The (un-) ethical interpreting researcher: ethics, voice and discretionary power in interpreting research

ABSTRACT This article discusses different aspects of research ethics, the researcher’s voice and discretionary power in interpreting studies. Research ethics is laid down in international conventions, which in turn are reflected in national regulations and ethical vetting. Discretionary power is understood as the leeway for making conscientious decisions within the rules and regulations governing a certain field. Although research ethics in interpreting has as yet received little scholarly attention, it is important that the field discusses aspects such as informed consent and the collection, analysis and reporting of data. This article uses three case studies to discuss how researchers can handle such ethical issues. Interpreting researchers often are or have been active interpreters, and this is yet another potential challenge for the field. Such duality potentially means that the researcher needs to navigate two ethical systems, that of the interpreter and that of the researcher – systems that may come into conflict with each other. It may also entail the risk of the researcher’s voice taking over the participants’ narrative.


Introduction
Research ethics covers a wide range of topics from informed consent to norms of reporting. Ethical considerations in any field involving humans may be challenging, although participating in a study within the humanities might not cause any physical harm to participants. This article will discuss ethics, the researcher's voice and discretionary power in research on interpreting, a field where humans are often involved in studies.
Ethics in interpretingthat is the ethical implications of an interpreter present in the room, or different aspects of the individual interpreter's ethicsis, not surprisingly, a widely researched and discussed area in interpreting studies (e.g., Hale, 2007;Inghilleri, 2012;Valero-Garcés & Tipton, 2016). Many countries and professional organizations have ethical guidelines for interpreters (Hale, 2007). Bendazzoli (2016, p. 24) points out the necessity for interpreting research to focus on the researcher and not only problems and methods. Indeed, despite the wealth of literature discussing the ethics of the practising interpreter, research ethics in interpreting has hardly been discussed at all.
Many researchers who conduct research in interpreting are former interpreters who sometimes also teach interpreting, so called practi-searchers (Gile, 1995;Pöchhacker, 1995). This entails that the interpreter/researcher often has a deep pre-understanding of the field. It also means that the collection of data, whether experimental or natural, may involve gathering data from colleagues. In some cases, in fact, the researcher's access to data may be granted only thanks to his or her collegial relationship with the research participants, as described for instance by Bendazzoli (2016) and Biagini (2016). The researcher's position in research which involves collecting data from former or present colleagues may be delicate in terms of handling results which are less flattering for the participants. The delicate position between colleagues' trust and researcher's duties may challenge the ethical obligation to report findings truthfully. In such cases the interpreting researcher has to navigate two ethical systems, that of interpreting and that of research. For the interpreting researcher who collects, analyses and reports research, it is important to be attentive to whose voice is really represented in the data: the participant's voice or the researcher's. In other types of interpreting research which do not involve peers or former or present colleagues, it may be important that the researcher understands what their discretionary power means in that type of research.
The present article discusses three authentic research cases from different areas of interpreting research: the first case deals with experimental research on simultaneous interpreting, the second case comes from an observational study of health care interpreting, and the third case describes an interview study with deaf interpreters. The overall aim is to discuss the ethics of interpreting research, the more or less present voice of researchers in the field of interpreting research, as well as the discretionary power of these researchers. The article does not aim to deliver any final answers to these issues but rather to open up for further discussions about these delicate and important areas of interpreting research. The basis for the discussion will be the three different cases that will problematize different aspects in regard to research ethics, interpreting ethics and the possible impact of the interpreterresearcher's professional background.

Interpreting research
Most studies on interpreting are done within the field of interpreting studies or in neighbouring disciplines and by researchers who have a (longer or shorter) professional background in interpreting. As mentioned above, early discussions on interpreters as researchers labelled these researchers practi-searchers (Gile, 1995;Pöchhacker, 1995). However, interpreting researchers today go through the same research training as researchers in other fields. Contributory factors to this development include the CETRA summer school in translation and interpreting hosted by KU Leuven, the fact that many interpreting programmes today are integrated at university and promote research training (Pöchhacker, 2016), as well as the publication of volumes that specifically focus on research theory and methods in interpreting (e.g., Angelelli & Baer, 2015;Bendazzoli & Monacelli, 2016;Hale & Napier, 2013). Interpreting researchers continue to have a practi-searcher profile, creating a double ethical challenge for the interpreting researcher as well as a very specific emic perspective.
