Contextual factors as an analytical tool: Exploring collaboration and negotiation in mental health interviews in prisons mediated by non-professional interpreters

This article proposes a framework for analysing interpreted events mediated by nonprofessionals. It is based on an examination of individual contextual factors rather than on traditional definitions of setting-based features. This approach promises to be more productive for the study of non-professional interpreting and for analysing contexts that do not fit into existing categories of setting. For these purposes, this article examines a corpus of 26 prison-based mental health interviews mediated by non-professional interpreters in order to analyse the collaboration and negotiation processes that emerge among the members of the communicative triad. First, it outlines contextual factors from a conceptual perspective. Second, it describes those contextual factors that are most relevant to analysing collaboration and negotiation processes. Finally, it describes the context of prison-based mental health interviews through the lens of these factors and examines their influence on specific instances of collaboration and negotiation extracted from this corpus.


Introduction
Framing analyses of interpreter-mediated interaction around the concept of setting raises legitimate questions. While consistencies in a setting allow for the creation of frameworks that facilitate analyses (e.g., conference vs public service interpreting), they can also limit our ability to develop theoretical constructs based on similarities across and differences within established settings. These issues become more relevant when we examine non-professional interpreting. Non-professional interpreters' (NPIs) lack of access to training and professional socialization opportunities (traditionally structured around settings) results in a lack of exposure to setting-based norms and expectations. Accordingly, NPIs' behaviour would more likely be conditioned by the interaction at hand and their previous experiences than by the setting itself.
Consequently, this article proposes and applies a framework for analysing interpretermediated events that is grounded in an examination of contextual factors, which offer a more granular view than setting-based analyses. For these purposes, I first outline contextual factors from a conceptual perspective, comparing and systematizing existing relevant typologies. However, given the breadth of potentially influencing contextual factors, any framework based on this approach has to be narrowed down to the specific issue to be examined (Alexieva, 1997). Therefore, second, I describe in greater detail those contextual factors that seem most relevant to the study of the active participation of all the members of the communicative triad in co-constructing interpreter-mediated interactions. Specifically, I investigate collaboration processes (i.e., participants supporting each other in meaningmaking, conversation management, goal achievement and relationship-building) and negotiation processes (i.e., participants challenging each other's interactional decisions and finding common ground to prevent a breakdown of communication). Finally, I use this framework of selected contextual factors to describe the context of mental health interviews conducted in prison and to examine their influence on specific instances of collaboration and negotiation extracted from a corpus of 26 NPI-mediated interviews of this type.

Limitations of setting-centred analyses
Scientific enquiry requires categorization to delimit our object of study, minimize confounding factors and facilitate generalizations. In Interpreting Studies, categories were originally created around interpreting modes and soon evolved to centre on the concept of setting. This was particularly meaningful as public service interpreting emerged as a branch of interpreting in its own right, as distinct from conference interpreting. This distinction was based on their contrasting contextual factors, mainly interpreting mode, types of participant, interaction topic, social and institutional constraints and communication goals (e.g., Abril Martí, 2006, pp. 35-86). However, as public service interpreting developed, it continued to branch out into separate subdisciplines and areas of professional practice marked by setting as the defining factor.
Whereas the surge in subdisciplines attests to a growing and diverse field, it also encourages a somewhat divisive view. Early researchers (Gentile et al., 1996;Gentile, 1997;Mikkelson, 1999;Roberts, 1997) argued that the exchange of ideas needed to develop strong theoretical concepts may be hindered by this division into setting-based, siloed compartments -a position that continues to draw support (Downie, 2021;Grbić & Pöllabauer, 2006). It is unarguable that the concept of setting provides a framework that facilitates analysis by assuming consistencies within that setting. However, these divisions may be imposing implicit, unconscious boundaries in our research (e.g., limiting literature reviews), and explicit, conscious hermetic separations (usually derived from the privileged status of some settings and their research). In such cases, settings may be imposing too narrow a framework for categorizing interactions and may also be limiting our ability to examine interpreting dynamics to their fullest potential.
On the one hand, setting-based differentiations may mask differences within one particular setting that could deepen our understanding of that particular interpreting reality -for example, when supposedly cooperative healthcare encounters become adversarial because of the controversial nature of the issue at hand (e.g., end-of-life decisions, reproductive rights). On the other hand, focusing on specific settings may also limit our ability to identify similarities across settings and, therefore, to compile sufficient empirical data to reaffirm or refute certain disciplinary, particularly contentious tenets (e.g., the interpreter's role).

