Introduction

1Ethnography has long rejected the fiction of the detached observer, recognizing that the researcher’s presence inevitably shapes the phenomena under study (Watson 2000; Pels 2014). Yet the practical implications of this epistemological stance remain undertheorized in multilingual clinical settings, where the researcher’s linguistic competence creates specific, materially consequential forms of involvement. When clinicians and patients do not share a language – or when shared repertoires are fragile – the researcher’s bilingual or plurilingual skills may be solicited, implicitly or explicitly, to clarify, translate, or gloss the meaning of a turn. This paper takes that moment seriously. It asks: what happens to the interaction, and to our methods, when the ethnographer herself becomes a linguistic intermediary? I approach this question through two, carefully analysed episodes recorded during long-term fieldwork in two third-sector clinics in Italy. Rather than treating the researcher’s intervention as “contamination” to be bracketed out, I frame it as analytically generative. These episodes afford a close look at how roles are negotiated, how accountability is redistributed, and how authority subtly shifts when a fourth voice – neither clinician nor patient nor official mediator – enters the pragmatic choreography of the consultation. In doing so, the paper reframes an all-too-familiar dilemma (Should I speak? Should I remain silent?) as a site where method, ethics, and discourse meet.

2The contribution is threefold. First, I offer a reflexive, linguistic-ethnographic account that centres role fluidity and positionality in the clinical encounter (Davies 2008; Emerson et al. 2011). The analysis tracks the emergence of the researcher’s voice, showing how footing and participation frameworks are rekeyed as soon as she speaks. Second, I draw on interpreting and mediation studies (Wadensjö 1998; Angelelli 2004; Gavioli, Baraldi 2011) to argue that the “researcher-as-intermediary” is not a mere stopgap for the absence of a professional interpreter, but a qualitatively distinct figure: one whose epistemic status, institutional mandate, and moral accountability differ from both clinicians and trained mediators. Third, I articulate the ethical and epistemic stakes of these moments through the lenses of cultural humility and epistemic injustice (Fricker 2007), asking how the researcher’s linguistic agency can support understanding without eclipsing patients’ voices.

3Empirically, the paper builds on an extended corpus of observations, audio recordings, and fieldnotes from outpatient services that cater to linguistically diverse populations. Within that corpus, I zoom in on two consultations – one from an outpatient clinic in Florence, one from a multidisciplinary clinic in Cosenza – in which my linguistic competence was directly mobilized by clinical staff. The analysis combines selective transcript excerpts with parallel fieldnote fragments, foregrounding the tension between what is said, how it is said, and how it is heard and ratified once the researcher speaks. Methodologically, this pairing makes visible the double inscription of the event: the interactional record and the researcher’s reflexive trace. Theoretically, I align with scholarship that understands healthcare talk as consequential practical action embedded in institutional asymmetries, yet I shift emphasis from macro accounts of power to the micro-mechanics of role hybridity and accountability in action (Roberts, Sarangi 2005; Rampton et al. 2015). The aim is not to provide a comprehensive model of interpreter-mediated care, but to interrogate a specific, under-discussed configuration: the ethically burdened, methodologically fertile moments in which researchers do language inside the encounter.

4The paper proceeds as follows. Section 2 integrates reflexive ethnography with interpreting studies. Section 3 details the field context and ethical stance. Section 4 presents the two vignettes and their analysis. Section 5 discusses implications for method and ethics, and Section 6 concludes by reframing these episodes as data rather than disturbance. Overall, I argue that the figure of the researcher-as-intermediary exposes the fragile equilibrium of clinical talk under conditions of linguistic diversity. It reveals how care, accountability, and knowledge are co-constructed not only by institutional actors and patients but also – at times – by researchers whose presence becomes audibly consequential. Recognizing, theorizing, and methodically documenting that consequence is essential if we are to move beyond the fiction of the “silent observer” and toward a more honest, rigorous account of multilingual care.

Theoretical Framework

5Ethnography has long abandoned the illusion of a detached, “fly on the wall” perspective. Rather, as Davies (2008) reminds us, ethnographic knowledge is always situated and partial, co-produced through the very presence of the researcher in the field. Emerson, Fretz and Shaw (2011) similarly stress that fieldnotes are not neutral recordings but interpretive renderings shaped by the ethnographer’s sensibilities, emotions, and embodied position in the scene. As Pels (2014) has argued in his historical approach to intersubjectivity in ethnography, the tension between objectivity and the researcher’s situatedness is not a problem to be solved but a productive condition to be theorized. Watson’s (2000) influential collection Being There further demonstrates that “being there” is never a passive state; fieldwork demands continuous negotiation of the researcher’s presence as both methodological instrument and moral agent. Reflexivity thus becomes not an optional add-on but a constitutive dimension of ethnographic practice.

6In multilingual healthcare contexts, reflexivity acquires an additional layer. The researcher is not only an embodied presence but also a linguistic resource, visible and audible to the participants. Even when silence is maintained, interlocutors are aware of the researcher’s potential capacity to understand, translate, or intervene. The researcher’s “non-participation” is therefore always conditional, framed against the possibility of participation. This echoes Goffman’s (1981) insight that participation statuses are multiple and fluid: one can be a ratified hearer, an overhearer, a bystander, or shift between these roles. For the ethnographer, these shifts are not simply analytical categories; they become lived dilemmas, demanding choices with methodological and ethical consequences. Reflexivity in this sense requires more than acknowledging one’s positionality; it requires recognizing how one’s linguistic agency can shape the very events being observed. The ethnographer’s body, accent, repertoire, and decisions to speak or remain silent all have material effects on how care is enacted and how power circulates in the consultation.