The emic and etic perspectives can be described as the researcher's approach from within or from outside, respectively (Pike, 1967). A traditional understanding of emic researchers is that they learn or integrate in the topic or the object of study to such an extent that they can provide readers with an insider's view of the topic, as exemplified by the modern anthropologist who lives with and within the studied community (Hannerz, 2010). Etic researchers, on the other hand, study the topic or the object with an outsider's view, studying it closely but from the outside. For the interpreter and interpreting researchers, the emic insider's perspective becomes even closer. Researchers with this special insider perspective can presumably never claim to investigate the well-known object of study from an etic perspective, but the issue arises of whether or not a researcher with such deep background knowledge in a given areafor example interpreters (which will be discussed here), nurses (Coar & Sim, 2006) or teachers (Hockey, 1993)can unveil their preconceived understandings of the area to study or question all aspects of the research object.
The emic research perspective, and conducting research with peers as participants, informants or subjects, should not be understood as solely relating to ethnographic research. In interpreting research, many research studies are carried out in an experimental context using quantitative methods but where the participants, informants or subjects are still peers of the interpreter-researchers. Perhaps more than any other research community, interpreting researchers conduct research on their peers (Johnsen, Halvorsen, & Repstad, 2009). In interpreting research, empirical databoth quantitative and qualitative, experimental as well as naturalmost often come from interpreters who are peers of the practi-searcher. 'Peer' in this context should not be understood as friend or colleague, but as someone belonging to the same professional group. This means that researchers not only mix with their object of study but are actually an active part in that community on the same conditions as other members, which in turn entails certain challenges of proximity and distance: In order to be credible as researchers, they need to maintain a necessary distance to their participants, even as their proximity gives them unique access to data material and participants (Bourdieu & Wacquant, 1992). Bendazzoli (2016, pp. 4, 24) points out the importance of interpreting researchers making their stance clear both as researchers and in the interpreting community. Bendazzoli's statement concerns ethnographic research, yet the statement is relevant to all types of research in interpreting. Research reports more or less clearly state whether research is done on peers (e.g., Duflou, 2016;Koskinen, 2008;Tiselius, 2013). In other research reports, such relationships with peers may not be as clearly stated (e.g., Cecot, 2001;Chmiel, 2016;Timarová, 2012). Although researchers may not know their participants personally, interpreters who carry out research on other interpreters do so on their peers, whether or not they know them well. I would argue that the position of the researcher is ethically crucial in all types of interpreting research involving interpreters as participants.

Research ethics
Research ethics is not a well-defined area and is tinged by culture, country-specific legislation and different research traditions (Bridges, 2016, p. 306). The Declaration of Helsinki is a global and morally binding set of principles for research on humans. Developed in the aftermath of the Second World War, the declaration lays down the individual's right to decide themselves, receive enough information to make an informed decision about participation, and have their welfare always come first (World Medical Association, 2013). Research ethics is so much more than not harming participants, however, and also consists of norms for the research community, the relationship of this community to the rest of the society, and how we view scientific practice (Rowson, 2010).
It is important to stress that the creation of basic principles for research ethics does not imply an ethics-free zone where no personal or general ethics are applicable (Hermerén, 1996). Personal morals or overarching ethics do in fact apply, just as for any area in life, but researchers discuss ethical standards more in detail in order to organize their work and protect participants. Such a position also means that research ethics can be defined as a type of utilitarianism, that is that a human should act according to what generates the best result for as many as possible, in contrast to deontological ethics, which implies that rules are made to be followed and the right decisions are those following the rules, no matter the consequences.

