“Do I Have to Sign My Real Name?” Ethical and Methodological Challenges in Multilingual Research with Adult SLIFE Learning French as a Second Language

: In 2017, Quebec’s Auditor General reported several major issues regarding government-funded French as a second language (FSL) courses, especially those intended for adult students with limited or interrupted formal education (SLIFE). To this day, no ofﬁcial framework or program exists for this speciﬁc population, a situation that the government of Quebec wishes to resolve. Our research team was thereby mandated by the Ministry of Immigration to conduct a large-scale multilingual study with the objective of gaining a better understanding of the realities and needs of the various stakeholders involved in low-literate FSL classes. We met 42 teachers, 24 French learning center directors, and 10 pedagogical advisors in individual interviews; we also led 107 group interviews with SLIFE in 26 languages, allowing us to meet 464 adult SLIFE enrolled in low-literate FSL classes from 11 regions of the province of Quebec, most of them being refugees. This article reports on the decision-making process in which we engaged to overcome the ethical and methodological challenges we faced at various stages of the data collection with SLIFE participants: recruitment, informed consent, conﬁdentiality, interview protocol design, instrument piloting, data collection, and data translation and transcription. To make informed decisions, we had to turn to literature outside SLA (i.e., refugee research and translation/interpreting literature) for guidance. In this article we discuss the limitations and contributions of our research to guide researchers who will conduct studies with similar non-academic samples/populations.


Introduction
As second language education researchers, we are used to conducting research with students who are not native speakers of the target language; we must adapt data collection procedures and obtain informed consent to make sure that participants understand both the nature of their participation and what is expected from them in terms of data collection. Since many studies draw on students from academic settings who are more familiar with the research process, the focus of adaptations in such research might then be mostly related to language issues, rather than about the research procedure per se. However, when research is conducted with samples from different academic and cultural backgrounds, some issues largely related to research ethics and data collection procedure might become highly sensitive. In our research, conducted with adult students with limited or interrupted formal education (SLIFE, see DeCapua and Marshall 2011;Laberge et al. 2019) learning French as a second language (FSL), we faced constant ethical and methodological issues as the participants were not familiar with either the language or research procedure (for example, the general understanding of research, importance of providing consent, confidentiality of data, risk and benefits for participants, etc.). We had to think about solutions that would be the most relevant to conduct a study with this population of learners.
In this article, we will present the decision process we went through as we were planning and conducting the study. We begin with the background of the study (Section 2) before moving on to the ethical guidelines, from the literature, for carrying out research with refugee-background participants (Section 3). Next, we present our research approach (Section 4), some general observations and comments about our experience on the field (Section 5) and finally the procedure we developed for data translation and transcription (Section 6). It should be noted that no empirical data are presented, given that the aim of the article is to provide ethical and methodological insights regarding multilingual research with SLIFE.

Background of the Study
Our research is part of a large-scale government project to improve French teaching services in Quebec, a French-speaking province of Canada. In the interest of providing a better framework for FSL teaching to adult students with limited or interrupted formal education (SLIFE), the responsible authorities 1 (see note) have mandated us to gain a better understanding of the realities of the stakeholders involved in French teaching/learning for these students. This picture was necessary in order to allow us to formulate recommendations for the development and implementation of a government framework program for teaching French to SLIFE.
To collect data that would allow us to shed light on the reality of French courses for SLIFE, we chose to meet with students and different stakeholders (teachers, pedagogical advisors, and center directors). The SLIFE we met in our study were adult immigrants enrolled in government-funded French courses. They shared the characteristic of each having reported less than nine years of schooling; on this basis, they were placed in courses specially designed for low-literacy learners (students with more than nine years of schooling are placed in regular classes and do not constitute our target population). In the French language classes for SLIFE, there is a great deal of heterogeneity on several levels: country of origin, languages understood and spoken, age, personal learning goals, language of origin (LO), literacy skills, French skills, school experience, etc. (Fortier et al. 2020).
When starting to reflect on the data collection, we first consulted research conducted with this specific population in our domain of expertise, second language acquisition and teaching research. Two main observations were made: SLIFE are largely underrepresented in second language acquisition and teaching literature (Andringa and Godfroid 2020) and, in studies involving data collection with SLIFE participants (i.e., Kurvers et al. 2015;Strube 2010), ethical and methodological issues are not thoroughly presented. Facing this literature gap and wishing to engage in a process guided by awareness of the participants' realities and by faithfulness towards them as individuals (Calafato 2022), we consulted refugee research as it most captured the particularities of our study population and could highlight important issues not to be missed (Mackenzie et al. 2007). Although our participants were not always refugees (some of them having come to Canada through other immigration paths), this field of research seemed to be the most closely aligned with the realities of our sample. We did find invaluable insights on how to conduct our research and believed that our reflection and development process could benefit other applied linguistics researchers working with language students from this specific migration context.

Ethical Guidelines for Carrying Out Research with Refugee-Background Participants-Stepping Out of SLA
The literature review highlights some of the relevant ethical principles in research, with respect to refugees and populations in situations of forced migration, that we consulted to guide our decisions. These will focus on communication with participants and recruitment (Section 3.1), obtaining informed consent (Section 3.2), as well as the risks and benefits of participating in social science research (Section 3.3).