7Recent scholarship has begun to address the ethnographer’s linguistic positionality as a methodological concern in its own right, moving beyond general reflexivity to examine how language practices constitute the researcher’s stance in the field. Danero Iglesias and Gibb (2017) have drawn attention to a persistent disciplinary silence on this issue. Reflecting on their own ethnographic trajectories – Gibb’s fieldwork on anti-racism in France, Danero Iglesias’s on nationalism in Moldova – they acknowledge that their doctoral theses and publications were “effectively silent” about their experiences of language learning and its effects on the research. This silence, they argue, is not accidental but symptomatic of a norm that treats linguistic competence as a prerequisite rather than a topic of inquiry. Their call to break this silence has implications for how boundary-crossing moments are treated analytically: not as embarrassments to be suppressed but as data to be examined. Kosiek (2023) pushes this further through his fieldwork among Ukrainian minorities in Poland and Romania. His fluency in literary Ukrainian – a variant that local speakers found “foreign and incomprehensible” – inadvertently shaped the social dynamics of his research: interlocutors assumed he was Ukrainian, adjusted their linguistic behaviour in his presence, and occasionally provided data that his presence had itself elicited. His experience illustrates that the researcher’s linguistic competence does not simply provide access to a pre-existing reality but actively constitutes the interactional field, a point with direct relevance for healthcare settings where the ethnographer’s language skills may be solicited by clinicians.

8Brooks (2024) extends this line of inquiry by examining moments of “communicative opacity” in multilingual antenatal consultations. When languages exceeded her repertoire – Hindi during a diabetes consultation, Portuguese during a booking-in appointment – she shifted her analytical attention from verbal content to embodied and relational dimensions: gaze, touch, tone, gesture. Rather than treating these moments as failures, she reframed them through a capabilities approach and developed a strategy of post-fieldwork collaborative translation, involving multiple translators who sometimes disagreed on the meaning and register of what had been said. Her work underscores that the meaning of multilingual encounters is never self-evident but always co-constructed through successive layers of interpretation. Most recently, Rickert et al. (2026) have proposed a four-dimensional model of researcher positionality constructed through language practices – access, alignment, authority, and affect – demonstrating that the researcher’s linguistic choices shape not only access but also the epistemological terrain of the research itself. What these four contributions share is a recognition that the ethnographer’s linguistic positioning is not incidental to the research process but constitutive of it. Whether through silence, opacity, fluency, or improvised mediation, the researcher’s language practices shape what can be known, by whom, and on whose terms. This recognition has particular salience in healthcare settings, where linguistic competence carries immediate clinical consequences and where the line between observing care and participating in it may dissolve without warning.

9Scholarship on interpreter-mediated interaction provides crucial insights for understanding the dynamics at play when researchers intervene linguistically. Wadensjö’s (1998) seminal work conceptualizes interpreting not as mechanical transfer but as dialogic coordination: interpreters co-construct meaning, manage turn-taking, and shape interactional trajectories. Angelelli (2004) further argues against the myth of interpreter invisibility, showing how interpreters are always active participants whose decisions have moral and epistemic implications. Within healthcare settings, studies by Gavioli and Baraldi (2011) highlight how mediators perform complex relational work: negotiating trust, reframing cultural categories, and smoothing institutional asymmetries. These contributions underscore that mediation is not limited to lexical transfer; it involves alignment, face-work, and contextualization cues (Gumperz 1982). When the ethnographer temporarily assumes this role, the dynamics shift. Unlike professional interpreters, researchers are not institutionally sanctioned to speak on behalf of others. Their interventions lack the formal authority of contracted mediators, yet they carry the symbolic authority of academic expertise and the relational capital of participant observation. This hybridity produces a distinctive form of mediation: partial, contingent, and often improvised. Importantly, while interpreters are trained to balance fidelity and neutrality, researchers intervene from a position structured by their research aims, ethical commitments, and personal investments. Their speech cannot be bracketed as merely facilitative; it is saturated with the dual identity of observer and participant. The “researcher-as-intermediary” thus represents a qualitatively different figure, one that unsettles both ethnographic conventions and interpreter-mediated models.

10The blurring of roles is not unique to multilingual healthcare. Roberts and Sarangi (2005) describe how professional identities in institutional talk are continuously negotiated, often producing role hybridity where boundaries between categories are porous. Rampton et al. (2015), in their reflections on linguistic ethnography, likewise stress the need to theorize the messiness of roles rather than smoothing them over. For the ethnographer, role hybridity takes on a distinctive form: she is simultaneously witness, participant, potential ally, and occasional spokesperson. This multipositionality creates what can be called epistemic ambivalence. On the one hand, speaking can facilitate understanding, support care, and prevent communicative breakdown. On the other, it risks silencing the patient’s voice, skewing the data, and raising questions of legitimacy. These tensions resonate with Fricker’s (2007) concept of epistemic injustice: moments where certain voices are systematically undervalued or over-interpreted. When the researcher speaks on behalf of the patient, however well-intentioned, there is a risk of testimonial injustice – of filtering or reframing the patient’s account through the researcher’s linguistic and cultural lens. At the same time, silence may perpetuate hermeneutical injustice, leaving the patient’s experience unrecognized within the institutional frame.