Voice in interpreting research
Voice is a complex concept in any setting, as it implies both the actual acoustic signal that carries a person's speech and also a person's opinions, ideas and presentation of self (Goffman, 1949;Linell, 2007). In interpreting, similarly to translation, it also refers to the conveyance of someone else's voice in terms of opinions, ideas and presentation of self, but sometimes also reflecting the interpreter's own voice (Wadensjö, 1998). A norm in interpreting is to interpret in direct form, thereby veiling one's own voice, although never completely (Duflou, 2012). The interpreting researcher also conveys voice, but often not the voice of a single individual, but rather voice as a representation of many individuals. The concept of voice in this perspective is not uncontroversial, and issues are raised on the right of a researcher to speak for a community or a practice. In post-colonial and feminist studies, it has been argued that 'the [researcher's] practice of speaking for others is often born of a desire for mastery, to privilege oneself as the one who more correctly understands the truth about another's situation' (Alcoff, 2009, p. 134). Voice, similarly to ethics in interpreting studies, has been researched from the perspective of the interpreter (for an overview, see Martin, 2016), but not from the perspective of the (possibly problematic) researcher's voice in interpreting studies.

Discretionary power
Discretionary power is part of the professional autonomy of any individual practising a profession governed by rules and guidelines such as interpreting or research (Molander, 2011, pp. 320-321). Discretionary power is the ability of professionals to exercise their own professional judgment in carrying out and making decisions within the rules and guidelines governing a profession. Molander gives the example of civil servants, who have to take decisions that are unique for every individual but nevertheless based on current legislation, provisions and guidelines. It is the civil servant's duty to interpret the law so that the best interests of both society and the individual are protected in the decision making. As each case is unique, however, the civil servant has to exercise discretionary power to fulfil his or her duty. A functioning discretionary power of the civil servant is crucial for the individual client's trust in the institutions in a modern society (Molander, 2011). Molander describes discretion as a concept with two dimensions, a structural one and an epistemological one. The structural dimension is related to the actions described above, namely the selection of a decision or action which is possible, permitted and thereby available to the decision maker. In an epistemological dimension, however, discretion refers to the reasoning the individual goes through in order to ground a decision or an action in a situation of indeterminacy. In this dimension, discretion was one of the intellectual virtues Hobbes identified in Leviathan, describing it as 'distinguishing and discerning and judging between thing and thing' (Hobbes, 1651(Hobbes, /1962. The legal philosopher Ronald Dworkin (1978), who used a doughnut hole to symbolize discretionary power as a small space of manoeuvre surrounded by a belt of restrictions, defined structural discretion as context-dependent and occurring 'when someone is in general charged with making decisions subject to standards set by a particular authority' (pp. 31-32).
Like civil servants, researchers are also governed by rules, regulations and guidelines. They have to make decisions in specific cases on the basis of both rules and their understanding of the situation. An area which is crucial for any research involving humans is the participants' trust, as participants trust researchers with information which may be private and personal or otherwise important for the individual. The decision of how this information is handled and analyzed often depends on the researcher's discretionary power. When this power is ill exercised, the decisions risk breaching the trust of the participant. By allowing the researcher to study and share this information, the individual participants trust the researcher not to spread that information and not to let any harm happen to them. The researcher cannot promise unfailing anonymity or complete secrecy. There are situations overruling the participants' interest of anonymity, such as the suspicion of research fraud or an obligation to testify under oath. This said, it must, of course, be absolutely clear to the participant that the researcher who promises anonymous treatment and publication of data will under no circumstances seek to share information that can be directly linked to the participant. Pendlebury and Enslin (2002) discuss the researcher's discretionary power in education research and stress that '[r]esearch, like teaching and many other human practices, requires that the researcher be entrusted with a range of discretionary powers, yet in granting such discretionary powers the truster opens the way for the trusted to use her discretionary power for questionable, and even harmful, ends' (p. 363). It is thus crucial for the researcher not only to gain the trust of the participant but also to honour that trust. How the researcher decides to honour the trust is part of the researcher's discretionary power.
