a new language (e.g., Gundel & Tarone, 1992;Muñoz et al., 2018).For instance, speakers of Spanish may find it relatively easy to learn Italian due to the close linguistic relationship between the two languages.Conversely, Chinese speakers may face greater difficulties when learning a language like French due to significant differences in linguistic structures and features.Notably, Schepens et al. (2013aSchepens et al. ( , 2013bSchepens et al. ( , 2016) ) have demonstrated that learners of Dutch who are also speakers of languages closely related to Dutch tend to perform better than those whose languages are less closely related.These studies emphasize that the smaller the linguistic distance in terms of grammar, vocabulary, and pronunciation between a first language (L1) and a second language (L2), the higher the learnability of L2.This concept is rooted in the idea of language transfer, where knowledge from one language can have both positive and negative influences on the learning process of another language.
Another influential factor in language acquisition is the number of languages in a learner's repertoire.Schepens et al. (2016) note that the impact of a linguistic repertoire on acquiring an additional language has long been a topic "wreathed in controversy" in second language acquisition research.However, determining the precise number of languages in an individual's repertoire is not a straightforward task.Furthermore, scholars use the terms bilingualism and multilingualism ambiguously.Therefore, in line with Aronin and Jessner (2014), the authors of this article consider bilingualism as a case of multilingualism.Consequently, the term "bilingual-ism" will only be employed when distinguishing between two languages in a speaker's repertoire.
The study presented here is shaped by Berthele's (2020) conceptualization of multilingualism.According to this concept, the author defines multilingualism as a natural category that distinguishes between (a) prototypical multilinguals who possess at least two languages at the same level of proficiency; (b) regular multilinguals, for whom one language is dominant, while competencies in the other(s) may be high; (c) and a limiting case, in which one language is dominant, with competencies in the other(s) limited to vocabulary or listening comprehension of household exchanges.One of the possible instances of (c) is the case of receptive multilingualism, also known as mutual intelligibility, and intercomprehension (IC).IC is defined as the phenomenon by which speakers with different first languages (L1s) are able to comprehend each other by using their respective L1s and/or a bridge language, that is, an L2 closely related to their interlocutor's L1 (Bonvino, 2015;Möller & Zeevaert 2015;Van Bezooijen & Gooskens 2007).IC can be considered the most direct way to achieve multilingualism (Peyer et al., 2010).This means that the speaker of Spanish, in the example above, who has never learned/acquired any other language before, can be considered a (receptive) multilingual since the linguistic proximity of Spanish with Romance languages allows them to understand these languages, albeit to varying degrees.While many studies have investigated the intercomprehension of unknown but related languages in comprehension and/or translation tasks, scarce attention has been devoted to the role of intercomprehension in tandem exchanges, when the Target Language (TL) of one of the partners is closely related to one of the languages in their linguistic repertoire.
In order to examine this role, this article presents the findings from a study that examined the learning of Italian in an ongoing digital telecollaboration mentoring program between Italian students and US students, the latter group composed of both Heritage Spanish Speakers (HSSs) and Non-Heritage Spanish Speakers (NHSSs).The main objective of this study was to explore the way mentors and mentees exploit their multilingual repertoires during interaction, since it is assumed that the similarity between Spanish and Italian is an important factor in the language learning process.

