Ethical issues in multilingual research situations: a focus on interview-based research

Abstract

Interview-based research in multilingual situations can present researchers with specific ethical challenges relating to language-based power play, data handling and presentation. Studies indicate favouring the L1 (first language) as an interviewing language may produce better quality data, but external pressures can favour English as the dominant research language. This article examines researcher perceptions and experiences of the ethical consequences of language choice and the practical issues involved. Interviews were conducted with five European researchers working on an interview-based project with experiences of diverse interviewing scenarios. The four moral principles of respect for autonomy, justice, beneficence and non-maleficence were used as a framework for analysis. The analysis revealed a nuanced picture of ethical issues in both L1- and English-oriented scenarios. This included potential misrepresentation and deculturalisation of the data in the former, and language-based power asymmetries in the latter. The findings highlight the importance of documenting ethics-related methodological details of language use, and advocates publication practices favouring the inclusion and foregrounding of L1 data.

Introduction

Although the European Commission has produced a significant body of research ethics policies and guidelines, there are as yet no specific guidelines for the handling of ethical issues associated with multilingual research. This in spite of the fact that, with 24 official languages in its portfolio, the European Union is language rich, and research within EU-funded programmes is designed to exploit its cultural diversity. A similar gap is evident on a smaller scale in ethics boards in higher education institutions (Koulouriotis, 2011), which Holmes (2016) has perceived as perpetrators of ‘monolingual regimes’ that need to be resisted in the process of carrying out multilingual research.

To our knowledge, research guidelines generally fail to address multilingual research situations. Broad ethical issues in research, such as differential power relations, have been given due attention (Cohen et al., 2018), but an understanding of how these are played out in the complex language arena within which multilingual research takes place is needed. This is particularly the case for interview-based research where language operates as the medium of data collection from participants. While The Routledge handbook of research methods in applied linguistics (McKinley and Rose, 2020) does include a chapter on the topic (Andrews et al., 2020), ethical issues are not a point of focus.

Nevertheless, ethical challenges have been noted by researchers who have experience of multilingual research. The following background highlights some examples of the concerns they have raised.

Following interviews with Chinese teachers initially in English and subsequently in Chinese, Janusch (2011) reflected that ‘no matter how much we may think we know about what someone really means by what they say in another language’ (p. 85), its interpretation by researchers using a dominant language is always coloured by a ‘perspective [that] comes from another cultural orientation and a position of power and privilege’ (p. 85). Janusch observed that switching to Chinese was empowering for the participants and allowed them to express themselves more fully and accurately, thus resulting in richer, more valid data and generally contributing to more robust research outcomes.

Other studies have revealed complexities regarding the issue of empowerment and its effect on data collection. There is, for example, some evidence to suggest that more natural avenues for probing and constructing meaning tend to present themselves in interviewing situations in which the interviewing language is the interviewee’s L1 (first language) but the interviewer’s L2 (second language), rather than a shared L1 (Chen, 2011). Furthermore, researchers such as Koulouriotis (2011) have commented that in some cases, participants have appreciated and felt a sense of pride at being given the chance to express themselves in a dominant language (in this case English), although this did not resolve issues related to ease of expression and required increased efforts to construct meaning during the interview.

Duff and Abdi (2016) document the problems faced in navigating the requirements of institutional ethics boards in Canada for multilingual ethnographic study. The authors present insights that show how standard protocols for English-speaking mainstream communities were inappropriately framed and worded for use in indigenous communities, even in translation.

A critical viewpoint regarding ethics and positionality in ethnographic multilingual research is provided by Holmes (2016). In recognition of the ubiquitous nature of ethical concerns within the multilingual research process, Holmes advocates more agentive ethical practices and stresses the importance of researcher reflexivity. Holmes encourages researchers to take an active involvement in questioning monolingual state and institutional practices, and generally fostering a more multilingual approach to research. This is particularly important for ethnographic research carried out on site, in order to promote social justice and solidarity.