11The notion of cultural humility (Tervalon, Murray-García 1998) offers a partial remedy, emphasizing openness, self-critique, and the recognition of power differentials in cross-cultural encounters. For researchers, cultural humility entails acknowledging the impossibility of neutrality and the necessity of reflexive accountability when linguistic agency becomes consequential. Methodologically, this means treating moments of researcher intervention not as errors but as data. Rather than erasing them in transcription or relegating them to footnotes, they can be analysed as sites where power, knowledge, and care intersect. This requires a shift in analytic stance: from a search for “pure” interactions uncontaminated by the researcher’s presence, to an acknowledgment that such purity is illusory. The ethnographer’s interventions are part of the ecology of the field and thus integral to the story that needs to be told.

12Taken together, these strands – reflexivity, interpreter-mediated communication, and role hybridity – provide the conceptual scaffolding for the analysis that follows. They direct attention to the researcher’s audible presence as both an ethical challenge and a methodological opportunity. They also anchor the argument that the figure of the “researcher-as-intermediary” deserves theoretical recognition in its own right, rather than being subsumed under categories of bias, contamination, or error.

Methodology

14Unlike large-scale surveys or interviews, the methodological emphasis here is on the interactional fabric of clinical encounters: the fine-grained turns of talk, the embodied cues, and the emergent positioning of participants. This makes discourse analysis and linguistic ethnography particularly apt tools. At the same time, it requires grappling with the unavoidable presence of the researcher as an active element in the setting. The full corpus encompasses dozens of consultations, varying in length from brief follow-up visits to extended diagnostic interviews. While the dataset includes examples of communication in a shared language (Italian or English), as well as encounters mediated by professional cultural mediators, this article narrows in on two configurations in which I was solicited to intervene linguistically: one sudden and unauthorized, one gradual and semi-institutionalized.

15These episodes were selected not because they are representative of all encounters, but because each crystallizes a distinct form of a recurring methodological dilemma. They are analytically rich precisely because they destabilize the convention of researcher invisibility, foregrounding the blurred boundaries between observation and participation. The data presented consist of extracts from the audio recording, transcribed to capture turn-taking and relevant linguistic features; parallel fieldnotes, written immediately after the consultation, which include reflections on emotions, hesitations, and ethical doubts; contextual observations, situating the episode within the broader ecology of the clinic and its routines. By triangulating these layers, the analysis demonstrates how the same event can be inscribed both as interactional sequence and as reflexive narrative, each carrying distinct epistemic value. A further layer of documentation consists of my ‘translation notes’: brief records written immediately after consultations in which I had intervened linguistically, documenting specific word choices, the simplifications introduced and the doubts carried about the adequacy of my renderings. These notes differ from fieldnotes in their focus on the micro-decisions of the translation act itself, rather than on the broader interactional or institutional context. My position in the field was shaped by several intersecting factors. As a PhD candidate affiliated with an Italian university, I was introduced to the clinics through formal agreements with their managing organizations. My role was officially described as that of an observer conducting a study on multilingual communication in healthcare. This institutional framing granted me legitimacy to attend consultations but did not authorize me to act as interpreter or mediator. It is important to explicate the institutional and ethical framework that governed my presence in the field. Participation was explicit and negotiated with the clinic coordinators and all professionals involved, through informed consent protocols approved by the ethics committees of the hosting institutions. The terms of my access were established through direct negotiation with clinic management rather than through formal written agreements, and my role was understood as primarily observational, attending waiting areas and consultations without formally participating in clinical or communicative exchanges.

16Yet my positioning within the clinics was, in practice, fluid and contextually negotiated. Particularly in the Cosenza clinics, where understaffing was more pronounced, I was occasionally drawn into practical tasks: administrative assistance, accompanying patients to diagnostic services, and, in some instances, providing ad hoc linguistic assistance when no mediator was available. This occasional involvement in a translational role was not sanctioned by any explicit protocol, nor was I prepared with specific training on how to manage the boundaries of my researcher role when the observational frame was challenged by the urgency of a clinical situation. This gap between a formally observational mandate and the interactional contingencies of an under-resourced field setting, is precisely part of what the present article aims to address, by examining what happens when the protocols that define the researcher’s role prove insufficient in the face of lived clinical encounter.

17Linguistically, however, my repertoire included Italian (as native speaker), English, and conversational proficiency in French and Spanish. Staff and patients were aware of this, and occasionally addressed me directly with questions about what had just been said. My physical presence in the consultation room – sitting slightly apart from doctor and patient – thus carried a latent expectation: that I could, if needed, facilitate understanding. This dual positioning created a zone of potential audibility. Even when silent, I was audibly available. When I did speak, I crossed a boundary: from observer to participant, from scribe to co-constructor. The methodological challenge was not simply to manage this crossing in practice but to decide how to treat it analytically.

18Yet the episode under discussion illustrates the limits of formal consent procedures: no protocol could anticipate the moment when the physician turned to me with an expectant glance. Ethical decision-making was situational, and writing fieldnotes immediately after became crucial for documenting the ethical tensions experienced. Given these complexities, I adopt an analytic stance that treats the researcher’s interventions as data in their own right. Rather than editing them out of transcripts or relegating them to footnotes, I examine how the researcher’s speech reshapes participation frameworks, footing, and epistemic authority. This resonates with calls in linguistic ethnography to embrace the messiness of fieldwork (Rampton et al. 2015) and to theorize the researcher’s role as part of the interactional ecology.