Contextual factors as key to analysing interpreter-mediated events
As an alternative to setting-based analyses, examining specific contextual factors offers a more granular view of each individual communicative event. Several scholars have implemented this approach in categorizing interpreter-mediated events (Alexieva, 1997;Angelelli, 2000;Mikkelson, 1999;Pöchhacker, 2004). Alexieva (1997) challenges categorizations based on single parameters (e.g., the setting or the nature of the input text) and proposes a multiparameter approach to account for the diversity of interpreter-mediated events. Rather than delineating clear-cut categories, this approach describes "families" with "central members (prototypes) and peripheral members (blendforms)" that are identified along a continuum (p. 156). This allows her to position interpretermediated events along a continuum of "universality vs culture-specificity" based on a number of scales that emerge from the analysis of these parameters: "distance" versus "proximity"; "non-involvement" versus "involvement"; "equality/solidarity" versus "non-equality/power"; "formal setting" versus "informal setting"; "literacy" versus "orality"; "cooperativeness/directness" versus "non-cooperativeness/indirectness"; "shared goals" versus "conflicting goals" (p. 169).
Whereas Alexieva tries to depart from the concept of setting (although her concept of "family" remains underdefined), Angelelli (2000) argues that contextual analyses are key to understanding the similarities and differences across interpreting settings more deeply. Using Hymes' (1974) theory of speaking, she provides a comparative description of a community and a conference interpreting event. Specifically, she analyses Hymes' fundamental notions (ways of speaking, fluent speaker, speech community, speech situation, speech event, speech act, rules of speaking and function of speech) for each event separately and then she compares the events through the framework of Hymes' speech components (summarized in Table 1). This analysis allows Angelelli to conclude that the difference between these two situations (settings) requires different performances by the interpreter: as the understandings of the communicative act itself are different, so are the abilities required of the interpreter and the demands made on them. Table 1 summarizes the parameters that these two models offer for analysing interpretermediated events. While there is noticeable agreement between both proposals, some categories are exclusive to one model and others are more detailed in one than the other (and even broken down into multiple parameters).  Alexieva (1997) and Angelelli (2000) Alexieva (1997) Angelelli (2000) Mode of delivery and production (i.e., interpreting mode) (includes distance between participants and directness of contact as corollaries) Participants (includes: number of participants; command of languages; involvement in the textual world; status/power relationships (influenced also by age and gender); role) Participants (focus on who is speaker, addressor, hearer, addressee)

Topic
Message content Text type and text-building strategies (i.e., orality vs literacy; including also non-verbal behaviour) Message form Key (i.e., tone, manner) Form of speech (i.e., varieties and registers) Genre Spatial and temporal constraints (includes: distance between participants and privacy as corollaries) Setting (understood as time, place and physical circumstances of speech act) Scene (understood as "the cultural definition of an occasion; the 'psychological setting'") Goal of the event Purposes-outcomes (i.e., expected outcome of event as recognized by speech community) Purposes-goals (i.e., intentions of participants) Norms of interaction (i.e., rules governing speaking) Norms of interpretation (i.e., how the belief system of the community shapes the understanding of utterances) Other typologies include factors that are not reflected in these proposals (either explicitly or at all) but which may be relevant to certain analyses, such as the professional status of the interpreter or the specific language combination (Pöchhacker, 2004, pp. 17-25) or marketrelated factors such as interpreter selection criteria, professional organization and regulations or the status of the job market, among other factors (Mikkelson, 1999).

Contextual factors for analysing collaboration and negotiation patterns in interpreter-mediated events
The breadth of existing contextual factors in any given event makes it likely that the factors that have the potential to be particularly relevant to a particular study -and therefore be crucial to establishing parallels and divergences with other instances of interpreting -will depend on the object of study. Indeed, Alexieva (1997) stresses that, of all the parameters at play in her analysis, the issue of distance versus proximity is the most important in determining the degree of culture-specificity of an event; this parameter is followed by noninvolvement versus involvement, equality versus non-equality and shared versus conflicting goals.
This section offers a proposal for the contextual factors that appear to be the most relevant to the analysis of the negotiation and collaboration patterns established in the communicative triad in order to achieve successful communication in interpreter-mediated events. For these purposes, existing categories have been expanded and/or refined to match the specific features of this particular analysis. This section also highlights the ways in which these factors are not necessarily setting-dependent. Parameters not included here are considered to have little to no impact on the dynamics to be observed. Nevertheless, this list should not be construed as being exclusive and may need to be expanded following future research on this topic. Likewise, items in this list may be relevant to exploring other research questions.