Cormier (2018) asserts that issues of language dominance will necessarily impact data handling, particularly when a dominant language is used for the purposes of translation and interpretation. There are concerns about the negotiation of meaning, which has been seen to affect data validity. Kapborg and Berterö (2002) analysed interviews carried out in English with the help of an interpreter when the participants’ L1 was Lithuanian. They identified threats to validity such as lack of awareness of summarised or modified responses in the process of data collection. Where the researcher also acts as translator, this adds another level of complexity. There is increased responsibility attached to the role of researcher-translator due to the ‘monopoly’ of power in the process of meaning construction (Shklarov, 2007: 535). Again, the use of reflexivity and accurate documentation of details pertaining to researcher positionality, and debatable instances of meaning-making, are suggested as ways of counterbalancing these risks (Shklarov, 2007).

Tietze (2018) has critically questioned language practices in the monolingual publication of multilingual research, arguing that at the publication stage ‘translation becomes invisible’ (p. 30) and a great deal of language richness is lost. Cormier’s (2018) reflexive means of counteracting the misrepresentation, inaccuracies and loss of richness of original data highlight the importance of transparency and entail a more detailed documentation of the language-oriented aspects of the methodology. In the case of translation, Cormier favours late (post data analysis) translation and the inclusion and foregrounding of original linguistic data in the presentation of the final paper. The latter choice would also counteract practices through which researchers have considered that ‘minority languages are rendered invisible’ (Holmes, 2016: 334) in multilingual research.

Mohamad Nasri et al. (2021) reviewed 35 curriculum and pedagogy studies that focus on methodological procedures as reported in qualitative multilingual research studies. They found that the majority did not report details of six core elements needed for ensuring acceptable levels of methodological rigour in multilingual research, including ethics-related aspects of the methodology. In particular, justification of language choice during data collection and analysis was often lacking, as were details of which language was used, at which point in the research process, and details of translator/interpreter roles and credentials. Related limitations were also often left unacknowledged. The importance of transparency in documenting the role of language in the data collection process in interview-based studies and for ‘invisible translation to be made visible’ has also been highlighted by Resch and Enzenhofer (2018: 144).

As is clear from the above, there is a need for consideration of the ethical implications associated with multilingual research. However, a deeper and broader understanding of the issues involved, as perceived and experienced by researchers, is needed to provide a holistic understanding of what ethically sound multilingual research involves. Ethically sound research behaviour needs to be situated and context-based (Hammersley and Traianou, 2012). Further, as Kimmel (1988: 34) points out, ‘ethical principles usually are not absolute, but must be interpreted in light of the research context’.

As linguists involved in carrying out the research for this paper, as well as in our role as researchers operating in multilingual situations ourselves, we face the challenge of articulating an ethical standpoint that reflects our views on the ethical use of language in multilingual research to provide a coherent approach to the handling of research situations that involve language choice. To help address this need, we used empirical data for a context-based analysis of ethical issues in multilingual situations. The emergent ethical issues are analysed in line with a principlist approach to ethics (Beauchamp and Childress, 2013) via the four principles of respect for autonomy, nonmaleficence, beneficence and justice. Beauchamp and Childress (2013) state that these four principles are ‘derived from the common morality’ (p. 13) and we therefore assume that they can be applied to settings other than biomedical ethics.

We address the question: What insights can researchers with first-hand experience of interview-based multilingual research situations provide on the handling of ethical issues and associated practical implications? It is hoped the findings will inform a broader understanding of the complexities of ethical issues that come into play when researching multilingually.

Methodology

Context

This empirical investigation forms part of a broader study on issues and challenges facing researchers carrying out interview-based research in multilingual situations conducted by the Academic Literacy Team (ALT) within the EU funded project ‘Strengthening Europeans’ Capabilities by Establishing the European Literacy Network’ (COST Action IS1401). Members of the ALT completed a questionnaire and were subsequently interviewed by the authors (also part of the team) about their research practices and experiences. The focus on ethical issues was included as an aspect of the broader study following a suggestion made during a pilot interview. The topic worked its way organically into discussions about issues and challenges facing researchers carrying out interview-based research in multilingual situations. Details of the methodology, including a description of additional procedures carried out for this investigation are given below.