Telecollaboration and negotiation of meaning
Telecollaboration (TC) can be defined as a type of online learning arrangement between geographically distant participants for the development of language and intercultural competence (Akiyama, 2018).The TC environment can be defined as "paradidactic" since it refers to a low-structured learning online context in which the teacher's role is primarily focused on promoting the meaningful use of the TL through exposure to input and/ or collaboration with peers, rather than teaching the languages themselves (Cortés Velásquez & Nuzzo, 2021a).
The growing body of research shows that TC programmes provide a natural environment in which learners, especially those in the context of foreign language learning, have the opportunity to engage in meaningful and goal-oriented oral communication using the TL (Lee, 2007;O'Dowd, 2006;Schenker, 2017;Thorne, 2006).Moreover, in comparison with faceto-face interaction, there is a greater amount of collaboration among learners (Chun, 1994), learners produce more language (Kern, 1995), feel more relaxed (Chun, 1994), and contribute more (Kern, 1995;Warschauer, 1996).The advantages of TC can be summarised in terms of the development of different types of competencies: intercultural, linguistic, pragmatic; other types of knowledge, such as digital literacies and multiliteracies; and the development of learning autonomy.Thus, they represent a valuable resource for language instructors whose interaction with students is limited to only a few teaching hours per week, which is not enough for students to achieve high levels of proficiency (Cortés Velásquez & Nuzzo, 2021b).Moreover, individuals learning languages other than English often lack opportunities to engage in meaningful conversations in their TL beyond the classroom.
Conversation stands out as a particularly rich context for language acquisition due to its ability to enhance various aspects of second language development.Interactions provide increased opportunities for learners to negotiate meaning, effectively connecting input, attention, and output (Long, 1996).The concept of negotiation of meaning refers to a specific type of interaction where interlocutors collaboratively strive to overcome communication breakdowns and achieve mutual understanding.Yuksel and Banu (2014) observed that a growing body of studies has confirmed the occurrence of negotiation of meaning (NoM) episodes even during video sessions.As highlighted by van De Zwaard and Bannink (2016), NoM events typically involve an initial indication of non-understanding and a subsequent repair sequence.Researchers have identified three distinct types of NoM episodes: (a) Clarification requests: episodes in which one speaker seeks assistance in understanding the other speaker's preceding utterance through questions in statements such as "I don't understand, " or imperatives such as "Please repeat" (Pica, 1987).These episodes are intended to elicit clarification, normally in the form of reformulation or repetition of the other student's preceding utterance (Clavel-Arroitia, 2019).(b) Comprehension checks: episodes in which one speaker attempts to determine whether the other speaker has understood a preceding message (Pica, 1987).In these episodes a speaker confirms that he/ she has understood or acknowledged the other student's acceptance (Clavel-Arroitia, 2019).(c) Confirmation checks: episodes in which one speaker seeks confirmation of the other's preceding utterance through repetition, with rising intonation, of what was perceived to be all or part of the preceding utterance (Pica, 1987;Clavel-Arroitia, 2019).For instance, expressions like "right?", "ok?" or "do you understand?" clearly show an effort on the part of one of the students to anticipate and prevent a break-down in communication.