Degree of interaction between participants and interpreting mode
In order for negotiation and collaboration to occur, it is essential that a certain degree of bidirectional or multidirectional interaction be feasible between the participants. This is greatly shaped by the information flow (dialogue vs monologue), which often also determines the interpreting mode. Monologue-based communication is unidirectional in nature; in contrast, dialogic interactions require a back-and-forth interactive process, where collaboration and negotiation are often an intrinsic part of the communicative process. Therefore, generally speaking, simultaneous interpreting as the mode of choice for monologue-based interactions would be more unidirectional, whereas dialogue interpreting would lend itself to greater flexibility to accommodate co-construction processes.
Another factor that influences the degree of interaction between participants is whether they share the interactional space. A shared interactional space allows all participants, interpreter included, [to] interact with one another directly through multiple verbal and embodied semiotic resources (such as gaze and head movement, posture and body orientation, gestures, facial expressions) (Davitti, 2019, p. 8).
However, the interactional space can be either physical or virtual (as in video-mediated interpreting). What is key is the ability to interact directly with one another, which requires the existence of open communication channels in every possible direction.

Degree of transparency of the language constellation
Participants in interpreter-mediated interaction may have some understanding of their cointerlocutors' language. These partially transparent language constellations (Müller, 1989) can alter the conversational dynamics and participant statuses, including the interpreter's, as the primary participants only partially understand one another (Meyer, 2012, p. 100). Besides the more obvious direct interaction between the parties (with the interpreter remaining on standby: Angermeyer , 2008), partially transparent language constellations also allow primary participants to engage to some extent in linguistic monitoring. By being able to follow, at least partially, the parts of the interaction that occur in the "other" language, they can assess and comment on the interpreter's linguistic choices (Anderson, 2012;Angermeyer, 2015).
However, a partially transparent language constellation is not necessarily a precondition of this monitoring function and, therefore, of negotiation and collaboration in multilingual conversations. Primary participants still react to interpreters' behaviours that are evident regardless of the language barrier, including -using Wadensjö's terminology (1998, pp. 107-108) -some expanded and reduced renditions and, most notably, zero and non-renditions (Martínez-Gómez, 2020).

Interpreter's interpreting background
All the participants' individual characteristics play a significant role in the construction of an interpreter-mediated event, including those of the interpreter, as is discussed in point 4.6. However, one key element to consider when studying collaboration and negotiation patterns is the interpreter's professional or non-professional status. Professional interpreters are trained in sanctioned best practices in interpreting, which provide them with tools and skills that enable them to face challenges successfully. However, these best practices are occasionally rooted in widely debated disciplinary norms (particularly the nature and boundaries of the interpreter's role), which have been found to contribute to professional interpreters' feeling restricted when needing to respond with flexibility to a given situation ("the interpreter's paradox" : Metzger, 1999).
Although it is yet to be determined to what extent professionals and non-professionals actually differ in their behaviours, it is nevertheless likely that NPIs require greater support from the primary participants. Primary participants' support may occur at the linguistic or the translational level (e.g., clarifying messages or finding equivalents), at the coordination level (e.g., turn length and turn-taking), etc. More significantly, individuals with little or no interpreting education or experience will probably be unaware of (and certainly not bound by) potentially limiting deontological regulation of their behaviours. They may be more sensitive to primary participants' needs and expectations, which may be the most immediately available criterion they have to guide their interpretation. Given this potentially stronger interdependency between participants, NPI-mediated interactions may allow for more productive examinations of negotiation and collaboration.

Environmental and institutional constraints
The public versus the private nature of an event and its level of formality may effectively limit the participants' freedom to act. Private and informal situations are more likely to offer participants (including the interpreter) the flexibility to intervene to clarify, expand, question or even re-conduct communication (Hale, 2007, p. 46). Other institutional constraints, such as time pressures to complete the interaction as soon as possible (e.g., in medical practices: Hsieh, 2006, p. 726), can also have a significant impact. Institutional constraints may also affect the interpreter-mediated event on a much larger scale. For instance, budgetary limitations or security protocols may determine the choice of interpreter.
Institutional norms and protocols can also limit participants' freedom to respond to the changing needs of the communication process. For example, in the courtroom, "the rules of evidence stipulate who can speak, at what time and in what manner" (Hale, 2004, p. 32). Likewise, similar restrictions apply to interpreters in events where their role is highly regulated. In US courts, established protocols require that interpreters seek permission for any intervention that is not considered strictly as an interpretation (e.g., a request for clarification). 1