Communication and Recruitment
The issue of communication between researchers and participants is an all-encompassing one in any research project involving human beings, in that it is critical from the recruitment of participants through to the analysis of the data collected. However, when researchers and participants do not share a common language, communication can be a major challenge. In these situations, the presence of a bilingual third party, often an interpreter or a local research assistant, is necessary to act as a linguistic bridge between researchers and participants (Wallin and Ahlström 2006). In their systematic literature review on researcherinterpreter collaboration in cross-cultural research, Shimpuku and Norr (2012) identify three forms that researcher-interpreter collaboration can take: (1) invisible assistance, which can be found in articles where very few details are given about the interpreter, his or her training, the arrangement of the collaboration, etc., and which presupposes that the interpreter's role is one of "literal translation" (which is, incidentally, widely criticized in translation studies; see Durieux 2007;Meschonnic 1999); (2) independent fieldwork, which involves the interpreter taking a leading role in the data collection by meeting participants without the presence of the researcher, which avoids interrupting the rhythm of a live oral conversation translation and may cause interlocutors to abbreviate their words to help maintain a flowing conversation (Gibb and Good 2014;Obijiofor et al. 2018) but may lead to a feeling of loss of control (anticipated or real) by the researcher; (3) integrated collaboration, which implies that interpreters are present in more than one phase of the project, for example, by participating in recruitment and/or translation and analysis of results, allowing "full collaboration between the researcher and interpreters so that the researcher's conceptual insights and the interpreters' cultural expertise can be incorporated" (Shimpuku and Norr 2012), but such an approach also poses feasibility challenges as well as being time consuming. Each of these forms has limitations that affect the collected data (Kosny et al. 2014), and researchers should take these into account when establishing their methodology.
The issue of recruitment also poses its own set of challenges. When recruitment is envisioned in communities, many studies among hard-to-reach populations (e.g., refugees in situ) use a contact person from the target community to recruit participants from their community (see Ellis et al. 2007;Obijiofor et al. 2018). These participants are in turn asked to identify other potential participants and so on. This approach, called snowball sampling, has several limitations: participants are likely to have some similar characteristics, since they are likely to share at least the characteristic of being part of the same community (religious, cultural, etc.) and not everyone has an equal opportunity to participate in the research, as a connection with the contact person or other potential participants is a minimum requirement (Ellis et al. 2007;Jacobsen and Landau 2003).
When recruiting in schools, access to participants is facilitated, but being in a closed environment, such as a classroom, can lead to participants feeling pressured to participate because they are seemingly "captive" (Block et al. 2013). These mechanisms are important to consider in recruitment, as they may influence the sincere and genuine willingness of potential participants to engage with the research.

Informed Consent
The issue of communication between researchers and participants is intimately linked to the issue of obtaining informed consent from participants. Indeed, obtaining this consent is dependent on a clear understanding of the project and the nature of the participation. It is therefore critical to have specific consent measures in place that meet institutional requirements (e.g., the researchers' home universities, the participants' academic institutions, and the funding agencies) while protecting participants. In research with refugees and those in situations of forced migration, the use of the consent form, which is widely used in social sciences, is subject to debate; it assumes that participants have access to a form in a language they understand, that their literacy level allows them to assimilate the information on the form, that they understand the nature of the participation they are being asked to make, and that they are familiar with the broader concept of informed consent (Mackenzie et al. 2007). Some authors suggest alternative ways of obtaining consent from participants, including providing access to information orally in a language understood by the participants as well as obtaining consent orally (Clark-Kazak 2017; Ellis et al. 2007;Hugman et al. 2011;Thapliyal and Baker 2018). These ways of doing things, which rely primarily on the oral exchange of information, may be more valuable or trustworthy to people from cultures with oral traditions (Clark-Kazak 2017; Ellis et al. 2007;Hugman et al. 2011). They may also seem less intimidating than many pages of text (Ellis et al. 2007) and may be particularly recommended where participants have limited literacy (Hugman et al. 2011).
Finally, when potential participants come from countries where authorities use coercive force to obtain the cooperation of citizens, the extent to which participants are willing to enroll on a completely voluntary basis without pressure or fear of reprisal must be considered (Block et al. 2013;Ellis et al. 2007;Hugman et al. 2011). Conversely, for a variety of reasons, including past abuse, some participants may actually have little trust in people they do not know (Kabranian-Melkonian 2015; Mackenzie et al. 2007;Obijiofor et al. 2018), thus dissuading them from volunteering to participate in the research. However, the question remains whether this issue of trust may be overcome by the trust that participants have in the community contact person, considering that if the community contact person feels it is acceptable to participate in the research, then so might other community members.