Participants

Five ALT members consented to be interviewed. All were multilingual European nationals fluent in English, had spent from 6 to over 20 years in academia, and carried out from 12 to more than 60 research interviews in multilingual situations. The participants are referred to by the pseudonyms Simone, Nina, Ella, Bessie and Lucy to preserve anonymity.

Data collection

The participants answered an online questionnaire that comprised 15 questions covering demographics, language background, affiliation and details of the linguistic context in which the research interviews they had carried out took place.

Following the preparation of an interview protocol, part of which was customised for each participant in accordance with their individual questionnaire responses, the researchers carried out an online, semi-structured interview of circa 60 minutes with each of the five participants. The first interview served as a pilot but produced rich and useful data and was therefore subsequently also included in the data with the permission of the participant. Since L1s were not shared, English was used as lingua franca. Towards the end of the interview, participants were asked to reflect on whether and how the interview might have been affected had it taken place in their L1.

All interviews were recorded and subsequently transcribed in full, two by one of the researchers and three by a third party. ‘Ethical issues’ had been included in the initial thematic coding scheme constructed after segmentation of the transcribed interview data. An inter-rater reliability test gave complete agreement on 76% of the codes. Since the discrepancies were superficial, subcodes were included for the relevant segments and the coding was considered reliable to a satisfactory level.

For the purpose of addressing the research question, a second layer of analysis was carried out. Audio recordings were re-examined in full without reference to the previous transcriptions. The parts of the interviews relevant to the research question were identified and compared to the segments coded ETHICS in the original transcripts. This provided a finer focus on ethical issues. Two further issues were thus identified: ethical issues relating to interviewers (as opposed to interviewees), and issues relating to language-based power and position during the interview situation. Thematic analysis was subsequently carried out on the full complement of ethics-related data. A list of topics was compiled, and a framework constructed categorising each topic into broader underlying ethical issues. Practical implications as perceived by the participants were included alongside corresponding topics.

Beauchamp and Childress (2013) four moral principles were used as guidelines for interpretation and discussion of the three broad ethical issues into which the topics were grouped. The three issues: language-based asymmetries of power, misrepresentation of language data and deculturalisation of language data are discussed below. This approach allowed for any interplay between the moral principles to be discussed in relation to each issue.

Ethics approval

Ethics approval was granted by the University Research Ethics Committee (UREC) of the University of Malta. An information letter with details relating to objectives and participation was sent to the participants and their consent obtained in writing. All references to potentially identity-revealing details, including mentions of specific projects, were removed from the data.

Results and discussion

As described in the introduction, ethical issues in interview-based research extend beyond the interview situation itself, through to publication. Our contention is that languages represent the people who speak them, and should be respected as such. Furthermore, in interview-based research, the language data is the product of individuals, and therefore, when researchers handle the data, they are not examining language generically, but handling words spoken by specific people.

In the discussion that follows, Section 3.1 addresses issues within the interview-situation itself, whereas Sections 3.2 and 3.3 examine issues related to the language data in the later stages of the research process. Transcription excerpts are numbered and followed by a bracket identifying participant and segment, such that in the first excerpt (Bessie: S6) refers to the sixth segment of the transcription of Bessie’s interview.

Language-based asymmetries of power

Resch and Enzenhofer (2018) have recently foregrounded the perception that language as a power-based construct exerts an empowering/disempowering influence. This happens in everyday life where language acts as a barrier to people speaking minority languages in multilingual communities, but also in situations where research is carried out multilingually. In situations where research is interview-based, power in setting up and handling the interview situation typically rests with the interviewer (Kvale, 1996: 3); asymmetry of power in favour of the interviewer is a pervasive characteristic of interview situations unless deliberate steps are taken to counterbalance it. Non-linguistic features of power asymmetry, such as the prerogative to choose the topic and ask the questions, remain a characteristic of multilingual interview situations. However, the focus here is on language-based asymmetries of power, often discussed in terms of ‘dominant’ (Cormier, 2018; Janusch, 2011; Resch and Enzenhofer, 2018) and ‘minority’ (Holmes, 2016) languages.