Heritage language speakers, intercomprehension and L3
As a country with a high number of immigrants throughout its existence, the United States is a nation formed by heritage language speakers.By heritage language we mean a language spoken at home by bilingual children of immigrant parents, with the language skills of these speakers covering the entire spectrum of language proficiency.Though a heritage language constitutes one of the languages in the linguistic repertoire of individuals living in multilingual societies, scholarly attention to the unique set of political, linguistic, interpersonal, and social conditions characterizing the heritage speaker's life in languages has only recently become the focus of research endeavours, most of which have originated in the United States, due to the linguistic history of the country, alluded to above.Thanks to a series of relatively recent handbooks, monographs, and collected volumes, the linguistic profiles of heritage language speakers have emerged as a unique subset of learners of interest in language acquisition studies.Indeed, their heritage and L3 language acquisition trajectories are the subject of new research studies (e.g., Montrul, 2016;Montrul & Polinsky, 2021;Polinsky, 2018).In the United States, the ever-growing number of Spanish-heritage speakers constitutes a significant subject of interest, with a wealth of studies in its own right, as exemplified by Potowski (2018).
The study of heritage languages and heritage speakers in the US is only of recent vintage, coinciding with the first National Heritage Language Conference1 .Heritage language research is contributing to our understanding of language acquisition, preservation, and attrition in the context of multilingualism, together with a growing appreciation for the effects of multilingualism on language behaviour (Brown & Bousquette, 2018).The status of heritage languages and their particular sociolinguistic environment is pertinent in assessing language performance (Rothman, 2009).In the case of heritage Spanish speakers, new waves of migration create cyclical multilingualism, reinforcing the presence of the heritage language and its evolution in the heritage speaker's identity.The acknowledgment that heritage speakers navigate linguistic and cultural blending throughout their lives as multilingual individuals, has surfaced as a distinctive aspect worthy of examination.Thus, heritage speaker multilingualism defies rigid classifications of L1 and L2, simultaneously questioning established national monolingual frameworks and language ecosystems.These frameworks have typically placed heritage languages in the context of a heritage speaker's past (Baker & Jones, 1998), rather than considering the current and future roles of heritage languages within contact-rich environments.
Over the past two decades, the field of third language (L3) acquisition has garnered increasing attention.The primary goal of these investigations is to delve into how learners' pre-existing linguistic knowledge shapes the process of acquiring an additional language, specifically the L3.
Distinguishing itself from the realm of second language acquisition, third language acquisition stands apart and cannot be simply construed as the mere amalgamation of various linguistic systems (Safont Jordà, 2005).Notably, a central focus of research in the context of L3 use revolves around the concept of cross-linguistic influence (Hammarberg & Hammarberg, 1993;Williams & Hammarberg, 1998).Furthermore, an exploration of typological effects on the acquisition of an L3 suggests that when a typological similarity between the target language (TL) and the languages already present in the learner's repertoire exists, it tends to foster positive cross-linguistic influences (e.g., Ard & Homburg, 1983;Kellerman & Sharwood Smith, 1986;Odlin, 1989).
To capitalize on these advantageous cross-linguistic influences, certain scholars (e.g., Bonvino, 2015) have introduced the pluralistic approach of intercomprehension (IC).This approach has developed through observing how speakers with different L1s effectively communicate using their own native languages and/or a bridge language -an L2 closely related to their interlocutor's first language.This makes it possible to avoid using a lingua franca such as English (Meissner & Burk 2001;Meissner et al. 2004;Möller & Zeevaert 2015;Van Bezooijen & Gooskens 2007).The studies from the field of IC provide cases addressing spontaneous comprehension and production of a third language, especially when the third language is also a Romance language.These studies have observed that the extent to which a learner understands a new language depends on several factors, including differences or similarities in phonology, orthography, syntax, the number of cognates that the languages share, the context, the number of languages in the learner's linguistic repertoire, and the metalinguistic awareness (Heeringa et al., 2013).Several studies have investigated IC in translation tasks (Swarte et al., 2015;Mieszkowska & Otwinowska-Kasztelanic 2015;Smidfelt 2015;Smidfelt & van De Weijer, 2019), showing how participants use the languages in their repertoire to comprehend a text in an unknown language.Others have analysed the IC potential through word recognition activities (Möller & Zeevaert, 2015).
However, to our knowledge there are no studies investigating the role of Spanish as a pivot language in the learning of L3 Italian from an IC perspective.By pivot language we mean a language in the students' linguistic repertoire (in our study, Spanish) that may facilitate the learning of a third language belonging to the same language family, in this case Italian.

Aim of the study
This exploratory study aims to investigate the use of linguistic repertoires by heritage Spanish speakers (HSSs) and non-heritage Spanish speakers (NHSSs) during small talk interactions in mentoring sessions with Italian partners.This study is framed by the following two research questions: RQ1: How do the HSSs and NHSSs use their linguistic repertoire when engaged in small talk during the mentoring sessions with Italian partners?RQ2: Are there differences between the HSS group and the NHSS group in the use of English, Italian, and Spanish?If so, does intercomprehension competence play any role in their telecollaborative sessions?

Context
California State University, Long Beach (CSULB) is designated as a Hispanic Serving Institution (HSI), with a student body of 40,069 students, 46% of whom are Hispanic.Within this substantial Hispanic population, a significant number of students identify as speakers of Spanish as a heritage language.As a result, CSULB has developed an Italian-language learning program tailored specifically for Spanish heritage speakers.This program has followed a trajectory similar to that of Portuguese for Spanish Speakers programs, which have taken root in universities across the United States (Carvalho et al., 2010;Oliva et al., 2019).Furthermore, there is a growing interest in teaching typologically related languages to heritage speakers of Spanish, as demonstrated by the work of Carvalho and Child (2018).
During the Fall and Spring semesters of 2018/2019, a telecollaborative program was implemented with students from Roma Tre University (R3) and CSULB.The telecollaborative program was structured using two types of collaboration, namely mentoring and exchange.These modalities correspond to those labelled by Ware and O'Dowd (2008) as e-tutoring and e-partnership.In this study, we focus only on the former, also widely known as telecollaborative mentoring.In the mentoring collaboration, one group (R3) was completing internship hours for a course on Teacher Education, therefore their role was meant to provide guidance and help in the learning of the Italian language for the students from CSULB.In the case of the exchange, the R3 and CSULB students worked on assignments following peer-collaboration strategies.
The main purpose of the program, then, was to provide CSULB students with the opportunity to engage in meaningful and goal-oriented communication in the TL beyond the scant opportunity to do so in the traditional educational context.At the same time, the program provided the R3 students with the possibility to practice their teaching skills.