Risks and Benefits of Research Participation
The issue of risks and benefits poses a profound challenge in research with refugees and forced migrants, who are often referred to as "vulnerable". First, the balance between participants' protection and autonomy is fragile; in trying to protect participants, researchers may unconsciously undermine their autonomy and agency by supposing to know what is best for them (Daley 2015;Mackenzie et al. 2007). Second, while it is true that some research, by virtue of its context and methodology, may present significant risks to participants (e.g., in situ research with refugee populations in transit), having refugee status does not necessarily equate to being "vulnerable". According to Perry (2011, p. 909), "vulnerability is a product of an interaction between participants, context and research design rather than an inherent characteristic [of the participants]". In other words, refugees are likely to be less vulnerable when they are in a familiar context (e.g., in a community organization or classroom or in their host country) than when they are in a refugee camp or in transit (Perry 2011).
In the context of research on forced migration or refugee populations, it is important for Western researchers to be sensitive to cultural issues that may be unknown to them. For example, in one-on-one interviews with an interpreter, it may be risky for some participants to reveal information if the interpreter is from an ethnocultural group with whom relations are strained (Jacobsen and Landau 2003;Kabranian-Melkonian 2015). The characteristics of ethnocultural groups (e.g., dress, hairstyle, accent, etc.) are not always apparent to Western researchers, so it is important to be sensitive to them and to seek advice from cultural representatives at an early stage (Jacobsen and Landau 2003). One way to do this is to pair participants with interpreters who speak the same language but do not necessarily belong to the same ethnocultural group (see Jacobsen and Landau 2003, for examples of pairing). In group interviews, this issue may manifest itself not only in the interpreter-participant relationship, but also in the relationships between participants; trust within the group can be undermined, for example, if participants have political "enemy" affiliations (Jacobsen and Landau 2003;Kabranian-Melkonian 2015).
Finally, in research involving refugees or forced migrants with precarious living conditions, there is a possibility that participants "[ . . . ] may have unrealistic expectations of the benefits of the research, believing that researchers may have the power to influence [their status]" (Mackenzie et al. 2007). This issue, also raised by Ellis et al. (2007), Gillam (2013, and Clark-Kazak (2017), reinforces the importance of well-informed participants and clear measures to ensure smooth communication.
This overview of the scientific literature provides avenues for reflection and essential steps for the implementation of empirical studies with populations that share certain linguistic, cultural, political, and migratory characteristics. These lines of inquiry were used as a guide in the process of preparing and implementing our large-scale needs assessment study, the details of which are presented below.

Research Approach
Our data collection process ranged from the planning stages (Winter), to piloting (Spring), and finally to the main collection (October 2019 to March 2020). In what follows, we will present the unfolding of our approach in relation to the following steps: the conception and implementation of the research instruments (Section 4.1), the recruitment and training of the research team (Section 4.2), the process of recruiting participants and obtaining informed consent (Section 4.3), and the progression of group interviews (Section 4.4), by highlighting our decision-making process, i.e., by reporting on our observations and the means we implemented to ensure the best possible ethical and methodological approach.
However, given that the focus of the article-the decision-making process itselfhappened mostly before the piloting phase, the presentation of our research approach does not follow a chronological path from the piloting phase to the main data collection. In fact, only few insights of the pilot study, which consisted in piloting the group interviews with 24 participants, from two different institutions, in four languages (Arabic, n = 11; Dari, n = 7; Spanish, n = 4; French, n = 2), led to changes in the main data collection. These insights are presented when relevant.

Conception and Administration of Research Instruments
To better understand the reality of SLIFE in French courses, and to properly identify the needs they have and the challenges they face, we opted, as mentioned, for group interviews with the students, as "a way to gather many opinions from individuals within a group setting" (Gibb 2012, p. 186). By proceeding in this way rather than through individual interviews, we wanted to have the opportunity to meet a larger number of students. In addition, group interviews have the advantage of being able to encourage discussion, as students can chime in to expand on a peer's responses or to add information (Baribeau and Germain 2010). The interview template we developed included the following themes: (1) languages and school (e.g., what is the first language that you learned? Have you ever been to school before taking French classes here?); (2) French (e.g., during the last month alone, did you need to use French in the following situations?-pictures of different situations were presented-in which situations is it easiest/hardest to communicate in French?); (3) French courses (e.g., what do you like the most in your French courses? What is the easiest/hardest?); (4) your comments (e.g., could French courses be even better? How?). The group interview instructions specified that not every question needed to be answered and that discussion was encouraged when appropriate. It should also be noted that in the original French version of the template, great care was taken in the wording of the questions to make them accessible and avoid complex structures (Perry 2011).
We then quickly determined, prior to the piloting phase, that group interviews would take place in the students' LO because their oral French skills were unknown to us (note, however, that some students did participate in a French interview, particularly when group formation in their LO was not possible and their oral French proficiency level allowed it). We therefore began to think about the type of interpreting that would best serve the interests of the participants and the project. It was unanimously agreed by the research team that direct interpretation, where the interpreter translates the researcher's questions from French to the LO and then from the LO to French for the participants' responses, would significantly increase the length of the group interviews, in addition to increasing the risk of losing potentially important information (Björk Brämberg and Dahlberg 2013; Gibb and Good 2014;Obijiofor et al. 2018). We therefore ruled out direct interpreting and also chose not to hire professional interpreters, except as a last resort for the rarer languages. Instead, we decided to explore the avenue of bilingual/bicultural assistants (Obijiofor et al. 2018), also known as local researchers (Jacobsen and Landau 2003) or community researchers (Berman and Tyyskä 2011), by recruiting and hiring university students, community workers, or trusted acquaintances. As genuine research partners, 39 bilingual assistants worked with us to conduct group interviews in 26 languages. These assistants were asked to conduct the group interviews on their own, using the interview templates translated into these languages by certified translators. Since the interviews were recorded, it was possible to back translate the audio files into French.
Once we decided on the data collection tool and methods, we were able to plan our visits to the learning centers across the province, but first we had to gather and train our research team.