Skaaden first introduced the concept of discretionary power to interpreting. She defines the interpreter's discretionary power as lying within the rendering and coordination of other people's utterances (Skaaden, 2013, p. 88), following theoretical discussions by Wadensjö (1998) and Pöchhacker (2008) on the interpreter's task. This task is an area restricted by codes of conduct and also in some countries by provisions and regulations, and the interpreter can be said to carry the conversation without owning it (Granhagen Jungner, Tiselius, Lützén, Blomgren, & Pergert, 2018). However, despite the importance of the issue, the present article will not deal with the discretionary power of the interpreter but that of the researcher. Discretionary power can be seen in light of the interpreter's or the researcher's neutrality and visibility. In the field of interpreting there is an ongoing discussion of the interpreters' role in relationship to rules, regulations and ethical guidelines. Bendazzoli (2016, p. 23) takes the discussion to interpreting research. I argue in this article that discretionary power is what the interpreting researcher has to exercise in order to manoeuvre delicate issues such as ethics and voice in research. By exercising discretionary power, the researcher will stay true to the data and the reporting of the data.
I will use the three dissimilar cases below, all from different types of interpreting research, to suggest how discretionary power, voice and ethics in interpreting research can be understood in this context.
Case 1: further investigations are needed to obtain more information: reporting results that may challenge participants When emic interpreting researchers carry out research on their peers, they are given the trust of those peers. There are many reasons why active interpreters would decline to give of their own time for experimental research on the interpreting practice. Most interpreters work as freelancers and may not be willing to give their time for what is often a lower remuneration than what a potential client would offer. Even if the lower remuneration were to be acceptable to them, many interpreters doubt whether interpreting is really a practice that can be studied rather than an innate skill (Tiselius, 2013), and some interpreters may feel insecure about their performance levels under experimental conditions and about how that performance will be interpreted by the researcher. Finally, the interpreting world is small, and in some contexts it would be easy to identify participants from factors such as language combination, gender or nationality, so participating interpreters must be willing to trust the researcher to handle their anonymity with care. As a result of these factors, most studies on interpreting are not based on a randomly selected material, and it is highly likely that the participants have self-selected to participate in the experiments.
When I started my research career, I was an active interpreter, and the study I carried out for my PhD sought to describe the development of experience and possible expertise among conference interpreters (Tiselius, 2013). The participants in the study were divided into three groups according to whether they had much, some or no experience of such interpreting. The highly experienced participants in my study were chosen among veteran colleagues who had been approached during joint interpreting assignments, when I introduced them to my project and recruited them to participate in my study, while the somewhat experienced participants were recruited from former students of mine who had just started their interpreting career. It can be argued that both groups were biased when it comes to their selection for the study: upon being recruited, the experienced ones were told that they were to be investigated precisely as highly experienced interpreters, while the somewhat experienced ones, given that they were my former students, may have felt obliged to participate in my research. The inexperienced participants, conversely, were students who were not mine and who were recruited from different language departments at my university.
The study also included a second group of experienced interpreters who had already participated in an experiment conducted in the 1990s by another researcher during their interpreting training. This group was truly unique as it, to my knowledge, was the first group which could form the basis of an intra-individual, longitudinal study. Recordings of the interpreters' performance (i.e., interpreted conference speeches) existed from that first experiment, making it possible to record and compare the interpreters interpreting the same speech as during that first experiment. Once again, the participants in this group were approached at our joint interpreting assignments. They were all more experienced than I was, and they were also recruited from the perspective of being investigated as experienced interpreters.
As I started to collect my data, I, naively, could not imagine that they would do anything other than confirm my hypothesis that experienced interpreters would outperform inexperienced ones, even when the same interpreter was compared to his or her younger self. At that point, to me, it was therefore not challenging to recruit these highly experienced interpreters, as the results were assumed to confirm what 'we already knew', namely that experienced interpreters are 'better' than inexperienced ones. Some of my more critical peers and friends told me I was flogging a dead horse, or perhaps rather a dead interpreter. Indeed, when I began collecting data, I mostly stressed the need to scientifically confirm generally accepted truths. My main concern at this point was how to guarantee my participants' anonymity even though full anonymity can never be granted, as discussed above in the introduction. The data were collected through voice recordings, so, naturally, there was a certain risk a participant's identity might be revealed if somebody listened to their recording, but no sensitive data according to the Swedish Personal Data Act of 1998 (concerning religion, race, sexuality or political opinions) were collected. Nevertheless, as a colleague of theirs it was absolutely vital for me, from a perspective of trust, that the interpreters could not be identified by accident, for instance by my mentioning the fact that they were participating, or by recognition, for instance by their language combination or a quote that would be typical for a certain person. I treated both the data and my own involvement scrupulously so as to be worthy of their trust. For me as a novice researcher, I understood my discretionary power as being a commitment to respect and anonymity.