Purposes (outcomes and goals) and nature of the event
Hymes argues for differentiating outcomes, as the expected outcomes of an event as recognized by the speech community, and goals, as the individual intentions of the participants. To analyse negotiation and collaboration patterns, it seems crucial to identify them, the extent to which they coincide, and the way in which they define the adversarial or cooperative nature of an event.
The socially expected outcomes of any communicative event are often based on the social and institutional roles that each participant plays. These outcomes, in turn, often determine the individual goals of the participants, as they act according to their assigned roles. When the individual goals of participants coincide between themselves and with the interaction outcomes, the encounter can be considered cooperative and lend itself to more extensive collaboration between the parties (e.g., "doctor and patient are partners engaged in a common struggle against an illness": Angelelli, 2004a, p. 16). In contrast, adversarial encounters are characterized by conflicting participant goals and socially sanctioned interaction outcomes that often benefit only one party to the encounter (e.g., a crossexamining attorney aims to prevent witnesses from delivering persuasive narratives: Harris, 2001).
The adversarial-cooperative dichotomy has traditionally been used as a distinctive factor between settings, particularly in public service contexts (e.g., legal-adversarial vs medicalcooperative). However, settings are not entirely homogeneous. For example, participatory lawyer-client interactions that empower the client (Hale, 2007, pp. 79-80) contrast with inquisitive police interrogations seeking inconsistencies (Berk-Seligson, 2009). Furthermore, certain encounters are characterized by mixed outcomes and goals, which in turn generate a combination of both adversarial and cooperative features.
In addition, participants may occasionally seek goals beyond their socially sanctioned ones, creating a disparity between goals and outcomes, usually because they have either multiple social or institutional roles, or personal goals external to the encounter. Moreover, the interpreter's goals can often have as strong an impact on the interaction as the primary participants'. This factor has traditionally been overlooked as research has focused on professional interpreters, who are less likely to have a personal stake in any given encounter. However, research on child language brokering has shown how young brokers' own interactional goals guide their behaviours and their translational choices, in alignment with either one or both of the primary participants (Orellana et al., 2003) or not (Bauer, 2013).

Interpersonal factors
In any interpreter-mediated interaction, a wide variety of social factors related to all the participants, including the interpreter, come into play -from personal attributes (e.g., gender, ethnicity, socioeconomic status, education level, age) to institutional roles, all of which coexist within larger societal frameworks and belief systems (Angelelli, 2004b, p. 29). The interplay of all these factors, together with the nature and goals of the encounter, shapes the relationships between parties. These relationships will, in turn, be crucial to examining and understanding collaborative and non-collaborative behaviours during interactions. Relationships may have an impact beyond the immediate exchange of information and be crucial to accomplishing larger communication goals. For example, caring and comfortable relationships between physicians and patients promote patients' greater adherence to treatment, greater satisfaction and less concern about illness (Angelelli, 2004a, p. 16).
These relationships, however, are not always built on even ground. Power structures create unequal relationships that invariably affect communication. As Alexieva (1997) argues, the major bases of power derive mostly from the social status of the primary participants institution-wise (their institutional affiliation and position within the hierarchy) and expertisewise (their prestige as authorities on the issues discussed). (p. 225) During witness testimony, for example, the party with institutional authority (counsel) controls the flow of information, whereas the witness remains restricted in the ways they can respond and react (Hale, 2004, p. 32).
Interpersonal relationships in the communicative triad often give way to issues of alignment, particularly between the interpreter and one of the parties; such issues are often founded on sociodemographic characteristics or institutional role. NPIs have traditionally been criticized for aligning with public service users, who may be their relatives or friends or share their national or ethnic background. As Tipton (2011)  However, alignments are also observed among professional interpreters. Asylum interpreters, for instance, tend to align themselves with the institutional representative, "either by facilitating the bureaucratic process or, more controversially, by assuming the role of institutional gatekeeper" (Maryns, 2015, p. 25). Further empirical research is needed to assess whether these behaviours may also have a constructive, rather than a disruptive, effect in certain situations (e.g., see the concept of "performance teams" in child language brokering: Valdés, 2003, pp. 95-97).

Method
Using contextual factors as our analytic foundation becomes particularly relevant when studying contexts where multiple settings overlap. This is the case with mental health assessment and treatment of incarcerated offenders in prisons. The following section summarizes a description of the main contextual factors identified in a corpus of 26 naturally occurring mental-health interviews mediated by non-professional interpreters (NPI) in two male prisons in Spain. These interviews feature a total of 10 prison therapists and counsellors, 26 inmates with limited Spanish proficiency (LSP prisoners), and 18 inmates acting as interpreters.
Audio recordings of these 26 interviews were transcribed following an adapted version of Atkinson and Heritage's (1985, pp. 346-369) Conversation Analysis transcription guidelines and then translated professionally. The transcripts were then coded following a three-step approach. First, I conducted a top-down analysis based on Wadensjö's (1998) rendition typology and Merlini and Favaron's (2005) footing typology. The goal was to identify the primary participants' and interpreters' utterances that "diverge" from the cannonical interactional format of interpreting -that is, those that do not conform to the traditionally expected "normal format of mediated discourse structure" ("Normalform der Mittlerdiskursstruktur": Knapp & Knapp-Potthoff, 1985, p. 457), close renditions (Wadensjö, 1998, p. 107) and reporter footings (Merlini & Favaron, 2005, pp. 279-280). Recapitulator and narrator footings were not included either, because collaboration and negotiation patterns were not considered to be affected by pronoun choice or the use of direct versus indirect style. It is in those utterances that "diverge" from the cannonical interactional format where negotiation and collaboration strategies are expected to emerge.
Second, I coded the selected utterances based on whether they reflected an instance of collaboration, negotiation or neither. The collaboration code was used in (single-or multiturn) instances where one or both of the primary participants and the interpreter jointly worked towards the interaction outcomes and goals (by constructing meaning, coordinating talk or engaging in relationship-building). The negotiation code was used in (single-or multiturn) instances where one or both of the primary participants showed discontent or disagreement with an interpreter's behaviour, either explicitly or implicitly.
Third, I analysed each of the segments coded as either collaboration or negotiation and linked them to the relevant contextual factors at play. The coding process was facilitated by the qualitative analysis software NVivo 11.