Training of the Research Team
In preparation for the group interviews, the bilingual assistants received 2.5 h of training, mostly in person, which covered the following topics: presentation of the project, research team, and objectives; the role of the assistant in group interviews; some general characteristics about SLIFE, including psychosocial characteristics and others related to literacy levels; confidentiality, anonymity, and the signing of a confidentiality agreement; the progression of the interviews and the presentation of the interview template. The interview template was presented to them in French and in their own language, and time was allowed for clarification, if necessary. This procedure of checking the template as a group made it possible to compensate in part for the lack of back translation, which was set aside due to a lack of resources. Because the assistants were trained in multilingual groups, we were able to navigate certain terms or turns of phrase to maximize our chances of having meaning-equivalent versions across languages (Bloch 1999).
During this training, consent forms were also presented in order to familiarize assistants with the various documents they would be handling (a simplified form for each respective LO, a simplified form in French, and a long form in French; see Section 4.3). Procedures for informing, recruiting, obtaining consent, and facilitating group interviews were also presented to the assistants, who were invited to discuss with us any elements that they might find culturally sensitive or problematic. For example, one assistant told us that in her culture it is frowned upon to ask someone's age, while others told us that it is a question like any other and that there is no awkwardness in asking it. Therefore, we offered the assistants flexibility, in the sense that they were in the best position to judge their comfort level and that of the participants. It should be noted that some assistants were contacted at the last minute and their training was cut short, which was reflected in the group interviews (see Section 5). Finally, the assistants were responsible for leading the interviews following the template in the LO of the interview. They were solely responsible for conducting the group interviews and managing the time.

Recruitment and Consent
To inform potential participants about our project in order to recruit them for group interviews, we had to implement a multilingual approach. It was essential that participants understood the project, its objectives, and the nature of the participation that was being proposed, and that they were able to ask questions freely in their LO. We therefore wrote a short introduction and recruitment text, which was translated into the 26 languages in our sample (see Appendix A for the English version). The text, written in a simple and accessible manner, introduced the project and its objectives, mentioned that students were invited to participate in a group interview in their LO about their experience in French classes, emphasized that they would be asked to give their consent orally, that a form would be signed on their behalf by the assistant, and that anonymity and confidentiality would be preserved in the analysis and dissemination of results. It was also stated that the discussions were digitally recorded and that only the research team had access to the audio files. Finally, it was specified that participation was free and voluntary and that refusal to participate would not entail any consequences or prejudice. A few examples of questions found in the interview template were finally presented to the students. In this way, each assistant had the same text, in his or her own language, and had to present it orally to the students.
For the information and consent process piloting, we were also accompanied by bilingual assistants. However, rather than relying on a text written in their language, the assistants repeated in the LO what one of the research team members said in French. The presentation was given in large groups, and when it was over, the students were invited to ask questions, which the assistants collected and which the researcher translated into French. This meant that there were several exchanges between the students and the researcher, always going through the assistant, which was possible because the piloting included only three languages (the most common in our sample, Arabic, Dari, and Spanish) as well as French. This piloting phase taught us that the presentation and recruitment stages of the project needed more attention. The presentation proved to be too short and insufficiently detailed, which had an impact on the question period following the project presentation taking much longer than expected. In particular, interested participants needed reassurance about anonymity and confidentiality. It seemed difficult for them to understand that participation and the dissemination of results were anonymous but that their name should still appear on the consent form. Some students also mistakenly understood that only those who had faced difficulty in their French classes were invited to participate. For instance, one participant told the assistant that his participation was not useful to the project because he felt everything was fine with his learning of French. Many also asked for examples of the questions that would be asked before making a decision about their participation.
In light of these challenges, we improved the information and recruitment process by providing more details on the information given during the presentation and by including sample questions. We also decided to proceed in a more efficient way to present the project to potential participants. Depending on the size of the institution where we were hosted, there were two scenarios: 1.
In the smaller French teaching centers, the presentation was made in all (or almost all) of the languages present, in front of all the students in the class. In order to emphasize the students' learning in French, the presentation was always made in French, then in a first LO, in a second LO, and so on. This procedure had the advantage of requiring less organization and planning, but, on the other hand, some students were left out because they were the sole speakers of their language and no assistant in that language accompanied us. Students interested in participating in the group interviews were then invited to follow us to join the students of their language recruited in the other class groups, and students not interested could remain in class. The classroom teacher had agreed to adjust his or her lesson plan for the students who remained in class.