The performances of the interpreters in my study were investigated both from a process perspective, that is on the basis of reported processing problems, instances of monitoring and strategy choices (cf. Englund Dimitrova & Tiselius, 2009;Ivanova, 1999), and from a product perspective, where peer assessors evaluated each interpreter's professional performance according to a scale, using written transcripts to avoid identifying the interpreters' voices (Tiselius, 2013). The results of this investigation were not at all as expected: When compared inter-individually, highly experienced interpreters did indeed outperform their somewhat experienced colleagues in regard to process and product; however, when I compared them intra-individually (i.e., the same interpreters over time), I could not find a significant difference in the performance of the interpreter as a novice and the same interpreter some 15 years later, despite all external signs of professional success (such as lengthy and regular experience, peer appreciation, career advancement and so forth). Even more puzzling, I could not find any significant performance difference between these highly experienced interpreters and the less experienced participants (i.e., the recent graduates).
My new headache in regard to discretionary power concerned how I should report these results. Naturally, I needed to report them truthfully, but could I really trust them? And what would these results do to the participants' future willingness to participate in research? In this particular case, the question of discretionary power had thus evolved from an anxiety about unwittingly revealing anonymized identities to an ethical quandary concerning how the findings should be reported. The solution in this case was to report truthfully, but also to scrutinize my methods so that they would not be the reason for any skewed results. I decided to conclude that the results of the study went against the initial assumptions, and also that in contrast to the expertise theory, where the concept of deliberate practice is a central component, my assumed experts did not recount any practice that could be labelled deliberate at all. Deliberate practice, as described by for example Ericsson, Krampe, and Tesch-Römer (1993), is conditioned by time assigned specifically for practice, particular goals of that practice and specific exercises to improve and refine the main skill, and deliberate practice is also presumed to be one condition to reach excellence in a field. In my study I could confirm neither, and I concluded that my participants' performance was stable when improvement was expected. Either it must be assumed that these interpreters were not experts in absolute terms as defined by Ericsson and Smith (1991), or else further investigations are needed to obtain more information about their specific careers (Tiselius, 2013, p. 92). It should also be noted that the field in general lacks sufficient data to draw major conclusions, and that the expertise theory has received substantial criticism around the concept of deliberate practice (Detterman, 2014).

Case 2: what is a passive observer? Doing observations as an insider
The second case comes from a study of communication over language barriers in highly specialized hospital care of children . The object of study in this case was how communication takes place with families with little or no proficiency in the majority language (Swedish in this case) and whose children are seriously ill. The hypothesis in this case was that communication between the medical staff and the patients' families takes place, whether successfully or unsuccessfully, both when the interpreters are present and when they are not. In order to understand how and when communication was enabled, field observations were chosen for this study. As this study dealt with sensitive data concerning ethnicity, illness and the involvement of minors, ethical approval was applied for, and granted, by Stockholm's regional board for ethical vetting.
During the study, I first observed the activities at the hospital ward in question for a week in order to survey ordinary, monolingual care. After the initial observation, families who had little or no proficiency in the majority language were identified. They were then approached through an interpreter in order to obtain informed consent. The procedure of informed consent was made both through an interpreter and with a written consent form translated into the family's language. The participants could ask for clarification through the interpreter or have the whole consent form read to them. The interpreters who participated in this process were recruited through an interpreting agency and were either trained or state certified. Once informed consent was obtained, I, the observer, followed the family both in informal communication in the corridor and more formal medical consultations, as well as while they prepared for procedures, both with and without interpreters. At some point during the observation period, the family was also interviewed through an interpreter. Some parts of the communication (interviews and formal consultations) were recorded, while the remaining observations were documented through field notes.