19The analysis therefore proceeds in two registers. On the one hand, it applies tools of discourse analysis to the transcript: turn allocation, repair sequences, code-switching, and stance-taking. On the other hand, it draws from reflexive ethnography to interpret the researcher’s decisions, hesitations, and post-hoc reflections. Juxtaposing these registers illuminates the tensions between the interactional event and the researcher’s embodied experience. Focusing on two vignettes inevitably narrows the scope of generalization. The analysis does not claim that all ethnographers in multilingual healthcare settings will face identical dynamics. Nor does it offer a prescriptive model of how researchers should behave. Instead, the aim is heuristic: to foreground a type of situation often relegated to the margins of field reports, and to argue for its analytic and methodological significance.

20Moreover, while the paper highlights the researcher’s agency, it recognizes the risk of over-centring the researcher’s voice. Care has been taken to balance this by situating the episode within the broader institutional context and by maintaining attention to the patient’s and physician’s perspectives as captured in the transcript. Methodologically, then, this study operates at the intersection of discourse analysis and reflexive ethnography. It builds on a corpus of multilingual consultations but zooms in on two, ethically charged episodes where the researcher’s linguistic intervention became consequential. The decision to treat these interventions as data, rather than contamination, is both a methodological choice and a theoretical argument: that ethnographic presence is never silent, and that the audibility of the researcher deserves analytic recognition.

Data Analysis: The Researcher as Intermediary

21The episode I want to focus on unfolded during a psychiatric consultation at the Florence outpatient clinic. The patient, a man in his early thirties from West Africa, spoke French and had only very limited Italian. He was undergoing treatment for trauma related to migration and had been prescribed psychiatric medication intended to stabilize anxiety and insomnia. At the time of the consultation, he reported experiencing persistent tingling sensations in his limbs. For the psychiatrist, the clinical issue was crucial: if the tingling had begun after the introduction of the medication, it could indicate adverse side effects that required adjusting the dosage or changing the drug altogether. If, however, the symptoms had preceded the prescription, they would point to a different causal chain and the need for neurological investigation. Establishing the temporal relationship between the onset of symptoms and the pharmacological treatment was therefore at the centre of the diagnostic process.

22The communicative resources available in the room were severely limited. The patient did not have the Italian necessary for fine-grained temporal distinctions, while the psychiatrist had no French. There was no interpreter present that day and no professional mediator could be summoned in time. After several unsuccessful attempts to pin down the chronology, the psychiatrist turned toward me – until then a silent observer – and asked whether I could help.

23What followed was a hesitant attempt at mediation. I addressed the patient in slow, broken French, supplementing my words with Google Translate on my phone. The transcript below captures the improvised nature of this exchange:

Doctor (to Researcher): Senti fa’ ’na cosa, parli francese spedita tu?
Listen, do you speak fluent French?
Researcher (to Patient, hesitant French): Vous pouvez parler lentement?
Can you speak slowly?
Doctor: Puoi chiedergli se prende ancora le medicine?
Can you ask him if he’s still taking the medication?
Researcher (to Patient): Vous prenez encore… les médicaments?
Do you still take the medicines?
Patient: Oui, mais parfois j’oublie. C’est pas facile.
Yes, but sometimes I forget. It is not easy.
Researcher (to Doctor): Dice di sì, ma che ogni tanto si dimentica.
He says yes, but sometimes he forgets.
Doctor: E chiedigli se ci sono degli effetti collaterali…
And ask him if he has any side effects…
Researcher (to Patient): Est-ce qu’il y a… effets secondaires?
Are there… side effects?

24At this point, the patient’s response was incomplete, accompanied by gestures and then by silence. The uncertainty of the exchange became increasingly evident. Eventually the coordinator of the clinic intervened, calling the reception centre where the patient resided to ask staff for supplementary information. Through this external inquiry, it was confirmed that the tingling symptoms had predated the psychiatric treatment and were already under specialist investigation. The psychiatrist therefore concluded that the symptoms were unlikely to be medication-related, but the difficulty of reaching that conclusion highlighted the fragility of diagnostic reasoning under conditions of constrained language access.

25Although the sequence was short, it condensed a number of dynamics. My participation was not planned; it arose under pressure, in response to communicative deadlock. The psychiatrist’s request already implied ambivalence, signalling both the urgency of the clinical task and an awareness of the inadequacy of my linguistic competence. The mediation that followed was fragmented and partial. My French was hesitant, my phrasing simplified, and my reliance on digital translation tools made the exchange halting and imprecise. The interaction revealed gaps that neither patient nor clinician could fully overcome. Yet despite its shortcomings, my intervention temporarily redistributed authority in the consultation. By selecting fragments, condensing responses, and reporting them into Italian, I provided the psychiatrist with information that he accepted as reliable and on which he began to act.