2.
In the larger French teaching centers, teachers were asked to direct students to the rooms assigned to their language. In this way, all students sharing the same language of origin were grouped together, and the assistant presented the project to all the students at the center who spoke that language at the same time. Students who were interested in participating in the group interviews were invited to stay on, while those who were not interested were invited to return to class. In cases where a large number of students of the same language showed interest, several rounds of discussions were planned, in succession, trying as much as possible to keep the number of students per group to no more than 10. This procedure required a great deal of planning but saved considerable time since all the presentations in the different languages were conducted simultaneously. The classroom teacher had agreed to adapt his or her lesson plan for the remaining students.
The presentation of the project and the recruitment of participants were carried out in 1 to 11 languages per institution, with an average of 3.89 languages per institution. However, even though the presentation of the project was made in the participants' LO and they could ask all the questions they wanted, misunderstandings persisted, such that at the consent stage, some participants still did not understand the nature of the project. Indeed, some students agreed to participate without necessarily understanding the nature of the participation that was required of them. For example, one person thought there would be food distribution and decided to return to the classroom when she realized that this was not the case. Another person, no doubt misled by the word recherche-'research', thought we were there to help her look for gloves she had lost in the school. The latter also chose to return to class without participating when she understood the nature of the participation that was required of her. Some misunderstandings also persisted to the point where a few participants from the same school completed the group interviews and admitted afterwards that they did not understand the consent forms (see Section 5). Despite these few glitches, our procedure allowed us to avoid recruiting by snowballing techniques (Jacobsen and Landau 2003), so that almost every student had an equal chance to participate, not just friends of friends (Obijiofor et al. 2018).
Overall, we recruited students in 19 schools, located in nine regions of Quebec. A total of 464 students participated in 107 group interviews in 26 different languages (see Table 1). Students were offered the option of participating in the group interview either in their LO, in a language in which they felt comfortable participating (e.g., a Tuareg speaker joined a group interview conducted in Arabic; moreover, as these cases were marginal, we consider in this article the languages of origin and the languages of the group interview to be the same), or in French (for the students for whom an LO interview was not available and who otherwise felt open to participating in French). The groups for the interviews included between 1 and 11 students, with an average of 4.34 students per group. Most participants were between the ages of 36 and 55 (46%); 56% of them were women and 44% were men. After being recruited and guided towards the bilingual assistant responsible for their group, the participants first had to consent to their participation in the study. From the very beginning of our research team's reflections, this issue of obtaining informed consent from the participants has been given considerable thought. We had to meet the requirements of the institutional ethics committee and of the ministries concerned, while applying the recommendations of the literature in relation to ethics and methodologies in refugee studies. We then found a formulation that was likely to meet these criteria. We decided to obtain oral consent in the LO from the participants for both piloting and main data collection, with few changes between the two phases. It was important, however, that a consent form be signed by the assistant and a witness (Hugman et al. 2011) who was a member of the research team, as evidence of the participant's consent. The participant's first and last name also appeared on the form. In addition, since many students have some reading skills in their LO, we felt it was essential to give them access to the written information on the form in those LO (Thapliyal and Baker 2018). We therefore created a simplified version of the information and consent form, which we had translated into Arabic, Dari, and Spanish for the pilot phase and into the remaining 22 languages of our sample during the main data collection (see Appendix B for the English version). Thus, for each student, three forms were required: (1) the simplified form, in their LO, which participants kept with them; (2) the short form, in French, identified with the participant's name, signed by the assistant and the researcher, and retained by the researcher; (3) the long form (including all the information required by the institutional ethics committee), in French, which the participants kept with them, with the idea of being able to refer to it if necessary with the support of someone close to them (family member, teacher, etc.) with the necessary French language skills. On the short, simplified forms (in the LO and French), space was provided for students to sign if they were able and wanted to. In this way, participants were not deprived of their agency, but could also rely on the process of providing oral consent and on the assistant when needed.
When the interview groups were small (less than five participants), problems were less likely to occur. On the other hand, some groups had as many as 11 participants, partly because it was difficult to anticipate the number of students who spoke the same language (student lists were not always up to date) who were present on the day of our visit and, above all, who were interested in participating. For this reason, and to avoid moving assistants unnecessarily, some interviews with larger groups were conducted, resulting in some confusion for both students and assistants with regard to the multiple consent forms. Nevertheless, in general, the presence of a member of the research team helped to resolve any problems that arose.
Several students who seemed interested in participating were dissuaded by the consent process. The manifestations of the students' distrust were diverse: some refused to give their real name and asked if they could sign with a pseudonym; others were not comfortable with being registered; some students wanted to participate in the interview individually rather than as a group; some women refused to participate in group interviews with men; finally, some students said they needed their father's or husband's agreement (attending the same center, but not necessarily in a class participating in the project) to participate. All these students were reassured that they were under no obligation to participate and could return to the classroom, but that if they wished to participate in the group interviews, they would have to agree to do so according to the established procedures. Most often, wary students preferred to leave and return to class, suggesting that the consent process was well understood and allowed students to make an informed choice. It should be noted that we did not observe any regular or systematic patterns related to ethnocultural groups, gender, age, or the LO among participants who chose to withdraw. We therefore believe that these withdrawals had little impact on our data and its representativeness with regard to the various ethnocultural groups and LO.

