Ostension , inference and response : analysing participant moves in Community Interpreting dialogues

Following a review of the methods employed in some recent studies, this paper proposes a way forward for pragmatics-sensitive research into actual participant moves in community interpreting events. Its aim is to overcome some of the objections that have been raised to methods in critical discourse analysis, conversation analysis and pragmatics and to relate microlevel analysis of participants’ utterances to the broader issues of role, power distribution, norms and so on that have dominated discussion of interpreter-mediated communication. Adopting a broadly ostensive-inferential view of communication, we examine the nature of the evidence that can be adduced in support of causal models and suggest that it is to be found in the real-time responses of the participants themselves to each other’s moves rather than in analysts’ imagined reconstruction of context, intentionality and acceptability.


Current trends
makes a plea for a cultural turn in interpreting studies so that it can "explicitly address questions of power and issues such as class, gender, race in interpreting situations".He implicitly criticises (2002: 49) interpreting research hitherto for having conducted "depoliticized, minimally contextualized experiments, carefully controlled by a researcher who assumes objectivity".In similar vein, Inghilleri (2005b: 125) speaks of the move in Translation Studies as a whole away from concern with "translated textual products" and towards translation seen as a "social, cultural and political act", linked to issues of power and control.
There can be no doubt that the perspectives opened up by the Cultural Turn in Translation Studies have proven immensely beneficial to the whole field.Acts of translating have at last been put in proper historical and cultural perspective and issues of power and ideology have come to the fore.Inghilleri (2003;2005a, b) has taken important steps towards a sociology/ethnography of public service interpreting, drawing principally on Bourdieu.With hindsight, some of the earlier studies -of both translating and interpreting -do indeed seem to have a very narrow focus.But the Cultural Turn in translation studies has tended also to involve a turn away from anything to do with linguistics of any kind.Thus, Tymoczko and Gentzler (2002;xiii) refer disparagingly to effort spent in "analyzing minute literary and linguistic differences, such that often the forest could not be seen for the trees".Arrojo (1998: 34-35) denies the potential even of "contemporary linguistically-inspired approaches" to grasp non-essentialist views of meaning and the heterogeneity of translation while Venuti (1998: 22) claims that "linguistics-oriented approaches… would seem to block the ethical and political agenda" that he envisaged in his own "minoritizing translation" project.
While this is not the place to rehearse and/or challenge such arguments regarding translation or to contribute to the debate on "common ground" between culturalist and linguistic approaches to translation studies (cf.Chesterman and Arrojo 2000), in what follows I want to make a strong case for the co-existence, alongside cultural, historical and ethnographic approaches to interpreting, of conversation/discourse-analytic and pragmaticsbased studies of interpreter-mediated events.For at some stage in our studies of community interpreting we have to engage with what actually goes on in interpreter-mediated dialogues, the linguistic and paralinguistic moves of each participant and the consequences of these for the unfolding of the event.Specifically, I want to investigate: • What methods have been employed in linguistic approaches to researching community interpreting?How have discourse analysis, conversation analysis and pragmatics been deployed in the study of these phenomena and how can we overcome some of the methodological objections to their use?

•
Where next?How can we relate micro-level analysis of participants' utterances to the broader issues of role, power distribution, norms and so on that would be called for by a cultural/ethnographic turn?

Methods in Community Interpreting research
In descriptive studies of interpreting in a community setting, a number of (overlapping) approaches can be distinguished.In what follows, we consider the interactional approach, conversation analysis, critical discourse analysis and some aspects of pragmatics.