26The tension between what was happening externally and what I felt internally became clear in my fieldnotes written immediately afterward. I recorded the sense of sudden responsibility when the doctor turned to me, the impossibility of refusing, the patient’s expectant gaze, and the doctor’s urgency. I described how I opened Google Translate and how each phrase felt clumsy, accompanied by worry that I was oversimplifying and missing nuance. I also noted the relief I felt when the coordinator eventually sought confirmation elsewhere, relief tinged with embarrassment and doubt about whether my intervention had actually been useful or had only added confusion. From an interactional perspective, my footing shifted rapidly. I moved from bystander to animator of the psychiatrist’s questions and the patient’s replies. Yet I was not a neutral conduit. Every utterance entailed selection and adaptation. Asking the patient “Vous prenez encore les médicaments?” was already a simplification of the doctor’s phrasing, and my Italian report that “he sometimes forgets” condensed the patient’s more elaborate expression of difficulty. This oscillation between relaying and coordinating is precisely what Wadensjö (1998) identifies as central to dialogic interpreting, though in my case it occurred without the professional norms or training that usually guide interpreters. This improvisation epitomizes role hybridity. I was simultaneously researcher, ad hoc interpreter, and uneasy participant. My words carried authority in the eyes of both doctor and patient, yet for me they represented a responsibility taken on reluctantly and with hesitation. The eventual recourse to external networks underscored the limits of this hybridity, showing how my mediation was a stopgap measure that could not sustain the epistemic demands of clinical reasoning. The ethical stakes of the episode were unmistakable. By speaking, I risked testimonial injustice (Fricker 2007) filtering the patient’s testimony through my limited French and losing nuances in the process. Had I remained silent, however, I might have contributed to hermeneutical injustice, leaving his experience uninterpreted and possibly leading to a misattribution of symptoms. Both options carried risks; neutrality was not possible. What was required instead was accountability, and the recognition that any choice – whether to intervene or not – had consequences for both patient care and knowledge production.

27The presence of Google Translate added another layer to the complexity. Digital translation provided a form of support, but also introduced its own opacity. It made my role not only that of a linguistic broker but also of a technological mediator, further complicating the interaction and heightening the sense of precariousness. What ultimately became clear was that this episode was not an isolated mishap but a manifestation of systemic fragility. The absence of a professional interpreter was not accidental but structural, tied to resource constraints and institutional priorities. My intervention was symptomatic of these gaps, a temporary patch in a fragile communicative infrastructure. The fact that the clinic had to draw on external staff to establish the timeline of symptoms demonstrates how diagnostic reasoning regularly exceeds the immediate bounds of the consultation, relying on dispersed networks to stabilize meaning.

28It is important, however, to balance this critical reading of the episode with an acknowledgment of its productive dimensions. The patient’s response to being addressed in a language closer to his own – however imperfectly spoken – was immediate and visible: he leaned forward, made eye contact, and began to engage more actively with the consultation. His willingness to share information about medication adherence, however fragmentary, suggested a momentary sense of being heard that the preceding monolingual exchange had failed to produce. The improvised mediation, despite its manifest limitations, temporarily bridged a communicative gap that institutional structures had not addressed, generating a transient but meaningful space of trust and therapeutic alliance. This positive dimension must be weighed against the risks. As Sayad (2002) warned in his analysis of nurses serving as “vulgarizers of culture” in French hospital settings for North African migrants, non-professional linguistic intermediaries risk reducing complex illness narratives to simplified cultural glosses, instrumentalizing cultural knowledge in ways that serve institutional efficiency rather than patient understanding. The episode thus carries a double valence: on one hand, it shows how improvised mediation can open communicative pathways that formal structures have foreclosed; on the other, it illustrates the risk that such mediation, precisely because it operates outside professional frameworks, may inadvertently flatten the patient’s experience to fit biomedical categories. Both readings – the generative and the reductive – must be held in productive tension if we are to understand the full complexity of these moments.

29In this light, the vignette reveals that the ethnographer’s role in multilingual healthcare cannot be understood as purely observational. When communication falters, the researcher may be drawn into the interaction, becoming an audible and consequential participant. Such interventions are halting, ethically fraught, and epistemically significant. They reshape participation frameworks, redistribute authority, and influence clinical outcomes, while also producing discomfort and self-doubt for the researcher. Far from being treated as methodological contamination, these moments can be reframed as data. They illuminate the fragility of multilingual care, the instability of ethnographic roles, and the systemic challenges of language access in healthcare. In short, they demonstrate that ethnography is always already participatory, even when participation is unplanned, reluctant, or inadequate.

A contrasting configuration: the researcher as ongoing linguistic support

30The Florence episode portrays the researcher’s linguistic intervention as a sudden, unplanned rupture of the observational frame. A contrasting episode from the Cosenza outpatient clinic, recorded during a psychological consultation in December 2023, illustrates a different configuration of the same dynamics – one in which the researcher’s role as linguistic intermediary was not a punctual emergency but a more continuous, integrated presence. Juxtaposing the two episodes reveals that the researcher’s audibility is not an anomaly but a structural feature of multilingual clinical settings.