Group Interviews
As mentioned in Section 4.1, it was quickly determined that the group interviews would be conducted in the LO and that they would be conducted entirely by bilingual assistants. For the group interviews, assistants were asked to rely on the template developed in French and translated into the other 25 languages by certified translators (Bloch 1999;Jacobsen and Landau 2003) where possible (otherwise, translations were completed by community organizations that worked with reliable bilingual people). After the presentation of the project and before the group interviews began, oral consent was obtained and consent forms were signed by the assistant and a member of the research team; the audio recording could then begin and the research team member would withdraw to observe, take notes, and answer questions from the assistant as needed. During the group interviews, questions were asked one after the other and participants were told to answer freely, with the result being that some questions were answered by all the participants while some others were not. When the group interviews were completed, members of the research team circulated to collect the consent forms. This procedure was the same for the piloting of our tool and for the main data collection. However, following the group interviews in the piloting phase, we met with the assistants so that they could provide us with feedback on the interview process. The main comments made by the assistants were related to the following issues: 1.
Assistants' instructions to participants for group interviews: ill-defined. One assistant reported that some participants became confused during the group interview. One participant reportedly spoke about things that were considered irrelevant by others, who told them that their comments were never relevant and always drifted away from the topic of conversation. The participant was offended by this and chose to leave the group interview.

2.
Problematic interview question: discomfort of the participants. One of the questions on the interview template asked about teacher and classmate appreciation ("How do you find (a) the staff at the establishment, (b) the teaching staff, and (c) the other students?"). This question created discomfort as it invited participants to express their opinion about other students in the class, perhaps in front of them. The group interview format did not lend itself well to a judgmental question that could relate to other group members.
In light of these comments, we wrote down the instructions that the assistants were to give to the participants (1-everyone is entitled to answer the questions; 2-no one is forced to answer the questions; 3-all opinions are important; 4-everyone has the right to speak; 5-discussion occurs with the assistant and not between participants). We reminded the assistants of the instructions during the training, and we inserted a printed copy of these instructions into each assistant's folder, which contained the interview template and the consent forms. Finally, we removed from the interview template the question about the appreciation of other students in the class.