Interpreting as interaction and conversation analysis
It is to Cecilia Wadensjö (1992 and1998) that we owe the most comprehensive description to date of interpreter behaviour in public service encounters.Rather than concentrate exclusively on the interpreter as a translator, Wadensjö shows how each participant in the triadic encounter affects each other participant's behaviour.Her distinction between the interpreter's translating activity and the activity of co-ordinating others' talk is amply documented and has become a standard reference in later studies.Shifts of footing, distribution of responsibility, role expectations, the choice between representing or re-enacting others' talk, attending to miscommunications are all shown to be routine features of interpreting as an activity and are analysed in the communicative detail of relatively short transcribed sequences.Of particular relevance to us here are three key aspects of Wadensjö's methodology.
The first of these is that her approach is resolutely descriptive, rather than normative.In this, she is representative of a trend in empirical interpreting studies (e.g.Berk-Seligson 1990;Hale 1997, 2001, Roy 2000) that calls for comment, if only because descriptivism has come in for a good deal of criticism in translation studies as a whole (e.g.Hermans 1999: 35-6;Chesterman and Arrojo 2000;Crisafulli 2002: 29-32).The main objections are firstly that descriptivism appears to aim at objectivity (whereas the essential subjectivity of the analyst is now generally accepted) and secondly that description means an absence of evaluation and thus isolation from social and political aspect of interpreting.Yet description in Wadensjö's studies -and those of others listed in this section -does not appear to be motivated by any claim to true objectivity.Rather, it stems from a desire to avoid interpreting studies as error analysis, in which concentration on prescriptive issues of quality may blind us to important regularities of actual behaviour.Indeed, descriptivism in interpreting studies can be seen as a reaction to the excessive judgementalism that still exists in the interpreter's immediate environment (see, for example, Anthony Pym's (1999) account of outsiders' views on interpreting at the O.J.Simpson trial).It is natural that the analyst of actual, contemporary interpreter behaviour might seek to adopt a more detached stance, without this implying omniscience or total objectivity.
The second key aspect of method in Wadensjö's and others' work concerns the treatment of data.Central to empirical accounts of interpreting activity in community settings is the use of authentic, non-experimental 1 data and an interest, inspired from Conversation Analysis (CA), in the organisation of talk, seen as reflecting and reproducing social organisation.These organisational elements include turn taking, adjacency pairs, repairs, topic management and gate-keeping -see, for example, Straniero Sergio (1999), Roy (e.g. 1996;2000), Bolden (2000), Davidson (2002), Fogazzaro and Gavioli (2004) and further work by Wadensjö herself (e.g. 1999).For these purposes, close transcription of data and a reluctance to posit contextual information beyond what emerges between participants within the exchange are required.Paying close attention to the context that emerges in and between turns at talk is methodologically attractive inasmuch as analysts can be seen to avoid positing contextual information for events in which they were not direct participants.In practice, though, an analysis that does not have access to verifiable contextual information that was influential on an exchange is likely to be impoverished, as Cicourel (1992) has shown 2 .For this reason, Wadensjö (1998: 97) does not limit her use of context to a strict CA approach but appeals to external context introduced where relevant.This procedure is further justified by the fact that she is a participant observer of the encounters she analyses.The micro-analytic approach of conversation analysis casts light on what evolves within an ex-change.Pérez González (2006), drawing on recent studies of institutional interaction, adopts a dynamic view of context as constantly shifting and jointly negotiated: "each contribution in the course of interaction is not only determined by what was said before, but itself also contextualises upcoming talk " (2006: 391-92).Studies of this kind are able to reveal a lot about particular linguistic features and their manipulation, especially in the courtroom.Hale (2001) considers the strategic use of question forms in court as control and manipulation and shows how the illocutionary force of these is frequently modified by interpreters (cf.also Rigney 1999).Pérez González (2006) treats non-restrictive relative clauses in a similar way, showing how recontextualisation cues are used to re-align witnesses or defendants.Bolden (2000), the study that is perhaps most explicitly conducted according to CA principles, provides insight into the way in which interpreter moves in a medical encounter are oriented towards institutional goals.This is convergent with the findings of Meyer (2002), who studies the use of modal verbs in a similar medical context.
The third element of Wadensjö's approach relates to her distinction between 'talk as text' and 'talk as activity'.The first of these is akin to early discourse analysis approaches that tended to treat the text as an object in itself 3 .In this view, items in texts are said to refer anaphorically to other items or exophorically to real-world entities, events, etc, and this is how coherence is established.But as Green and Morgan (1981: 176) observe, "relevance and coherence, far from being linguistic properties of texts, are functions of the relation between observed acts on the one hand, and goals, intentions, purposes, and motivations inferred or inferrable by the hearer, on the other".By treating the data of speech events as evidence of participants' plans, goals and (inter)actions, interpreting research of this kind focuses on 'discourse' (as the negotiation of meaning among participants) and places human agency at the centre of communication.It thus opens the door to considerations of institutional constraints, power relations, role negotiation and so on.