The observations were focused on the communication, but in the communicative event they had an unstructured and exploratory character. During these sessions I participated as a passive observer (Minkler & Wallerstein, 2011), meaning that I did not actively take part in the caring or treating process although I did communicate with the participants at some points. An oft-discussed impact in this type of event is the observer's paradox, whereby observers may change the observed event simply by their presence (Labov, 1972, p. 209). Though the observer's paradox is not an ethical issue per se, in interpreted events even passive observers may find themselves in situations much more troubling that the one Labov described, for instance if they discover a mistake in the communication. The researcher may have to decide whether it lies in their discretionary power to rectify the miscommunication or not.
In the case described here, the challenges were not related to my relationship with the participants but to how my presence impacted the communication event or even the consultation. By observing a family over several days, I obtained a unique insight not only into the family's medical situation but also into their social situation. This unique perspective led to situations where two or even three norms conflicted with each other and impacted either the norm of being a passive observer or the ideal of health care on equal grounds. I felt that my ethical commitments as a researcher went against both my personal moral values as a fellow human being and my professional ethics as an interpreter. In this case, I would argue that the researcher's discretionary power lies in decisions on whether to intervene or not in different situations. The following example will serve as a case in point.
During a preparatory meeting between the medical staff (including the treating physician, nurse and nurse's assistant) and a patient and family with little proficiency in Swedish, the discussions revolved around certain test results and the further treatment plan. The patient was very young, and the family stayed in the hospital with the patient; and depending on whether the patient responded positively to the treatment, the patient and family would either be discharged quite soon or would have to stay at the hospital. As long as the patient had to stay, different test results would be coming in on different days, and interpreters would have to communicate these results to them, but as soon as the patient was discharged the interpreter-mediated communication of results would not be necessary. As interpreters represent a significant cost, there is a reluctance to book them in advance for fear of having to pay them in case they show up and the family is not present. In this example, the physician decided during the meeting that an interpreter would only be booked if the family was still at the ward when the results arrived. I was present at this meeting, and knowing the interpreting service industry as I did, I could foresee that when the results arrived, it would be too late to book an interpreter with the language combination in question. I also knew that if a booked interpreter is cancelled 24 hours before a scheduled appointment, no fee is charged.
The passive observer's ethical requirement to only observe and not interfere in any course of action conflicted with my personal moral values of patients being provided with the best possible care, as well as with my professional interpreting ethics of ensuring that information be transferred correctly. Is there any ethical wiggle room here where an active intervention, which will change the course of action of the treatment process, can be made for the greater good of the patient? Section 1.6 of the Swedish Patient Act of 2014 states that the goal of health care is to give the entire population access to good and equal heath care that respect the individual's dignity and rights. Section 1.8 also states that paediatric care should take the best interest of the child into consideration. I was not part of the health care team but decided to exercise my discretionary power and intervene to inform the physician about the conditions for booking an interpreter, as I did not see how the care would be in line with the Patient Act if an interpreter was not present. The immediate result was that the patient and parents received treatment in line with the act, and the intervention will also be clearly stated when I report my findings.
Case 3: whose voice is it anyway? Finding a voice through several layers of possible bias The final case discussed here will deal with the voice we hear in research reports. The study that will be the basis for this case was an interview study of deaf interpreters (Skaten, Urdal, & Tiselius, 2018). The aim of the study was to understand deaf students' experiences of taking an interpreting course together with hearing students. Interpreting programmes at university open to both hearing and deaf students are very rare, but are an important contribution to an inclusive society.
Deaf interpreters interpret between different sign languages and are also better than their hearing colleagues at understanding people who are less proficient in their given sign language. They are an important bridge between, for instance, public authorities and newly arrived immigrants with a different sign language than that of the host country (hearing sign language interpreters typically only master their own nation's sign language) or individuals with limited sign language proficiency due to disability.
Despite the importance of an inclusive society, deaf interpreters are rare, and the training of deaf interpreters is even rarer. As part of a project to develop a sign language interpreting training on equal grounds in Norway, we found it important to learn from the experiences of the students and teachers at one of the rare programmes where both hearing and deaf sign language interpreters are taught together as one group. Our aim was to use critical discourse analysis (Fairclough, 1992) to identify how language use revealed the deaf interpreters' construction of identity (Burr, 2003) and how language use also manifested different types of recognition (Honneth, 2005). The data consisted of in-depth interviews with deaf sign language interpreters. The interviewed interpreters used their mother tongue sign language, and the interviews were conducted using interpreters between the given sign language and spoken English, which was not the interpreters' first language combination. The interviews were then transcribed to written English and analyzed on that basis. The transcriber had proficiency in both sign language and English, but none of the two languages was the transcriber's first language.