General Observations and Comments on Data Collection
Several general observations emerge from the data collection approach, both in terms of its strengths and limitations. First, the choice to collaborate with bilingual assistants (n = 39) constitutes one of the strengths of our methodology, as it allowed for a long-term collaboration (Wallin and Ahlström 2006;. Resch and Enzenhofer 2018). Some assistants conducted as many as eight group interviews over the five months of data collection. In addition, the collaboration was generally exceptional, and we developed excellent relationships with the assistants. Several of them have expressed an interest in working together again if needed. They were instrumental in the success of our data collection. Connections were also made between the assistants and the participants; although the assistants may have represented the educated elite, thereby creating a power imbalance in their favor (Ellis et al. 2007), it appears that goodwill, empathy, and caring prevailed in most group interviews.
Moreover, by choosing integrated collaboration (Shimpuku and Norr 2012), we were able to avoid important limitations related to live interpreting. For example, research biases can be induced, even unintentionally, by the person interpreting, because they have to make many quick decisions (Björk Brämberg and Dahlberg 2013;Jacobsen and Landau 2003;Kabranian-Melkonian 2015). Moreover, the pace of conversation may seem unnatural because sentences or ideas are interrupted by live oral translation, and speakers may tend to abbreviate their words to help maintain a fluid conversation (Gibb and Good 2014). With an emphasis on integrated collaboration, we placed our full trust in the assistants, who were involved from recruitment to data translation (see Section 6).
The main data collection technique, enhanced by the adjustments made following the piloting phase, still revealed some limitations of our approach, mainly due to time or resource constraints. For example, the fact that we travelled to nine regions of Quebec led to days of data collection where the research team and the assistants were split into three schools at the same time. It was also complicated to obtain the list of languages spoken by the students well beforehand to allow us to train the assistants. This was especially the case for rarer languages, for example, Tigrinya, Tagalog, or Khmer, where we felt we were in a rush to train the assistants. Hence, despite the precautions taken and the efforts invested in assistant recruitment and training, we encountered a few glitches. In one case, the recruitment of a rare language assistant in Quebec was laborious and led to the problematic hiring of a person who did not fully understand the project. Despite the training (conducted by telephone), this assistant seemed unclear about her role. She seemed stressed when she realized that she would have to approach a group to recruit students who spoke her language. She was reassured by the researchers in particular, but also by the other assistants present that day. Another isolated case, where an assistant recommended by a community center behaved arrogantly towards the research team, made it difficult to recruit students. The students who spoke the language in question did not seem at ease with this assistant. They did agree to participate in the group interview in that language, but the community center was alerted to the unprofessional attitude of the assistant. Finally, the major problem that we discovered when reading the translated French transcripts of the group interviews (see Section 6 below) was that one assistant, who conducted five interviews, was giving her opinions to the students. For example, if a student talked about a problem he or she was having in daily life because he or she did not speak French, this assistant went so far as to say: "You need to go find someone you know who can help you!". The assistants had, however, been told that they were not to give their opinion, whether positive or negative. This may have made some participants uncomfortable and thus violated the ethical requirements we had established.
In this sense, the training provided to assistants could have benefited from certain improvements: we observed that the assistants who had been trained in person for 2.5 h and in groups with assistants of other languages had a better understanding of the project, its objectives, and the role they would play with the SLIFE. We also noticed that the assistants with whom we had the chance to work for several weeks seemed more confident and showed a better understanding of the project than those with whom we only worked once. However, even among the more experienced assistants, closer attention should have been paid to the fact that their role involved being as neutral and detached as possible, and that they should not offer personal comments or opinions. To this end, it would probably have been beneficial to test their interviewing skills (Kosny et al. 2014), which we were unable to do due to time constraints. The training should have also included resources to which assistants could direct participants if sensitive information was disclosed during the group interviews (Clark-Kazak 2017). Although the interview template did not include discussions of sensitive topics, it would have been prudent to provide tools for assistants to respond appropriately if sensitive information was shared (Kohlenberger et al. 2017) and to reassure and guide participants as necessary.
Nevertheless, communication between the various parties (research team and assistants, assistants and students, research team and school correspondents) generally went well throughout the implementation of the data collection, despite the organizational challenges inherent in our multilingual process. However, one particular case should be noted where communication was more difficult. Indeed, after our visit to one of the participating schools, we were told by some teachers and our contact person at the center that several teachers did not appreciate our visit. In particular, they said that after participating in the group interview in their LO, a few students returned to class with documents they did not understand (the consent forms), and that they seemed confused in general about the project and its goals. We took this situation very seriously and all assistants on subsequent collection days were again made aware that it was essential that participants have a clear understanding of their participation before starting the group interviews. As this situation seemed to be related to assistants who were participating in the project for the first time, the importance of providing comprehensive and quality training was reiterated.
Finally, the scope of the data collection and the number of participants met (n = 464) made certain principles of risk mitigation difficult to apply, for example, the "[negotiation of] reciprocal benefit [between the researcher and the participants] in the context of an iterative consent process", during which participants are asked to express their expectations, their motivations for participating in the research, and their concerns (Mackenzie et al. 2007, p. 311). It was impossible to allow time and space for students to express their expectations and motivations because we met with a great number of students and because we had to perform the group interviews according to the schools' schedules. It is also possible that participation in a group with other students from the same center prevented participants from sharing some information. We did not hear anything about this, but the inability to share information in confidence remains an inherent limitation of the group interview (Jacobsen and Landau 2003).
It goes without saying that this data collection gave us access to a significant amount of data to analyze, a step that also led us to various questions addressed in the following section.

Data Translation and Transcription
The 107 interviews conducted with students enrolled in French courses for SLIFE totaled just over 75 h of recording time in 26 different languages. While the French interviews were transcribed by Francophone research assistants, the remaining 99 interviews presented greater transcription and translation challenges, echoing the issues raised in particular by Ganassin and Holmes (2013) regarding the complexity of multilingual data translation and transcription in their study conducted with refugee women. Initially, we wanted to have the interviews transcribed into the LO by student speakers of these languages, and then translated into French by certified translators, but evident time and budget constraints led us to find alternative solutions, with the limitations that these entail. We therefore proceeded with large-scale recruitment of student speakers of the various LOs, whom we commissioned to translate/transcribe the interviews from the oral LO into written French. We worked with 23 people to translate/transcribe the audio files. Some of these people were also part of our team of assistants, but great care was taken in cross-referencing audio files, so only a few rare language speakers translated/transcribed interviews they had conducted themselves. This cross-referencing of files was put in place to avoid the temptation for translators/transcribers to change the translations/transcripts to appear in a better light. Finally, to support the translation/transcription work, several documents were sent to the selected individuals: a procedure for securely exchanging audio and transcribed files with a file transfer site and general translation/transcription guidelines, including transcription conventions and a transcription template (Kosny et al. 2014). A research assistant with several years of experience in the field of translation was responsible for coordinating the translations/transcriptions. It should be noted, however, that some group interviews were not translated/transcribed, particularly when we could not find a speaker of these languages (Sango, n = 3 and Tigrinya, n = 1) who was available and sufficiently fluent in written French, or in the case of the interviews conducted in French. Because of the low level of the participants' oral competence in French, the thread of the conversation almost exclusively depended on paraverbal elements (e.g., gestures, facial expressions, etc.) and the audio files (n = 4) were virtually unintelligible without these elements. Thus, it was practically impossible to access the true meaning of what participants shared. Interviews were also lost due to technical problems with the recording devices (n = 2). From this point, data analysis was conducted.