The ethical issues surrounding interviews in multilingual situations often revolve around language choice or participant autonomy, where the principle of respect for autonomy dictates that ‘to respect autonomous agents is to acknowledge their right. . . to make choices’ (Beauchamp and Childress, 2013: 106) and this ‘involves respectful action, not merely a respectful attitude’ (p. 107). Various interviewing scenarios exist, in which it may or may not be possible for interviewees to use their preferred language, usually their L1 (Schembri and Jahić Jašić, forthcoming; Cortazzi et al., 2011). In L1-oriented scenarios, interview data is collected in the interviewee’s L1, either because they are unable to speak English or by choice. Similarly, English as the dominant research language (Englander, 2014; Lillis and Curry, 2013) may be deliberately chosen as the interviewing language over the interviewee’s L1 or become necessary as lingua franca in English-oriented scenarios (Schembri and Jahić Jašić, forthcoming). The data indicate that situations may ensue in both types of scenarios that require ethical consideration. This subsection will focus on English-oriented scenarios where, however, the interviewer and the interviewee also share an L1. The choice of using either language is therefore present and an ethical responsibility to respect autonomy and participant choice exists. Ethical issues that ensue when the L1 is chosen instead will be discussed in sections 3.2 and 3.3.

Some of the effects of interviews taking place in English when the L1 could have been chosen instead are described by Bessie:

if you add another layer which is not perceived as necessary for the interviewee, if we both shared the same first language and we add a layer just to make things easier for the interviewer, it’s not fair for the interviewee because that’s adding difficulty for him or her (Bessie: S6)

In a similar situation, Nina (S8) also felt she was adding ‘another layer of complexity’ which was potentially a ‘burden’ for the people who had offered to be interviewed for the project, even though they had given their consent to be interviewed in English. Naturally, lack of ease of communication is exacerbated with lower levels of proficiency. This was the case with one of Nina’s interviewees who, in Nina’s view, ‘didn’t quite get her point across or didn’t feel like she was fully . . . able to explain to me her experiences’ (S8).

The fact that such situations are not likely to produce optimal data is at once clear, but a more detailed picture of the effects is given by Bessie, who favours conducting interviews in the L1, and Ella, who tends to prefer using English. Bessie describes such situations as follows:

and in some cases that we tried. . . to use English. . . if the other is looking for a word, you help them, then the others start to speak in [the L1], then you heard something in [the L1], then you translate this same thing, and then it’s kind of a mess (Bessie: S6)

Ella’s description of two interviews she carried out in English is similar:

at times they couldn’t come up with the English words, so they would say it in [the L1] and I would translate it, or they would use some English expression that . . . I wasn’t sure what they meant, and I would try to clarify with my own words, and I had to confirm that what they were saying was basically . . . what I understood (Ella: S2)

From an ethical standpoint, these situations reveal a struggle for self-expression that may leave the interviewee feeling inadequate, possibly to the extent of relying on the interviewer’s input to construct meaning, as was the case with Ella’s interviewees. The reality is also that the situation is conducive to lapses into codeswitching, defined here as the use of more than one language ‘in the same conversation or sentence’ (Gardner-Chloros, 2009: 4). Apart from the specific challenges these may present to translate (see, for instance, Halai, 2007), codeswitching data may be considered of questionable validity when interviewer input is needed to resolve issues of meaning making.

It seems fair to say that the choice of English as interviewing language when proficiency levels impede ease of communication can have a disempowering effect on the interviewee and tends to confirm Janusch’s (2011) contention that richer, more valid data is collected in the interviewee’s L1. Furthermore, if this results in a lowering of the interviewee’s self-esteem, the practice may be considered ethically questionable (Robson and McCartan, 2016: 211) and in breach of the principle of non-maleficence, which ‘obligates us to abstain from causing harm to others’ (p. 150); and the principle of justice, in particular its interpretation as a multi-faceted principle operating on the basic premise that ‘social justice is concerned with human well-being’ (p. 261). Using English in such cases can also be considered as missing an opportunity to enhance rapport and trust by matching interviewer/interviewee characteristics (Cohen et al., 2018: 136), and therefore holding back on a potentially benevolent act.