Prison mental-health interviews as context: A description of contextual factors and their influence on instances of collaboration and negotiation
This section summarizes the analysis and description of the above-mentioned contextual factors in the specific context of mental-health interviews conducted in prison. It also provides a few selected examples from the corpus in which the influence of these factors . Contextual factors as an analytical tool: Exploring collaboration and negotiation in mental health interviews in prisons mediated by non-professional interpreters. Linguistica Antverpiensia,New Series: Themes in Translation Studies,20, 38 becomes evident in the development of collaboration and negotiation strategies among the members of the triad.

Degree of interaction between participants and interpreting mode
All the interactions in this corpus are dialogical (interviews), occur in person and are interpreted in the bilateral mode in eight language combinations (Arabic, Bulgarian, English, French, German, Romanian, Russian or Turkish in combination with Spanish). As explained above, dialogical communication interpreted bilaterally, especially if conducted in a shared interactional space, is more conducive to accommodating co-construction processes. In these interactions, not only can turns-at-talk be more easily negotiated, but body language also provides an additional layer of immediate meaning that contributes to the participants' assessments of communication success, which enables them to interject and repair more proactively. Unfortunately, body language was not included in this analysis, because video recording of the interviews was not allowed.
Excerpt 1 illustrates a common occurrence: an instance of linguistic collaboration, where the NPI needs support either to understand a term or to provide an equivalent for it. In this case, the interpreter expresses difficulty with the word antecedente, signalled through his hesitations and his code-switching into Spanish. The dialogical nature of the interaction and the interpretation mode, coupled with the shared interactional space, facilitates the therapist's intervention, that is, to rephrase her question and clarify her meaning.

Degree of transparency of the language constellation
In prison-based interactions, partially transparent language constellations are common, mostly as a result of the inmates' increasing proficiency in the institutional language(s). In fact, one of the main strategies to overcome language barriers in prisons worldwide is to provide (sometimes mandatory) language courses for inmates (Martínez-Gómez, 2014, p. 251); this is the case in the prisons studied here. Many prison systems also offer language training for their staff or have hiring initiatives and employment incentives to create more linguistically diverse workforces (Martínez-Gómez, 2014, pp. 250-252). In Spanish prisons, though, courses are offered for only a limited number of mainly Western European languages (English, French), which only rarely coincide with the most common linguistic profiles of prisoners. In this study in particular, 21 LSP prisoners reported speaking at least some Spanish, whereas only the four therapists who were interviewing English-speaking inmates had a basic knowledge of the inmates' language. . Contextual factors as an analytical tool: Exploring collaboration and negotiation in mental health interviews in prisons mediated by non-professional interpreters. Linguistica Antverpiensia, New Series: Themes in Translation Studies,20, 39 The partial transparency of the language constellation influences the way turns are organized during the interaction. Specifically in this corpus, 23 of the 26 LSP prisoners engage in direct communication in Spanish at some point during their interviews. In these situations, and following the cues of primary participants, the NPI relinquishes his turns and adopts a standby mode, aware that communication is being effective without his contribution, and so he joins in again only when the need arises.
In Excerpt 2, which occurs at the beginning of a session, the NPI initially assumes language opacity and takes the second turn to render the counsellor's question. However, the LSP prisoner's direct response in Spanish in turn 3 signals that the NPI's participation is unnecessary at this moment. Accepting this negotiation of translation relevance (Baraldi & Gavioli, 2010), where a primary participant signals that one or more turns do not require translation, the NPI remains silent during a few adjacency pairs. After turn 8, the NPI interprets the 2. Likewise, this transparency in the language constellation facilitates monitoring of the NPI's renditions by the primary participants (particularly the LSP prisoners). In Excerpt 3, the presence of cognates for certain drug terms allows the LSP prisoner to understand and correct the NPI's expanded rendition. This instance of negotiation seems particularly necessary for the LSP prisoner, because the NPI's rendition threatens his social image vis-à-vis the therapist (face, in Goffman's terms, 1967, p. 5