Critical Discourse Analysis
A strikingly different reaction to the "positivist (…) tendency to regard language texts as 'objects' whose formal properties can be mechanically described without interpretation" (Fairclough 1989: 27) is the approach known as Critical Discourse Analysis (CDA).It is an approach to discourse analysis that brings to the fore the social conditions of the production and reception of texts and, therefore, the very topics of power, class, gender, race, etc. suggested as urgent issues for investigation by Cronin (2002) see section 1 above.In this way, CDA provides a bridge between linguistic studies of interpreting and those inspired by the Cultural Turn.The essential subjectivity of the analyst, the ideological nature of all discourse, post-colonial perspectives on the economic and political constraints governing all interaction are common ground between both fields.CDA treats 'discourses' as reflections of belief systems and of institutional power (contrast the view of 'discourse' mentioned above).
A fine example of CDA in community interpreting studies is provided by Pöllabauer (2004).Examining the role of interpreters in asylum hearings in Austria, she pays particular attention to "role conflicts, discrepant role expectations, the asymmetrical power distribution" and "existing (traditional) norm systems" (Pöllabauer 2004: 153).The power wielded by the interpreter is shown in the effects of interruptions, admonishments, omissions, tending to serve the interests of the institution.Miscommunications are shown to be due to "lack of shared background and linguistic resources" (ibid.171) between officials and claimants.While the analysis relies on transcribed data samples, the author relates these to underlying motives and institutional constraints (e.g. the interpreter's use of 'we' to align herself with the immigration officer and thus speak as the voice of the institution).Similar analytical procedures -and similar findings -are apparent in Mason (2005) and Maryns (2006: 2), who speaks of the "discursive construction of what counts as evidence" in such hearings.
It should not escape our attention, however, that CDA methods have been the subject of much criticism within linguistics and discourse analysis as a whole.Let us briefly identify four major objections that have been raised.They are (1) accusations of circularity in argument, ( 2) the fact that many analyses are not replicable, (3) the privileged position that analysts appear to grant themselves and (4) the absence of any consideration of cognitive factors (see Chilton 2005).The first of these criticisms relates to an alleged tendency in CDA to select texts for analysis because they exemplify discourses of power and control, to mine them for linguistic evidence of these processes and then to adduce the evidence in support of the original claim, thus producing a circular argument.Without entering the debate on the validity of this allegation (on which, see, e.g., articles in Toolan 2002), we can note that in this field the pitfall is easily avoided if analysts do not pre-select data in order to prove a particular point but constitute a corpus and treat all parts of it equally (i.e.counter-examples as well as supporting evidence).Indeed, the use of corpus linguistics can provide a good deal of the independent evidence that CDA analyses require (see Stubbs 1997).The second criticism, non-replicability, can also be avoided if studies of community interpreting make their methods explicit so that their findings can be tested by others (although the data on which studies are based can often not be made available to later researchers for obvious reasons of confidentiality).The third point is rather more intractable.It is that CDA scholars often appear to assert that particular instances of language use have particular effects.For example, Widdowson (2004: 151) objects to Fairclough's (1992) comment in an analysis of a text on antenatal care, "the message that comes across is one of re-assurance…": This may be the message that comes across to [the analyst] but he cannot know, of course, whether the same message comes across to the readers for whom this text was designed