The obvious solution seems to be to allow interviewees the choice of interviewing language, as Nina (S15) suggests. The choice may still be English, but it can be a positive choice when it is motivated by familiarity with certain topics in that language, perhaps as a result of training, for example. Both Lucy (S17) and Simone (S13) commented they might have had problems discussing technical aspects of the topics we interviewed them about had they been using their L1. This supports Cortazzi et al.’s (2011: 508) comment that ‘interviewers should consider that some concepts are more easily explained in one language rather than another’.

Interviewees may also, however, decide to choose English for reasons that have nothing to do with familiarity, as Simone points out:

people sometimes. . . feel obliged to speak in a foreign language, for instance if you work with teachers, they might feel that if they choose another language that would reflect on their confidence in speaking the foreign language. . . the same with students, so they might want to speak in English because they feel. . . they will project themselves in a better way (Simone, S3)

Koulouriotis (2011) noted that interviewees will sometimes avail themselves of what they feel is the privilege of using the dominant language, even when they might have to struggle to achieve adequate levels of self-expression. The ethical issues in such cases are not quite as clear-cut since any resulting feelings of inadequacy can be interpreted as being self-imposed, generating conflict between respect for autonomy and beneficence. Researchers must take positive action to contribute to the welfare of the persons involved by ‘helping to allay fears and other conditions that destroy or disrupt autonomous action’ (Beauchamp and Childress, 2013: 106) but there can be a fine line between beneficence and paternalism.

English-oriented scenarios may also present challenges to interviewers, and it is important to bear in mind that their well-being should also be considered. Bessie describes situations occurring in international European projects when interviewers who are uncomfortable using English use it to carry out interviews for the benefit of interviewees who speak English as a L1:

the difficulties were related to the degree of comfortability or mastery that the interviewer had of English as a second language. . . they say that they don’t feel comfortable, they were asking [the interviewees] all the time ‘do you understand me?’ (Bessie: S10)

The burden of expression and feelings of inadequacy are once again self-imposed. However, in this case, Bessie’s comments support Chen’s (2011) contention that the resulting interviewing dynamics may present more natural avenues for probing and clarification.

In conclusion, one may reflect that the linguistic power asymmetries present in the various interviewing scenarios are probably best addressed in terms of dominant and marginalised languages, rather than linguistic power play between interviewers and interviewees. In English-oriented scenarios, English as the dominant research language tends to be preferred by interviewers and sometimes even by interviewees, even when either may find themselves struggling to express themselves. Language dominance-related ethical issues may consequently not be as easy to resolve as simply allowing interviewees the choice to use their L1 may suggest. The more far-reaching pressures that motivate language choices will tend to make themselves felt at a later stage in the research process in L1-oriented scenarios, as the next two subsections reveal.

Misrepresentation of language data

This section and the next focus on ethical issues surrounding the handling of language data. We argue here that researchers have an ethical obligation to handle language data such that its use beyond the interviewing context is representative of its contextual meaning, as intended by the interviewees, as fully and authentically as possible. This is in line with Temple et al.’s (2006) foregrounding of the importance of contextual awareness in data handling in multilingual situations. Furthermore, since language is an intrinsic part of one’s identity, misrepresentation of language data is considered harmful in that it shows lack of respect for the autonomous right of agents to ‘hold views’ (Beauchamp and Childress, 2013: 106) and is misrepresentative of aspects of their identity. Researchers may be aware of this; Nina comments that she invited her interviewees to look at the excerpts she used from their data and asked them ‘do you recognise yourself in this?’ (Nina: S16).