Interpreter's interpreting background
Most prison interpreters, both in Spain and elsewhere, are often inmates themselves (Martínez-Gómez, 2014, p. 244). Those participating in this study tend to be selected on the basis of language competence (as perceived by the primary participants), immediate availability and/or trust (either by the LSP prisoner or by the prison staff member). Their developing language abilities usually require a greater need for support for their linguistic and translational skills, as seen in Excerpt 1. Likewise, as NPIs without any prior interpreting training, they bring a certain degree of openness to their role. Their previous experiences acting as NPIs (for half of them, of over a year) are likely to inform their decisions in these interviews.
Furthermore, similarly to the primary participants, interpreters bring their full selves to each interaction, including their condition as incarcerated offenders. The same factors that shape the behaviour of the LSP prisoner because of his prisoner status (adversarial-cooperative tensions, individual goals, power structures, etc.), as discussed below, also shape the behaviour of the prisoner-interpreter. Relationships and trust-building processes are affected by the presence of another inmate in the interaction. The NPI's possible alignment with either party may change the interpersonal dynamics of the conversation, especially when this alignment may shift at any point. Most of the following excerpts illustrate this added layer of interpersonal idiosyncrasies to varying extents: the NPI's contribution to interactional goals (also potentially to improve his own face) (Excerpts 5 and 7); his interest in improving his peer's face (Excerpt 6); or his attempt to introduce a new conversation topic out of his own interest (Excerpt 4).

Environmental and institutional constraints
The environmental conditions in which mental health interviews take place in prison are highly conditioned by the security protocols governing every aspect of prison life (Baixauli-Olmos, 2013). Common protocols to avoid the non-essential movement of inmates in the complex may constrain the primary participants' ability to choose certain NPIs for a specific encounter. Infrastructure and physical safety guidelines may hinder the need for privacy and confidentiality that exists at the core of the therapeutic alliance. All the interviews analysed in this study (and described in this article) took place in meeting rooms in the residential block where the inmate resides -again, to avoid non-essential movement. Many of these rooms led off heavily trafficked common areas and the door was kept open in most cases, possibly for safety reasons (especially because of the researcher's presence). Moreover, this need for privacy runs counter to the institutional expectation that all information pertaining to inmates would be disclosed to the prison administration (Campbell, 2004, pp. 791-792).
Other aspects of the infrastructure also impose conditions interpreted events in prisons: equipment, lighting, ventilation, acoustics, etc. (Baixauli-Olmos, 2013). In some interviews in this corpus, noise emanating from the nearby common areas (especially when the door was kept open) often made multiple repetitions and clarifications necessary.
Furthermore, as some of these interviews show, inmates sometimes use these interactions to raise issues unrelated to the purposes of therapy (e.g., ensuring that arrangements for release have been made, facilitating communication with family members, understanding legal documents, etc.). Both LSP and NPI inmates seem to use these encounters as a direct channel of communication with someone who has institutional knowledge, regardless of their area of expertise. This is particularly important when they seem to experience prison professionals' having limited availability to meet with inmates. In fact, as one inmate mentions in one of the interviews: "I want to know while I'm here because obviously she's not here every day." Excerpt 4 illustrates a request by the NPI for the counsellor's help: the NPI takes advantage of the LSP prisoner's request to check on an unresolved procedural matter (getting credit for time served) and reminds the counsellor to look into the same issue for him too. This triggers a 10-turn conversation between counsellor and NPI in which this personal request is negotiated: the counsellor acknowledges the issue, but reminds the NPI to maintain the focus on the LSP prisoner ("hoy hemos venido a lo que hemos venido"), while also committing to returning the following day to help the NPI.

Purposes (outcomes and goals) and nature of the event
Two types of mental health interview were identified in this corpus: 23 intake and 3 followup interviews. Intake interviews take place both on the arrival of a new inmate to the prison or to a specific residential wing. Their main purpose is to gather information about the prisoner's background and current situation in order to assess his mental health status and potential needs. This assessment results in either proposals for psychological treatment (individual or group programmes for specific issues) or referrals to other services in the prison, or it is used as part of the information considered for a variety of administrative processes (e.g., assignment to a residential wing, job placement, etc.). Follow-up interviews engage more deeply with the psychological issues identified during intake, including exploring the factors that led to criminal behaviour. They usually aim to provide further individual psychological treatment or to update information on the mental health status for administrative purposes (e.g., furlough, disciplinary action, etc.).
These institutionally defined outcomes coexist with socially defined outcomes for imprisonment and for mental healthcare, which generate a hybrid context where both adversarial and cooperative features coexist. On the one hand, the punishment purposes of incarceration sentences come into play: deterrence, incapacitation and retribution (adversarial) and rehabilitation (cooperative). On the other, prison mental healthcare is considered mostly cooperative, as its ultimate purpose is identifying and implementing the best course of action for the client's treatment. But it may present serious adversarial features that stem from the fact that inmates are often involuntary recipients of psychological treatment (Campbell, 2004, p. 791). This reluctance to undergo mental health treatment, however, was not observed in the vast majority of interviews in this corpus. Several hypotheses may explain this, particularly a self-selection bias and interpersonal factors.
Regarding the former, five scheduled interviews were cancelled during the fieldwork because inmates refused to participate in the study. This may possibly be explained by lower levels of engagement in psychological treatment. Regarding the latter, see Section 6.6.