Participants
The participant sample (N = 69) consisted of two groups of undergraduate students, 15 mentors who studied Second Language Teaching at R3, and 54 mentees, first-year Italian language students at CSULB attending two dif-ferent courses: ITA100A "Fundamentals of Italian for Spanish Speakers", (N = 30); and ITA 101A, a regular beginner course of Italian (N = 24).ITA100A was specifically designed for heritage speakers of Spanish (HSS), and therefore, only students who self-declared as Spanish speakers were allowed to enrol in this course.The curriculum for ITA100A incorporated course materials and teaching strategies, based on intercomprehension, aimed at utilizing the students' entire linguistic repertoires in acquiring Italian, with a particular focus on Spanish as a pivot language in the learning process (Donato, 2017;Travers, 2017).The use of intercomprehension as a methodological approach has gained backing through the creation of instructional materials and teacher preparation at CSULB (Donato et al., 2020).This support aims to cultivate cross-linguistic and cross-cultural parallels between Spanish, Italian, and English within the Italian for Spanish Speakers language learning program.This program spans three semesters and has garnered national recognition (Looney et al., 2021).
The ages of the CSULB students ranged from 19 to 24 years old, while the R3 students were pursuing their Master's degrees and were between 23 and 28 years old.Within the CSULB Spanish-speaking population, they could be classified as belonging to either the regular or limited cases of multilingualism categories (see section 1).These students demonstrated English dominance due to their education in English-medium schools, while varying degrees of proficiency in Spanish allowed for extensive language use and linguistic variation in Spanish.However, proficiency in Spanish was not tested.The R3 students were speakers of English as L2.No other language was self-declared.
All the mentors identified as women, whereas only three participants from CSULB identified as men.Every mentor was randomly assigned three or four mentees.Mentors and mentees met eight times during the semester through Zoom©.These 45-minute to one-hour meetings were designed to help the learners of Italian complete homework assignments while being exposed to as much target language input as possible.