Our findings indicate that researchers operating within L1-oriented scenarios face specific challenges at the data handling stage of the research process. Issues mainly revolve around the necessity of translating L1 data into English to report their research. Bessie explains:

it’s difficult to guarantee that the meaning has been respected by someone who is not part of the research, who don’t understand the situation in which the interview was done. . . and that’s an ethical issue, reason for the researchers not to send an interview to. . . a professional translator, so that requires us to be the researchers and the translators at the same time, it’s the only way to guarantee that ethically we preserve the meaning, we preserve sense (Bessie: S14)

Bessie’s misgivings about misrepresentation of data handled by outsiders are supported by Choi et al. (2012: 660), who emphasise the importance of constant communication between translators and researchers following instances of differences between their own understanding of interview content and its subsequent translation by the translators hired for the project. In an earlier study, Temple et al. (2006) also examined language-related issues in the handling and analysis of L1 data requiring the intervention of interpreters/translators. They foregrounded the importance of context in meaning-making and contended that this is only available to those directly involved in the research. As aforementioned, however, this added responsibility also presents the risk of researchers monopolising the process of meaning construction (Shklarov, 2007). Furthermore, researchers are not necessarily trained translators. As Simone points out: ‘it’s just assumed that yes you have that skill and it’s taken for granted but it’s not an easy thing’ (S10). Resch and Enzenhofer (2018: 138, following Przyborski and Wohlrab-Sahr, 2008) have suggested researchers should implement a critical self-assessment of their skills in such cases.

Lucy suggests that in order to counteract the problem of misrepresentation, the consent form should specify that interviews will be translated, and the interviewees given the opportunity to go over the English version ‘to make sure the translator hasn’t distorted anything’ (Lucy: S16). She has herself, however, not always been able to do so:

it’s not because I don’t think it’s a good idea, it’s because of time constraints and because sometimes you don’t know how much of the data is getting used, and then by the time we decide to use it. . . they won’t remember what they said anyway (Lucy: S16)

A more delicate issue arises when misrepresentation of language data is wilful. Nina describes issues of transparency when interviews are translated in part:

if people want ‘well, I want to see these interviews and I don’t speak [the L1], where are they?’ . . . if you can’t provide that transparency then readers of your work have to trust you even more, that you’re not just making stuff up. . . because they don’t understand what this means anyway, so I can just. . . tweak this data to fit whatever I want it to say (Nina: S12)

She points out that such problems also arise during supervision, when supervisors of dissertations and theses are not familiar with the L1 (Nina: S12), a point Simone (S9) also makes.

In summary, researcher insights indicate that multilingual research situations lend themselves to language-based forms of misrepresentation of data when interviews are carried out in the L1. This goes some way towards counteracting the likelihood of obtaining better quality data in L1-oriented scenarios. Researchers such as Holmes (2016) and Cormier (2018) have advocated reflexive practices in reporting multilingual research to mitigate such problems; however, studies such as Mohamad Nasri et al. (2021) and Cortazzi et al. (2011) reflect a tendency to occlude methodological detail regarding the handling of language data.

Deculturalisation of language data

The deculturalisation of language data can be considered an aspect of data handling that has a marginalising effect on the cultural identity of interviewees and affects the authenticity of the data in terms of cultural context. Our findings reveal ethical concerns about the effects of publication in English on the reporting of research based on L1 data. One concern is how this may affect the selection of excerpts:

there’s a lot of extra work in translating things, and a lot of the times you need to also contextualise the translation itself so it’s very, very difficult, and I think it does affect the way you select excerpts to include, because there could be some really good excerpts but you think ‘ooo this is difficult to convey in a good way, so I’m gonna find something else’ (Simone: S9)

An important point here is that the difficulty often lies in the cultural specificity embodied in the excerpts. Nina explains:

some cultural linguistic contexts. . . are just very difficult to represent in English, you would have to do so much explaining, I remember that there was one term in [the L1] that people would often use about country music saying that it’s [speaks in L1] which means it’s not clean enough for the living room. . . it’s something that. . . you don’t bring it into the house, but it’s also something that you use about cats that are not trained to do their business outside. . . and I’m just like ‘how do I represent this?’ and sometimes you just can’t (Nina: S10)

This example makes it clear that retaining culturally-specific concepts involves not only more effort, but more dedicated text for explanation in a writing environment where word count is at a premium. As Simone also points out, excerpts are not always equally effective in translation, and there is a tendency for researchers to choose less culturally-specific examples. A method in which researchers involve translator research assistants in discussions about both oral and subsequent written translation of interviews has been advocated by Inhetveen (2012). This is intended to refine the research process in multilingual situations and identify ‘semantic fields and phenomena that are connected with obviously difficult translation decisions’ (Inhetveen, 2012: 42). However, a doubling of translation efforts by the same translator may be difficult to justify in the face of time pressures and limited financial resources.