Many of these interviews illustrate NPIs' awareness of institutional expectations and goals
and their active engagement to ensure the LSP prisoner's full participation in the interaction. Limited educational levels, a lack of familiarity with institutional norms and procedures, emotional distress and the use of vehicular (non-native) languages, among other factors, heavily influence LSP prisoners' ability to engage fully in mental health counselling. In this context, NPIs seek to ensure that the LSP prisoner understands the therapist's questions and recommendations and offers relevant responses. They often do so by engaging in exchanges that go beyond traditional interpreter's renditions ("close renditions" in Wadensjö's terms, 1998).
This is particularly visible when NPIs aim to repair LSP prisoners' replies that appear irrelevant or insufficient to meet the therapist's expectations, namely, utterances that are considered "not immediately translatable" (Baraldi & Gavioli, 2015, p. 58). Through follow-up questions, which sometimes even suggest potential answers, they prompt LSP prisoners to repair, clarify or expand their initial replies. These dyadic exchanges between NPIs and LSP prisoners provide an opportunity to work out problems associated with perceived information gaps that, left unfilled, may run counter to the data-gathering purposes of the exchange (incomplete dates of birth, unspecific time periods, unstated locations of events, etc.).
Occasionally, however, these dyadic sequences also respond to "non-linguistic differences in the perspectives of the interlocutors" (Baraldi & Gavioli, 2016, p. 43), which could affect the achievement of conversational goals. In Excerpt 5, the therapist intends to assess whether the LSP prisoner has substance-abuse problems. After asking about drug use, she enquires about the LSP prisoner's alcohol consumption. Her laughter after the NPI's rendition of "normal", coupled with the NPI's concern with potentially different cultural habits between France and Spain, may have served as a trigger for the NPI's initiative to request further clarification. In order to facilitate the gathering of essential information from the LSP prisoner, the therapist seems willing to relinquish her control over the conversation momentarily. These behaviours, through which the NPI appears to support the LSP prisoner's full participation in the encounter, also contribute actively to the information-gathering purposes of the interview. In this sense, the NPI is closely collaborating with the therapist in accomplishing an interaction outcome. However, at the same time, this behaviour may also be triggered by the NPI's personal goal of improving his social image and his relationship with the therapist. This is particularly common when NPIs answer the therapists' questions directly, as will be seen next.

Interpersonal factors
Of all the social factors (personal attributes, institutional roles, and societal frameworks and beliefs) that shape the behaviours of all the participants, the power structure inherent in the prison system appears to be the most salient in this context. Prison therapists' power goes beyond the communicative event over which they have control in a variety of respects: topic selection, duration, scheduling, etc. As institutional representatives, they also have the authority to trigger punitive responses to disruptive behaviour. As members of decisionmaking bodies (treatment teams, disciplinary boards, etc.), at least in Spanish prisons, they participate in procedures that affect the inmate's life conditions in the long term (e.g., decisions about furlough, parole and educational or work opportunities). Given the weight of the therapists' assessments in inmates' prison life and future prospects, it is logical that inmates would want to present themselves as cooperative and trustworthy.
Relationship-building is often relatively implicit in NPIs' attempts to contribute to the perceived interactional outcomes of the event, often by repairing the LSP prisoner's responses (Excerpt 5) or by answering questions directly on his behalf (Excerpts 6 and 7). NPIs may consider and use these initiatives to signal to therapists that they can be helpful and trusted. In rare cases, however, the parties' own interpersonal goals interfere with one another. In Excerpt 4, for instance, the NPI's interest in contributing to the institutional outcomes of the interaction (and his personal goal of appearing collaborative) clashes with the protection and promotion of the LSP prisoner's face and triggers an immediate negotiation and repair by the latter.
This implicit relationship-building also emerges in other interactional features, such as humour and displays of genuine concern for an inmate's troubles, which coexist and sometimes outweigh the suspicion and wariness that often characterize prisoner-staff relationships (Campbell, 2004, p. 792). In fact, friendlier relationships benefit not only the inmates' personal interest in being in the therapists' good graces, but also the therapists' need to build the trust-based therapeutic alliance that is essential to productive psychological intervention.
However, relationship-building occasionally becomes explicit, with participants being openly proactive, particularly the NPI in trying to enhance the LSP prisoner's face. In Excerpt 6, the NPI answers on behalf of the LSP prisoner (implicit relationship-building in his own favour) and even justifies this initiative by explaining that they are cellmates, which the counsellor accepts explicitly ("claro muy bien") and even follows up with a request for confirmation. At that point, the NPI takes this opportunity to provide, unprompted, his assessment of the LSP prisoner's character ("buen chaval"). As a more veteran inmate, he may consider it useful to build on the trust that he may have developed with the counsellor over time to facilitate relationship-building between this newer prisoner and the counsellor. Once more, the counsellor accepts this intervention explicitly ("muy bien"), but immediately re-conducts the focus of the conversation to the LSP prisoner. His shift in pronoun use, addressing the LSP prisoner directly, signals his intention that the NPI returns to a reporter footing (Merlini & Favaron, 2005, pp. 279-280). Although most of the NPI's direct responses are accepted by the primary participants, in a few instances the triad negotiates these behaviours jointly. Whereas LSP prisoners are more likely to adopt strategies that avoid open confrontation, therapists have the institutional authority to be more explicit about their expectations of everyone's role. In Excerpt 7, the therapist openly expresses discontent with the NPI's direct response and asks him to correct his behaviour. It is worth noting that this same therapist agreed with a different NPI intervention earlier in the interview (Excerpt 5). This illustrates how therapists need to balance their need to exert their institutional power, which can be alienating for the inmates, with interactional outcomes, including building a therapeutic alliance, which is intimately related to interpersonal dynamics. As is revealed later in the interview, this LSP prisoner cannot leave his current residential block because of conflicts with other prisoners. The fact that this issue indicates his relationships inside the prison and possibly also his emotional wellbeing may have guided the therapist's decision. In the process, the therapist slightly jeopardized the trust-building process with the NPI in order to gain a deeper insight into the LSP prisoner's mental health status.