There are several possible sources of error in this type of design, and it is necessary for the researcher to ask whose voice is represented in the data: the interviewee, the interpreter, the interviewer, the transcriber or the researcher. The researcher's discretionary powers in this type of design lies in controlling the different sources of error. Ashby (2011) argues that the researcher should go beyond the aim of 'giving voice' to somebody and instead facilitate 'voice and agency'. She goes on to argue that if the analysis and reporting of this type of qualitative, interview-based research is understood as facilitating agency, then it 'recognizes the interdependent and dynamic nature of voice and critical qualitative research' (Ashby, 2011, p. 13). Voice in this type of research is created through the process of research and is not something that was there from the beginning, only for the researcher to subsequently unveil and report it.
Our study sought not only to facilitate agency in the way Ashby describes it but also preserve the interviewee's voice in the material. First of all, the interviewers were also the researchers analysing the material, entailing that at least one of the researchers was present when the interviews were conducted and was thereby aware of the co-construction of meaning in the interview situation; indeed, on most occasions, two or three of the researcher-interviewers were present. Second, two of the three researchers were fluent in sign language, although not the same one as the interviewee, which would have been ideal. Nevertheless, the fact that they were fluent in a sign language meant that they could follow the interviewee at least superficially and thereby gain a first-hand impression of the voice of the interviewee. Third, the narratives of the main participants (i.e., the deaf interpreters/students) were backed up by narratives from their classmates and teachers, thereby allowing us to confirm our impressions. Finally, we also had the possibility to ask follow-up questions to the participants while doing the data analysis, though the participants unfortunately did not have time for us to share our post-analysis interpretations and conclusions with them and thereby further enhance our readings of their experience.
Linking this case back to the researcher's discretionary power, I believe that in this case such power concerns respecting what the interviewees said and the meaning of what was said. There are several layers of possible bias here, both our own pre-conceived opinions as well as the interpreters rendering of the information possibly also coloured by pre-conceived ideas. There is a delicate tension between respecting collegiality (i.e., assuming that the interpreter colleague, to the best of his or her knowledge, conveyed everything that was expressed) and following the researcher's ethical imperative to question their data and analyse them as thoroughly as possible. The discretionary power lies then in how the researcher decides what measures should be taken both before and during the analysis to make sure that the participant is understood.

Conclusion
This article has discussed ethics in interpreting research and the voice and discretionary power of interpreting researchers. By reviewing three cases from my own research practice, I have argued that understanding and exercising discretionary power is important for the interpreting researcher and that research ethics may at times be in conflict with both interpreting ethics and personal moral values. I have also aimed to show how the researcher should consciously reflect on how the participants' voice is represented in research reports and how to avoid having the researchers' own voice dominate. I have also argued that the focus on the researcher that Bendazzoli (2016) calls for is relevant both for research with experimental data and natural data. I hope thereby to have given food for thought on crucial areas of research ethics.

Disclosure statement
No potential conflict of interest was reported by the author.

Notes on contributor
Elisabet Tiselius has a PhD in interpreting, and is an active interpreter since 1996. She is a Swedish state authorized public service interpreter, accredited to the EU and member of AIIC. At Stockholm University she is focusing her research on cognitive aspects of dialogue interpreting (VR grant 2016-01118) within the SPRINT research group. Together with the research group on Childhood Cancer Health Care at Karolinska Institutet, she studies communication over language barriers in highly specialized cancer care (partly funded by Swedish Childhood Cancer foundation). In collaboration with the research group on Interpreter and Interpreter training at the Western Norway University of Applied Sciences, she investigates training of deaf interpreters. At Stockholm University she is Director Studies for interpreting and involved in the programs for public service, conference and signed language interpreting. She tweets @tulkur ORCID Elisabet Tiselius http://orcid.org/0000-0002-2285-6729