Methodology
For the purpose of this study, we selected 10 dyads: five dyads in which the CSULB student was a HSS (ITA 100A); and five dyads with NHSS ( ITA101A).HSS declared that they only spoke English and Spanish, while NHSS did not speak any other language besides English.
In Table 1 the participants are presented, specifying the dyad, the group (HSS and NHSS) and the role (mentors and mentees).Note that in the table below more than one student was assigned to mentors AC, BB and GM.This is due to the higher number of mentees compared to mentors.For the purpose of this study, we did not consider this an obstacle in the observation of spontaneous linguistic exchange, since our aim was to analyse what was produced during their interactions.As the primary objective of the mentoring sessions was to provide an opportunity for mentees to practice Italian and receive assistance the mentors tended to focus more on linguistic forms than meaning during these sessions.To examine the NoM, while excluding the negotiation of form (which was not the primary focus of this study), we specifically selected the initial and final segments of each session.This decision was based on the observation that greetings and small talk occurred more frequently during these moments of the video sessions.To ensure comparability in terms of session length, we transcribed and coded the first and last 5 minutes of each session for every dyad.As a result, the data collected encompassed approximately 13 hours of recorded videoconferencing, which was subsequently transcribed and coded.
All sessions were recorded and transcribed employing a simplified version of the Jeffersonian transcription system (Jefferson, 1983).Subsequently, these transcriptions underwent coding using QSR NVivo 1 software.During the initial phase, a small subset of the data was collaboratively coded to ensure alignment with the coding criteria.This collaborative process of dialogue and deliberation among the researchers ultimately led to the finalization of the coding framework.In the second phase, each researcher independently completed the coding process for their designated data segments, punctuated by periodic meetings to ensure adherence to the predefined coding schema.Any uncertainties that arose were deliberated until a consensus was reached.To ensure data quality, a dedicated researcher cross-validated the data and formulated the queries used for the analysis.
To address our research questions, the researchers coded the NoM episodes, which included confirmation checks, clarification requests, and comprehension checks.Additionally, they coded the languages used in the data.Subsequently, we calculated the number of tokens (i.e., transcribed words) produced by every dyad (both HSS and NHSS), in every language.Therefore, every time the participants used a word or a phrase in one of the languages: English, Spanish, and Italian.The decision to quantify tokens rather than utterances was deliberate, as the latter holds a more indistinct definition, varying between a single word, a phrase, or an entire sentence.
Here we propose some examples of the NoM episodes coded.Example (1) is a clarification request, where the student (DTJ) struggles to understand the question and uses her English to continue the conversation with her mentor (AKS).In Example 2, we illustrate a confirmation check.This episode took place as the mentee (DCS) was recounting having gone to bed early the previous day.The mentor (AC) subsequently inquired whether she accurately grasped the mentee's account, seeking to verify her understanding.In Example (3), we present a comprehension check.This episode took place after the mentor asked the mentee if she had already had dinner (n.15).The exchange shows that the student (ZA) understood the question but needed confirmation and thus used English.Therefore, in order to negotiate meaning, participants use their English rather than producing in the TL, Italian.In Example (4), we show how languages were coded.In the excerpt, the mentor (GM) inquired of the mentee (LC) whether she had already com-menced recording the video session.This question in Italian triggered a NoM episode that spanned 16 turn takings.Within this transcript, the text coded as Italian is visually emphasized in grey, while the portions coded as English are displayed in black.