Conclusions
This article argues that analyses of contextual factors can be more productive for understanding and explaining interpreter-mediated interaction than examinations based on settings as theoretical categories. In fact, studying interpreting within the framework of siloed settings may limit our ability to discern commonalities of interpreting as a social activity, both across and beyond settings (Downie, 2021). I would also argue that strict and somewhat prescriptive definitions of setting hinder the research discovery process in communicative events that do not match those features. Such definitions can have the effect of predisposing the researcher to situate their findings as exceptions or, even more problematically, to force a framework onto their data that does not match the reality being studied.
As an alternative to setting-based analyses, this article proposes in-depth examinations of contextual factors, particularly a selected group that will be relevant to the research question at hand and that will have a certain degree of explanatory or predictive power over it. Whereas some of these contextual factors will probably align with those identified for particular settings, others will not. This article offers an example of this type of analysis as applied to the study of collaboration and negotiation patterns in interviews for assessing prisoners' mental health. The excerpts summarized here show that, for example, the immediate presence of all the participants and the bilateral mode of interpreting allow primary participants to help the NPI to deal with terms that he may be unfamiliar with. They also illustrate how the partial transparency of the language constellation allows primary participants to interact directly with each other and to signal when they need interpreter support. The NPI's awareness of the therapist's outcomes (i.e., gathering accurate information) leads him to seek direct clarifications and even to provide answers on behalf of the LSP prisoner, thanks to the prior relationship between the inmates and the NPI's knowledge of his peer's background and current situation. At the same time, these initiatives are possibly also triggered by the NPI's own goal of appearing helpful to the therapist, given that her opinion may be key in future decisions about the NPI's penitentiary situation. His alignment with the LSP prisoner, who may also want to project a positive image of himself before the therapist for the same reasons, may also lead the NPI to include comments that promote the LSP prisoner's face.
These factors also affect the primary participants' behaviours and reactions to the NPI's initiatives. Whereas a therapist may agree to the NPI's sharing of factual information or seeking clarification, they may disapprove of interventions that may hinder their professional ability to assess the LSP prisoner's emotional state, which could directly jeopardize a crucial interaction outcome (e.g., Excerpt 7). In the same vein, the LSP prisoner may be more than willing to let the NPI cast him in a positive light, as that could possibly improve his relationship with the therapist. But he will certainly be quick to correct information that threatens his face if he is able to (e.g., Excerpt 3). Indeed, as these interviews show, collaboration processes entail not only agreement and support (e.g., by filling in the linguistic gaps), but also monitoring and re-conducting the conversation when necessary, even by the less powerful party (negotiation). Primary participants employ multiple strategies to accept and reject these interventions: from relinquishing turns, laughing or even verbally approving a particular comment, to claiming turns back, correcting the NPI or explicitly disapproving of his behaviour.
The factors identified in this article have proven useful to describe these NPI-mediated mental health interviews in prisons. Looking ahead, further research will be needed to assess their generalization potential and their predictive power, and also to determine whether this list needs to be shortened or expanded.