Ian Mason
[…]. [W]e might ask 'Comes across to whom?' We just do not know whether this comes across to the prospective patients since they have not been consulted.
In other words, what evidence do we have of actual meanings received?Alternatively, if the analysis is no more than one reader's interpretation then what authority can the analysis claim?As analysts, we cannot claim to speak for actual participants unless -and here is the crucial point -we have gathered evidence of actual user response.Thus far, experiments designed to elicit user responses in interpreted encounters are very few (see, however, endnote 1).
The fourth objection listed above relates to cognition.CDA stands accused of paying insufficient attention to "what goes on inside people's heads" (Chilton 2005: 6), that is, what conceptual activity goes on during discourse processing.Yet, as Chilton suggests, cognitive linguistics offers theories of, e.g., metaphor (Lakoff and Johnson 1987) and 'conceptual blending' (Fauconnier and Turner 2002) that might assist in understanding how discourses form in the mind and spread from one mind to another.(See also Cicourel 2004Cicourel , 2006 on the need to integrate cognitive approaches into CA and Wodak 2006 for an attempt to do so in CDA).
These brief observations do justice neither to the achievements of CDA nor to the cogent criticisms that have been made of it.They do, however, suggest that CDA remains a useful tool in community interpreting research: the objections need to be addressed in the methods we adopt for data analysis and, above all, evidence from outside the transcribed interaction itself needs to be adduced in support of any claims about the discursive effects of particular uses.

An ostensive-inferential account of interpreter-mediated communication
Despite Wadensjö's (1998: 47) caution about the "cognitive bias" of pragmatics-based studies, which concentrate on "intentions and inner states of mind" rather than "activities in context", there can be no doubt that pragmatics offers an additional dimension to the analysis of interpreter-mediated communication, especially in the light of the alleged cognitive deficit of CDA, noted above.A body of pragmatics-oriented research now exists.For example, Meyer (e.g. 2002) draws on the action theory of Jochum Rehbein in his analysis of the interpreting of modality in announcements of planned action by medical staff.A number of scholars draw on politeness theory to show how participants pay attention to face management (e.g.Knapp and Knapp-Potthof 1987;Cambridge 1999;Mason and Stewart 2001;Pöllabauer 2004).
For the purposes of the present article, I want to draw on Relevance Theory (RT) (Sperber and Wilson 1995), specifically in order to cope with one of the methodological problems of CDA noted above 4 .RT's underlying model posits that communicators -including interpreters -will seek to maximise the contextual effects they create in exchange for the lowest expenditure of processing effort.Intuitively, this formulation of the principle of relevance seems particularly apt in the case of interpreters, who are constantly conscious of the need to be brief (efficient) and to-the-point (effective) because of the perception that their interventions hold up or lengthen the communication process.Some evidence of the operation of this principle in interpreters' output is available in Mason (2004).
RT is said to be an "ostensive/inferential" model of communication.In this conception of communication, speakers provide ostensive cues of what they intend to communicate while hearers infer what was intended from the evidence provided, including the ostensive indicators.For our purposes, it is important to see that, whereas we may not be able to show actual inferences made by actual participants in an exchange (except where these are made explicit by a hearer's response to what was said), we can identify ostension.Sperber and Wilson (1995: 49) define ostension as "behaviour which makes manifest an intention to make something manifest".Put differently, the basic information contained in an utterance is said to be accompanied by indications as to the communicative intention of the speaker in uttering it.We want to suggest that ways of making this communicative intention manifest include the use of "contextualisation cues" (Gumperz 1982: 131), defined as "any feature of linguistic form that contributes to the signalling of contextual presuppositions" and including prosodic phenomena, style switching, lexical and syntactic choices.These cues are, for example, central to analyses such as those conducted by Wadensjö (1998), Roy (2000) and Pérez González (2006).Equally significant are the nonverbal cues (gaze, facial expression, gesture, posture) that also seem to function as guidance to the hearer on the ways in which the speaker wants the utterance to be understood.More generally, we can say that markedness (the use of less expected or more unusual forms of expression) and salience (drawing attention to particular parts of an utterance) function as crucial aspects of ostension.
The reason why, on the other hand, the study of inferencing is more problematic is that the observer/analyst is not the intended or actual receiver in community encounters.Each participant has his/her own purposes, interests, pre-dispositions and so on.As Widdowson (2004: 76) observes: "What is relevant in text is what the users choose to make relevant in relation to what they are processing the language for".These intangible purposes are what Widdowson (2004: 79) calls 'pretext', the sum of purposes, assumptions and pre-dispositions that a user brings to his/her processing of text.Maryns (2006: 6), in her study of asylum interviews, uses a very similar concept, 'pretextuality', defined as "the entire set of contexts people have access to before they enter the interaction".While Maryns focuses on users' prior experience, Widdowson allows greater space for users' intentions but there is no reason why pretextuality should not include both.The implication of this insight is that, whereas we may be able to identify cues of os-tension in utterances, we cannot in any principled way predict what inferences will be derived by particular participants.How then do the words spoken in an interpreted exchange, the evolving micro-context and the institutional context and the pretexts of all participants work on each other in the joint construction of meaning?In our discussion so far, we have identified three approaches, often thought of as mutually incompatible: CA, CDA and RT.The speech events we examine in community interpreting, often using CA methods, are far from the situation of the idealised speaker/listener that is central to RT.It should be noted, though, that Sperber and Wilson (1997) have suggested that, whereas their model was necessarily devised at a level of abstraction, there is no reason why it should not be used at more concrete levels of analysis.And Watts (1997), among others, shows how this can be done in an empirical study (of verbal interruptions).At the same time, it is undeniable that RT has avoided consideration of the institutional factors that are clearly influential on community interpreting encounters.A methodologically careful version of CDA, we have suggested, can supply this.