The consultation took place in mid-December 2023 at the Cosenza outpatient clinic, during a follow-up visit with the psychologist. The patient, whom I call Dev, was a man in his late twenties from South Asia – India, specifically – who spoke English with moderate fluency but very limited Italian. He had been referred to the psychological service for trauma-related symptoms: persistent headaches, chronic fatigue, and severe insomnia. These complaints had intensified since his arrival in Italy, where he lived alone while his wife and young son remained in India. The circumstances of his departure were tied to what he described, in fragmented terms, as “political problems.”
I was present in the room in a role that had, by then, become semi-institutionalized: the psychologist knew I spoke English and had, over successive sessions, come to rely on my presence as a communicative bridge. I was not formally designated as an interpreter, yet my function had quietly shifted from observer to what might be called “support interpreter” – paraphrasing, reformulating in simplified English or Italian, filling the gaps that opened whenever Dev’s words faltered between languages.
During this session, the psychologist attempted to explore the cultural meanings of his symptoms. She asked, gently: “In your country, when a person has headaches or feels very tired, how is that usually explained? What is the cause?” There was a long pause. Dev looked down, then replied in halting English interspersed with fragments of his own language: “In my country… sometimes doctor say… because of… thinking too much.” I repeated his words softly, partly to confirm I had understood, partly to make them audible to the psychologist, who nodded and waited.
The psychologist then asked about his family – his wife and son in India – and about fear: whether the symptoms had started before or after “the political problems.” Dev’s answer came as a barely articulate admission: “You know this feel and the danger” – a fear without a precise object, a fear that had settled in the body. The psychologist, drawing the threads together, offered a reframing: “I understand better now why some days you have more headaches or less energy. It is the worry for them.” Dev nodded slowly, and in the silence that followed, something shifted in the room: not resolution, but recognition.
Afterward, writing my fieldnotes, I found myself returning to the image of a language – broken between memory and care. Each of Dev’s hesitations had been a site where the body translated what words could not say, and each of my paraphrases had been an attempt, however imperfect, to hold that meaning open long enough for it to be received (Fieldnotes and translation notes, 18 December 2023).

31Unlike the Florence consultation, where my involvement was sudden and reluctant, here my role as linguistic support had become a normalized part of the clinical encounter. I was not called upon to resolve a crisis but functioned as a continuous communicative interface: paraphrasing, reformulating, bridging the gaps between languages. This normalization, however, did not eliminate the ethical and epistemic tensions identified in the first vignette. Each paraphrase involved selection; each reformulation shaped what became visible to the clinician. The choice to render “thinking too much” as a culturally specific idiom of distress rather than translating it into biomedical terms (e.g., “rumination,” “anxiety”) was itself an epistemic act, one that preserved – at least partially – the patient’s own framework of understanding.

32The expression “thinking too much” is well documented in cross-cultural psychiatry as an idiom of distress that bridges somatic and psychological registers across diverse cultural contexts (Kirmayer 1996; Kleinman 1988). In Dev’s utterance, the headache is not merely a biomedical symptom but a culturally embedded category of suffering, linked to moral causation and embodied memory. The psychologist, to her credit, received this formulation as part of the therapeutic process rather than reducing it to a diagnostic checklist. Yet the passage from the patient’s fragmented multilingual narrative to the psychologist’s interpretive reframing (“It is the worry for them”) involved a chain of translations – linguistic, cultural, epistemic – in which my voice played a constitutive role. As I noted in my fieldnotes, each hesitation, each word left suspended, became a site of shared interpretation between patient, therapist, and researcher. The meaning was not produced by a single voice but through the interaction of voices that pursued, corrected, and translated one another.

33Taken together, the two episodes – one from Florence, one from Cosenza – reveal that the researcher’s linguistic involvement in multilingual healthcare is not a single configuration but a spectrum. At one end, the sudden, unauthorized intervention in an emergency communicative breakdown; at the other, the gradual, semi-institutionalized absorption of the researcher into the clinic’s communicative infrastructure. Both configurations raise the same fundamental questions about epistemic authority, about whose voice is amplified and whose is filtered, and about the conditions under which alternative epistemologies of illness can find expression within biomedical settings. The difference lies in visibility: in the Florence episode, the crossing of the observational boundary was dramatic and uncomfortable; in the Cosenza episode, it had become so habitual that its consequences risked going unexamined. It is precisely this normalization that makes the second vignette analytically significant. When the researcher’s mediation becomes routine, the epistemic choices embedded in every paraphrase and reformulation may escape reflexive scrutiny – unless, as this article argues, they are systematically documented and treated as data.

Discussion

34The two episodes I have described may appear minor at first glance: a few turns of talk in which I reluctantly mediated between psychiatrist and patient, followed by the intervention of the clinic coordinator to stabilize the diagnostic process. Yet the brevity of the exchange should not obscure its significance. In those short minutes, the assumptions that underlie ethnographic presence, the fragility of clinical reasoning, and the ethics of participation all came into sharp relief. What unfolded in the consultation room was not simply a pragmatic solution to a communicative problem, but a moment that exposed the methodological and epistemic stakes of doing ethnography in multilingual healthcare.

35One of the most striking lessons of this vignette is the inadequacy of the enduring figure of the silent observer. Ethnography has long problematized the myth of detachment – a critique now well established in the discipline (Watson 2000; Pels 2014) – but in multilingual contexts, the specific ways in which this fiction collapses take on distinctive, materially consequential forms. My mere presence in the room was already audible in potential: both the doctor and the patient perceived me as someone who might intervene if needed. When the communicative impasse became insurmountable, this latent potential was activated. In the episode from Cosenza, this potential was activated more gradually, absorbed into the routine of successive consultations until it became structural. The neutrality I had attempted to maintain dissolved in an instant, replaced by an active role that changed the course of the encounter. The idea that ethnographers can simply “blend into the background” is untenable in such contexts, where linguistic competence itself is a resource that others may mobilize.