RESULTS AND DISCUSSION
In this section, we will present the results of the study, addressing the research questions.To delve into the RQ1 ("How do participants utilize their linguistic repertoire in mentoring sessions?), a detailed analysis was conducted on a total of 49,603 tokens, which are documented in Table 2.The distribution of these tokens across various languages utilized during the sessions is also presented for reference.Overall, English was the most used language, accounting for 58.07% of the total number of tokens.It was followed by Italian, the language studied by the mentees, which made up 41.83% of the tokens.Spanish was the least frequently used language, representing only 0.23% of the tokens.It's worth noting that although they did not self-declare Spanish as a language they spoke, some mentors still used it (14 tokens).Likewise, one of the mentees (RPA from Dyad4), who did not indicate her proficiency in French, occasionally used this language, albeit infrequently (6 tokens).The distribution of the language used according to the informant's role in the dyad (i.e., mentor or mentee) is radically different for mentors who used Italian and English with almost the same percentage: (52.98 vs 46.97%).Mentees, instead, used mostly English (71.47%), followed by Italian (26.57%).
The NoM episodes were also analysed.In the Table 3, the three types of NoM episodes, namely clarification requests, comprehension checks, and confirmation checks, are shown in relation to the number of tokens in each language.In the NoM episode, the language most frequently used was English (59.67%), followed by Italian (32.44%), and Spanish (7.89%).In their con-versations, mentors and mentees tended to use English more frequently for clarification requests (67.32%) and confirmations checks (50.06%).Comprehension checks took place more frequently in Italian (47.83%).
The second research question asked was whether there were differences between the HSS group and the NHSS group and if intercomprehension competence influenced their telecollaborative sessions in any way.In order to answer this question, we compared the number of tokens produced by every dyad in every language in each group.Table 4 provides an overview of the distribution of tokens used by mentors and mentees in the HSS group, revealing that mentors, on the whole, employed a greater number of tokens in comparison to their respective mentees.The languages employed by both mentors and mentees included English, Italian, and Spanish.Out of the total number of tokens Italian accounted for the highest proportion at 56.98%, followed by English at 41.37%, and lastly, Spanish at 1.62%.Mentees used English more frequently than their mentors (58.01%vs. 30.53%),while the mentors used Italian extensively when interacting with their mentees (69.38% vs. 37.95%), aligning with expectations due to their role in utilizing the TL.In the NHSS group, as shown in Table 5, mentors and mentees demonstrated a similar token count, with 13,832 tokens produced by mentors and 12,334 tokens by mentees.Notably, mentees exhibited a higher frequency of English usage compared to their counterparts (81.72%).Conversely, mentors used Italian more frequently (64.12%).Surprisingly, Spanish was also used within this group, albeit in minimal quantities (0.09%).It is important to note that this group belonged to the standard course, specifically catering to mentees who did not speak Spanish.The occurrence of Spanish can possibly be attributed to the sociolinguistic context, as these learners may have some level of familiarity with Spanish words due to their exposure to Spanish while living in Southern California.
Noteworthy differences are evident between the two groups, particularly regarding the usage of Italian and English.It is noteworthy to observe that Spanish, which one might assume to have a higher occurrence due to its presence in the linguistic repertoire of HSS group mentees, was not as frequently employed.Upon conducting an inter-group comparison, it becomes evident that the HSS group employed Italian significantly more often than the NHSS group (57.50% vs. 27.95%).This suggests that the intercomprehension competence of the HSS group might lead to an increased use of Italian and their capacity to smoothly transition to Italian right from the beginning.This may be facilitated by the HSS students' ability to leverage their heritage language, Spanish, while negotiating meaning.The HSS group consistently exhibited a considerably higher frequency of Italian usage in comparison to the NHSS group, which predominantly relied on English.
The following examples aim to show some of the differences between the two groups.Example (5), taken from a dyad in the HSS group, shows how the NoM and the usage of Spanish helped the mentee in the conversation.At first, she used English to ask for a word, but soon after she successfully attempted to say it in Italian.She also used Spanish while talking about clothes, possibly because she assumed that the word "ropa" (Eng."clothes") might be similar to the Italian word "vestiti".To delve deeper into the use of the languages, we conducted an analysis of the number of tokens generated by each type of NoM episode across different languages within the two participant groups.We have chosen not to differentiate between the number of tokens used by mentors and mentees, as the NoM is a complex conversational phenomenon, and we believe it is more informative to not present them separately.Table 6 presents the distribution of NoM episodes per language within the HSS group, while Table 7 illustrates the distribution within the NHSS group.Table 6 shows that the HSS group used mostly English (55.82%) to negotiate their meaning during communication breakdowns, especially during clarification requests and confirmation checks.However, Italian was also frequently used, particularly in the comprehension checks (50.58%).Spanish was rarely used overtly, but when used, it was used mostly for clarification requests.Table 7 shows that the NHSS group also used mostly English during the NoM (67.94%).Participants spoke mostly in English, especially during clarification requests and comprehension checks.Confirmation checks were produced mostly in Italian (56.74%).As we would expect, Spanish was rarely used by this group, but when used, it was mostly for clarification requests.
In comparison, the HSS produced more tokens when negotiating meaning (2,822) in relation to the NHSS (1,304).Both groups tended to negotiate frequently in English, but the HSS group produced more tokens in Italian (961) than their counterpart (371).Evidence of intercomprehension competence among the HSS group is demonstrated by the fact that Italian was frequently used by both mentors (69.38%) and mentees (37.95%).The situation was completely different in the NHSS group, in which Spanish was not part of the learner's linguistic repertoire.The most frequently used language in this group was English, by both mentors (64.12%) and mentees (81.72%) alike.
Upon examining the data presented in both Table 6 and Table 7, a notable distinction arises regarding the frequency of NoM episodes used by the two groups.As previously mentioned, the HSS group demonstrated a higher incidence of NoM episodes, indicating a greater need for meaning clarification and comprehension during their interactions.The greater need possibly reflects a greater desire and motivation for clarification, since they realized they were getting closer to comprehension, and thus activated a bridge, or pivot, for doing so.This observation suggests that the HSS group engaged in more frequent instances of language negotiation, subsequently creating additional opportunities for language learning within their interactions because Spanish as a pivot language brought them closer to successful comprehension.
Aligned with Grosjean's language mode hypothesis (2013), the observation indicated a consistent pattern where English predominantly served as the language for clarification requests.This suggests that participants leaned on their first language (L1), rather than their heritage language (Spanish in the case of the HSS group), for negotiating and making conversational adjustments.However, Italian was also frequently used.This suggests that Spanish does not emerge as a productive language in terms of participants using it extensively to solve communication problems.Instead, Spanish functions as a pivot language, serving as a supportive tool that aids in understanding Italian, the target language (TL).This analysis offers valuable insights that align with previous observations (e.g., Ard & Homburg, 1983;Kellerman & Sharwood Smith, 1986;Odlin, 1989) on the beneficial impact that a typologically similar language might have on the acquisition of an L3.