Conclusions
The purpose of this article was to share the reflections we came across-and the decisions we made-while setting up a large-scale data collection with 464 SLIFE in 26 lan-guages. At the beginning of our research project in 2019, we lacked information on which to base our data collection judiciously; to our knowledge, no multilingual methodological approach conducted with SLIFE in an educational context had yet been published. We therefore had to be creative while demonstrating ethical, methodological, and scientific rigor in order to achieve our research objective, which, to reiterate, was to gain a better understanding of the reality of the stakeholders involved in the teaching-learning of French to these students in Quebec (Canada). We believe that the second language research community can benefit from this experience and gain awareness of specific issues related to working with students coming from various immigration paths, and to collecting data with groups of individuals coming from a large variety of linguistic and literacy backgrounds.
More precisely, the one aspect of our research process that deserves the most attention is our close collaboration with the bilingual assistants, which proved to be an asset for the project. By relying on an integrated collaboration model (Shimpuku and Norr 2012), bilingual assistants were involved in different phases of the project (i.e., recruitment, interviews, data transcription) and could therefore incorporate their cultural expertise to the conceptual and methodological expertise of the research team. Although cultural tensions can persist within groups of people sharing the same language, we believe that letting the bilingual assistants in charge of their group interviews is a way to offer a culturally reassuring environment for the participants to express themselves freely. Yet, this collaboration with bilingual assistants did not go without challenges, these being mostly related to their training. Indeed, in cases when training had to be cut short, mostly because we were only informed at the last moment of the languages spoken by the students in the school to be visited, it had negative consequences during the interviews (see Section 5). Assuring quality training of these bilingual assistants is thus highly important for collaboration with these key members to work at its best.
Another problem we faced, but to which we do not have an easy solution, is that a few students were left behind as soon as the presentations of the project started in their school, when no bilingual assistants could present the study in their LO. These students would do their best to understand the presentation of the project given in French, and they would stay in class afterwards while many of their classmates left the classroom to take part in group interviews in their LO. Given the choices that we had to make, it is unclear what could have been done differently, but we nevertheless believe that it remains problematic to leave some students behind.
Finally, many second language acquisition and teaching researchers call for their colleagues to step outside of their comfort zone and to gain experience carrying out research with learners who currently remain invisible in research (Andringa and Godfroid 2020). We can thus anticipate that the challenges we faced working with SLIFE will be met by others, and hopefully what we reported in this paper can guide their reflections. Ultimately, not only research methods and ethics with understudied populations will be enriched but, more generally, second language acquisition and teaching research will engage in a path toward being less biased and more inclusive (Andringa and Godfroid 2020).  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to the terms of the research contract.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Appendix A Presentation and Recruitment Text
Hello, We are a team from the Université du Québec à Montréal and we are working on a research project with an objective to portray French courses. We come to your centre to meet you and also to meet with the teachers and the director.
This morning, we are here to present the project and answer your questions and this afternoon we would like to have group discussions in your native language, to talk about French courses. We want to know your experience and better understand your needs. All opinions are important, even if the courses are going very well for you, or even if the courses are not going as well or they do not meet your expectations. No one is obliged to participate: Your participation has no influence on your grades in the French course or on your immigration status. You do not make money by participating either: Your participation is voluntary.
During the discussions, we will record your answers with a small device without a camera, with a microphone only. We do not ask for your names or any information that allows for your identification. In addition, only members of the research team will have access to the records: Not your teacher, neither the ministry nor other students. Your participation is therefore anonymous and confidential.
The discussion groups last for about an hour. Three to five people who speak the same language as you will participate in the discussion. First, you will be asked to sign a form that indicates that you understand the project and want to participate. If you need help signing, I can help you. Then, I will ask questions about French in your life and about the courses. The questions will be like:

•
In your life, in which situations do you need to speak French? • What do you like about French courses? • What's the most difficult thing for you in the courses?
There is never a right or wrong answer: We simply want to know your reality and your opinion.
Take some time to think and talk with your colleagues during lunch time, and we will come back after lunch and ask if anyone wants to participate and then we will go and form the discussion groups.
Thank you very much! Do you have any questions?

Information and Consent Form
Hello, You are invited to participate in a research on your experience in French classes. The objective of this project is to portray the French learning of immigrants with little education or literacy skills in Quebec according to different people involved (students, teachers, educational advisors and directors of centres offering French courses). To participate in this research, you must participate in a group interview in your language with a member of our team.
The personal information collected during the interview is confidential and anonymous: It will not allow for you identification. If at any time you wish to stop the interview, you can do so without giving any reasons and without any consequences.
If you agree to participate in this research, say "I agree". For any questions, you can contact the persons responsible for this research: This collection of information is carried out by the Université du Québec à Montréal on behalf of the Ministry of Immigration, Francization and Inclusion who commissioned this action-research project.
Thank you very much for your participation.
To be completed by the interpreter ______________________ (name of the participant) agrees to participate in the research. I obtained his/her consent orally. In Quebec, two ministries are responsible for providing French teaching services, including FSL courses for SLIFE, to immigrants: these are the Ministère de l'Immigration, de la Francisation et de l'Intégration (MIFI) and the Ministère de l'Éducation (MEQ).