36This raises the question of how to conceptualize the role the researcher assumes in such moments. Interpreters, as Wadensjö (1998) and Angelelli (2004) have shown, coordinate talk and bear responsibility for communication within a framework of training and institutional mandate. My role was improvised and unauthorized, yet carried the symbolic weight of academic expertise. For the doctor, my words had the authority of linguistic access; for the patient, they offered a fragile bridge to being heard. What I embodied was a hybrid role that cannot be neatly assimilated to existing categories. The hybridity was not only institutional but epistemic. As Roberts and Sarangi (2005) note, professional roles are often negotiated and fluid, shaped by interactional contingencies rather than fixed scripts. My speech – whether in the clinic in Florence or in the successive sessions in Cosenza – carried information on which the psychiatrist began to base clinical reasoning, yet it was simultaneously marked by doubt and inadequacy on my part. For me, every phrase felt like a compromise, an unstable attempt to balance fidelity to the patient’s words with the urgency of providing the doctor with something usable. The dissonance between the apparent clarity of my translated statements and the felt uncertainty of their production highlights the gap between outward function and inward reflexivity.

37Ethical dilemmas were deeply entangled in this moment. By speaking, I risked what Fricker (2007) has called testimonial injustice: filtering the patient’s voice through my interpretive lens. Yet silence would have perpetuated hermeneutical injustice, leaving his perspective opaque to the institutional framework. What was required was not neutrality but accountability: the recognition that these moments shape both clinical outcomes and ethnographic knowledge.

38The use of Google Translate adds another layer. It demonstrates the creativity with which participants improvise under pressure, yet reveals the precariousness of relying on digital tools without guaranteeing nuance. This hybridity of human and digital mediation further complicated my position, reminding us that contemporary multilingual encounters are rarely shaped by human interlocutors alone.

39Beyond the interactional and ethical dimensions, both episodes underscore the systemic conditions that make such improvisation necessary. The absence of a professional interpreter was not incidental. It was a structural feature of the setting, reflecting the underfunding and marginal position of third-sector clinics that must constantly compensate for the shortcomings of the national healthcare system. My intervention was symptomatic of these structural gaps: a temporary patch in an infrastructure that routinely fails to provide stable access to language support. The fact that the clinic eventually relied on external staff at the reception centre to clarify the chronology of symptoms illustrates how fragile and distributed the process of diagnostic reasoning can be in these contexts. Meaning is not stabilized within the walls of the consultation room but through networks that extend across institutions and actors.

40The dynamics observed in this episode resonate with an emerging body of literature that examines the ethnographer’s linguistic stance as a constitutive dimension of fieldwork. Rickert et al. (2026) identify four dimensions through which language practices construct researcher positionality – access, alignment, authority, and affect – all of which were activated in the psychiatric consultation described here: my French provided access to the patient’s account, my translation aligned his words with biomedical categories, my researcher status lent authority to the reported information, and the affective dimension of the encounter – my discomfort, the patient’s relief – shaped the interactional dynamics.

41The relevance of these frameworks to the present data is direct. Unlike Brooks’s (2024) antenatal consultations, where communicative opacity could be revisited through post-fieldwork collaborative translation, the clinical setting analysed here afforded no opportunity for post-hoc verification – the consequences of my translations were immediate and irreversible. And unlike the disciplinary silence that Danero Iglesias and Gibb (2017) describe in their own earlier work, this article treats the researcher’s linguistic interventions not as background noise but as foreground data. Kosiek’s (2023) observation that the researcher’s linguistic competence actively constitutes the interactional field finds its clinical counterpart here: my presence as a potential linguistic resource altered the consultation before I ever spoke, and once I did speak, each paraphrase became a consequential act of epistemic selection. For medical anthropology more broadly, the two vignettes complicate the provider-patient binary. Care emerges here as the product of a wider ecology that includes researchers, technological tools, and institutional networks, through which power and authority circulate in unpredictable ways.

42The episode also points beyond communication dynamics toward the epistemological underpinnings of medical practice. Biomedical reasoning presupposes linear, standardized accounts of symptom onset, and by condensing the patient’s words, I inadvertently aligned them with biomedical expectations, facilitating care but narrowing epistemic possibilities. This points to a dimension of the encounter that the available ethnographic material can only partially illuminate: the epistemological contestation embedded in care systems that operate across radically different frameworks of illness. We cannot fully know what other epistemologies of illness the West African patient might have drawn upon had the communicative conditions permitted – what somatic metaphors, relational explanations, or spiritual frameworks might have surfaced had a more sustained and culturally informed dialogue been possible. Yet the second vignette offers a glimpse of what such a dialogue might look like. When Dev described his headaches as caused by “thinking too much,” he invoked an idiom of distress that operates across somatic and moral registers, one that cannot be adequately captured by biomedical categories such as “anxiety” or “rumination.” The psychologist’s willingness to receive this formulation as therapeutically meaningful – rather than translating it away – depended in part on the researcher’s mediation, which preserved (however imperfectly) the cultural texture of the patient’s account. This suggests that the epistemic gap exposed by the Florence episode is not merely a matter of missing linguistic resources but of missing cultural-epistemological competencies. A medical anthropologist with area-specific knowledge – of West African healing traditions in the first case, of South Asian idioms of distress in the second – could have mobilized different interpretive frameworks, not to replace biomedical reasoning but to create what Tareau et al. (2024) describe as a “personalized approach to health mediation,” where ethnobotanical, spiritual, and relational dimensions of illness are integrated into the clinical encounter rather than filtered out by translation. This is not an argument for the professionalization of anthropology in the abstract, but an observation grounded in the specific failures and partial successes documented here: when the researcher’s competence falls short, it reveals what a more sustained anthropological presence might accomplish. The implication is that interdisciplinary collaboration between clinicians and anthropologists with relevant cultural expertise should not be conceived as an ideal but as a practical response to the epistemic fragility that this article has documented.