CONCLUSION AND PEDAGOGICAL IMPLICATIONS
In this paper, we conducted an exploratory study of language use within the context of telecollaborative mentoring.Our findings revealed that the mentees predominantly used English during the analysed conversations.This observation can be attributed to the fact that all the mentees were in their first semester of learning Italian.However, while non-heritage Spanish speakers (NHSSs) mainly relied on English, heritage Spanish speakers (HSSs) actively engaged with the target language by frequently incorporating Italian into their conversations.Although Spanish was seldom employed as a means of communication, it assumed a pivotal role in facilitating comprehension of Italian, rather than being extensively relied upon to address communication difficulties.We note, then, that the HSS group exhibited a superior ability to grasp and communicate effectively in Italian compared to the NHSS group.This ability was possibly a result of the adept use of both Spanish and English, a skill with noteworthy implications for Italian acquisition.
However, the linguistic behaviour within the HSS group did exhibit certain variations, also influenced to some extent by the mentors' tendency to consistently use English.Such variations could be mitigated through minimal pre-dyad training to raise tutors' awareness of their role in the telecollaboration activity, namely, by providing more opportunities for exposure and negotiation of input, as well as encouraging their mentees to produce output.By making mentors aware of the enhanced potential the Spanish language brings to HSS learners, they can take on a more dynamic role in the mentoring activity, fostering intercomprehension and generating interaction with their mentees.To achieve this, input-based tasks that promote Italian language comprehension from the HSS group should be designed and proposed.An input-based task is a type of focused task in which learners process input through listening or reading.While L2 production from learners is not mandatory, it is not prohibited either (Ellis, 2009).Such tasks are typically designed with two main purposes: a) engaging learners in understanding the input; and b) drawing learners' attention to specific linguistic features within a meaningful context.
Nevertheless, it is important to note that the impact of the course materials and the methodological approach employed in ITA100A, "Fundamentals of Italian for Spanish Speakers", cannot be definitively established at this time.While it is our contention that these instructional components exert a favourable influence, their efficacy warrants validation through subsequent research endeavours.
As emphasized, the intricate interplay between pivot language and intercomprehension within third language acquisition warrants attention in the realm of multilingual education.This domain presents an exciting avenue for both research and practical implementation within the paradidactic setting of telecollaboration, offering valuable insights for heritage language and third language acquisition.
To conclude, we would also like to acknowledge the unique situation of the Spanish heritage speakers in our study, noting that the inclusion of heritage speakers and their repertoires in L3 studies is a new frontier that scholars have only recently begun to explore.As migrations and mobilities bring increasing numbers of heritage speakers into contact zones where L3 acquisition is taking place, heritage speakers as multilingual subjects are destined to become an ever-increasing source of data on L3 acquisition.This study, which addressed the linguistic repertoire and dynamics of language use of HSS at CSULB, has been conducted within the emerging purview of the rich complexity heritage speakers bring to research on L3 acquisition.It is hoped that this work may serve as a model going forward.

Table 2
Languages used in the dataset -Tokens per language Pojavnice u svakom jeziku

Table 3
Number of tokens used per NoM episodes and language Broj pojavnica prema vrsti pregovaranja o značenju u svakom jeziku

Table 4
Number of tokens per language:HSS Group Broj pojavnica po jeziku: HSS

Table 6
Number of Tokens per NoM type produced in every language: HSS Group Broj pojavnica u pregovaranju o značenju u svakom jeziku: HSS

Table 7
Number of Tokens per NoM type produced in every language in NHSS Group Broj pojavnica u pregovaranju o značenju u svakom jeziku:NHSS