The problem of presenting translated excerpts is exacerbated by the fact that these are often not accompanied by the original version. Bessie reflects:

I remember some time ago when we were publishing results in English and the interviews or the original data was collected in another language, what we usually did at that time was to add a footnote in which we offered the original quotation in the language in which [it] was recorded, and more and more this has been disappearing (Bessie: S14)

She explains that she herself has difficulty keeping up this practice because editors often consider it unnecessary and insist on shorter versions (Bessie: S14). Apart from being a methodological issue because translation is occluded (Tietze, 2018), Holmes (2016) points out that such practices also render minority languages invisible, thus effectively contributing to a deculturalising effect. Simone contends good practice would entail giving ‘prominence to the other language and just to the idea that “well, this language was there”’ (S10), but also comments that ‘often there is no space’ (S10) for both languages.

Researcher perception therefore indicates the practical challenges in the handling and presentation of original language data have ethical reverberations. A great deal of the added cultural value of data collected in the L1 risks being lost because it is not made visible at the publication stage. However, its inclusion requires more time for translation and for the annotation and explanation of culturally rich data to make it accessible to the wider research community. One must also appreciate that researchers sometimes face a choice between publishing better quality research that is more time-consuming or publishing more papers that make fewer demands on their time. Unfortunately, the institutional environment within which they operate does not always reward the former (Berman and Tyyskä, 2011; Jahić, 2016).

The perceptions of the researchers we interviewed indicate that deculturalisation of language data is an infringement of ethical principles that operates through occlusion, by actions that reduce the visibility of culturally specific data and present language data in translation only. Counteracting deculturalisation would therefore entail allowing culturally rich data and original language data more visibility, in what can be interpreted in ethical terms as benevolent research acts. It is possible to argue that less harm is inflicted through avoidance than through positive harmful acts. However, we believe the marginalising effect of minority languages is harmful on a broad scale to the speakers of those languages as well as to the interviewees whose data is directly involved, undermining their cultural identity in a research environment that it has the potential to enrich.

Conclusions and recommendations

This analysis set out to examine researcher insights of interview-based multilingual research situations with a view to providing ethics-oriented guidelines for practice in the field. The data reflect a nuanced picture of ethical concerns that foregrounds the importance of context-based ethical decisions involving language dominance and language choice throughout the research process. Researcher insights indicate interviews in English may be disempowering when interviewees lack appropriate proficiency and also carry the risk of compromising the quality and validity of the data. On the other hand, the need for translation in L1-oriented scenarios lengthens the research process and may result in misrepresented or deculturalised data.

Ethically sound research contributes to research excellence. However, understanding what constitutes ethically sound research in terms of language choice in multilingual interview-based research is not a straightforward one. Although we do not aim here to provide a definitive definition of excellent multilingual research, we are of the opinion, nevertheless, that this analysis brings us closer to defining what it entails. The ethical nuances surrounding language choice foreground the importance of documenting ethics-related methodological details of language use and their potential limitations in research reports. In essence, the quality of multilingual research is enhanced when researchers make situated language choices and account for them appropriately in their research reports.

In view of this, we propose the following recommendations:

• • At the recruitment stage, make potential interviewees aware of the language choices available

• • Prepare multilingual interview protocols

• • Invite interviewees to switch between languages (codeswitching) where possible

• • Communicate closely with translators about interview content

• • Implement critical reflexivity when researchers take on the role of translators

• • Give interviewees the opportunity to go over the translated version of their interview

• • Document ethics-related methodological details of language use

• • Devote space to culturally-rich examples

• • Promote original interview language visibility

It is hoped the article will increase awareness of the ethical issues involved in interview-based research in multilingual situations and promote the foregrounding of L1 data as an element conducive to a scientific practice that is more reflective of the cultural diversity of society at large.