Conclusion

43The analysis of these two vignettes has shown how quickly the boundaries between observation and participation can collapse when ethnographers find themselves in multilingual healthcare settings. What began, in Florence, as a consultation aimed at clarifying the side effects of a psychiatric medication turned into a moment in which my presence became audibly consequential. The psychiatrist’s request, the patient’s expectant gaze, and the lack of institutional linguistic support drew me into the communicative infrastructure of care. My hesitant French, my reliance on Google Translate, and my improvised translation choices all became part of the diagnostic process. That the figure of the “silent observer” is a methodological fiction has been well established in ethnographic theory (Watson 2000; Pels 2014). The contribution of this article lies not in restating that insight but in showing how this fiction collapses in specific, materially consequential ways in multilingual healthcare settings. Ethnographers are not invisible in the field; they are perceived, addressed, and sometimes relied upon as interlocutors. In multilingual contexts, linguistic competence – or even the assumption of competence – becomes a resource that others may mobilize, and refusing to respond is not a neutral act but a choice with consequences. The ethnographer’s audibility, whether exercised or withheld, shapes what unfolds in the consultation room.

44At the same time, the episode underscores that when researchers become intermediaries, they occupy a role that is neither equivalent to that of a professional interpreter nor reducible to that of a companion or bystander. The authority attached to their words derives from a hybrid positionality: unauthorized yet trusted, hesitant yet treated as reliable, marginal yet central in moments of communicative breakdown. This hybridity is not just institutional but also epistemic, as the researcher’s words contribute to the production of clinical knowledge even as they are marked by doubt and inadequacy. Ethically, the vignette highlights the impossibility of neutrality. Speaking risks filtering the patient’s voice, while silence risks leaving it unheard. Both paths carry dangers of epistemic injustice. What is required, therefore, is accountability: a reflexive awareness that our choices in these moments have implications not only for research but also for care. Documenting these choices through transcripts and fieldnotes, and analysing them as data rather than as contaminations, is a way of honouring that accountability.

45Perhaps most importantly, both episodes demonstrate that individual improvisation takes place within systemic conditions of scarcity, reflecting the precarious infrastructures of language access in healthcare. It is important to acknowledge that this analysis rests on two vignettes drawn from different clinical settings and does not claim to represent all multilingual healthcare encounters. The heuristic value of the episodes lies precisely in their detailed, reflexive analysis of critical moments: moments that condense dynamics often dispersed across longer stretches of fieldwork. As Pels (2014) has argued, the asymmetrical epistemological breaks that characterize cross-cultural research are not obstacles to be overcome but conditions to be theorized; the vignette analysed here offers one such theorization, grounded in the specificity of clinical practice. Ingleby’s (2012) work on health literacy as a moral task further reinforces the point that access to care is never merely a technical problem of language matching but involves deeper questions about whose knowledge counts and on what terms. The structural conditions documented in this article – the absence of professional interpreters, the reliance on improvised mediation, the distributed nature of diagnostic reasoning – are not anomalies but features of a healthcare landscape shaped by what Ingleby (2009) has called the politics of care for migrant populations.

46These findings carry implications for how ethnographic research in multilingual healthcare is designed, conducted, and documented. The Florence episode revealed that formal research protocols, however carefully drafted, cannot anticipate the moment when a clinician’s expectant gaze transforms the observer into a participant. The Cosenza episode showed that this transformation can also occur gradually, without a dramatic threshold, making it even harder to recognize and document. Together, they suggest that the question is not whether the ethnographer’s linguistic agency will become consequential in such settings, but when and how. This recognition should reshape the way researchers are prepared for fieldwork. Rather than treating role boundary infractions as aberrations to be avoided, training programs should equip ethnographers with strategies for navigating them reflexively – including explicit discussion of the epistemic consequences of translation choices, the ethics of speaking for others, and the analytical value of documenting one’s own discomfort and uncertainty. At the level of research design, protocols should build in provisions for systematic documentation of moments of researcher linguistic involvement, treating them not as deviations from the plan but as anticipated features of the interactional ecology. Kosiek’s (2023) practice of maintaining separate “headnotes” on the effects of his linguistic presence, and Brooks’s (2024) use of collaborative post-hoc translation to interrogate her own field recordings, offer concrete models for this kind of reflexive documentation. Finally, the analysis points toward a reconfiguration of the institutional relationship between universities and healthcare settings. The improvised mediation documented here arose from a structural absence – the lack of professional interpreters – that the researcher’s presence temporarily and inadequately filled. A more sustainable model would embed anthropologists with relevant cultural and linguistic competencies within care teams, not as extractive observers but as collaborative participants whose epistemic contributions are recognized and integrated. This would move beyond the individual ethnographer’s ad hoc choices toward a systemic response to the communicative fragility of multilingual care.