Some applications
In order to pull together various strands from what precedes and point towards future research, let us consider three examples, two of which are cited from data made available in CDA and CA approaches to talk in asylum/immigration hearings (Barsky 1994;Baraldi 2006) while a third is made available here for the first time.In each case, I want to suggest some links between the distinct methodologies outlined above.

Assumptions in discourse and mutual cognitive environments
In some recent work (Mason 2006), I tried to show how each turn in interpreter-mediated exchanges potentially refocuses the direction of the exchange.In this way, distinct local contexts can emerge, not necessarily shared by all participants.A sequence cited by Barsky (1994: 142) is reproduced here in order to illustrate this communicative dilemma, frequently faced by interpreters, and to suggest an avenue for further exploration.

Sequence 1
A claimant, who had been arrested and tortured, is questioned about the circumstances of his detention.The official's turns at talk conform to a recognizable pattern, familiar within the genre of asylum hearings.Each turn is a question and the official confines himself to questions alone.Of more significance, however, is the fact that, considered together, we can see that these questions reflect a set of assumptions about routine sequences of events, familiar through repeated use, that typify particular social occasions and thus become sets of default assumptions (i.e.speakers and hearers will tend to assume that a stereotypical sequence of events will be followed in the absence of evidence to the contrary).This particular routine might be summarised as the 'police questioning routine': Arrest -Interrogation -Request for lawyer/contact with friend or family -Refusal of request -Continuation of interrogation.This routine is all the more familiar in the Western world in that it is played out in countless television dramas, detective stories and so on.Each of the questions asked in Sequence 1 appears to emanate from such an expectation or, from a pragmatics perspective, to presuppose it.Thus, asking for the assistance of a lawyer assumes a world in which it might prove beneficial to do such a thing while seeking to communicate with one's family presupposes that there would be some useful goal to be achieved by doing so.The claimant's responses, however, point to a quite different set of assumptions: that to ask for a lawyer, for example, entails recognition of the legitimacy of the arrest and plays into the hands of the oppressors; that joining a trade union is so dangerous an activity that it is best to keep it secret even from one's closest relative.It is clear that the official and the claimant are operating on the basis of different sets of premises.Now, Barsky does not present this particular sequence as having been mediated by an interpreter 5 .The purpose of reviewing it here, however, is to suggest that the "mutual cognitive environment" (Sperber & Wilson 1995: 41) required for relevant communication to take place may often not be established in encounters where a wide cultural gap exists and that this is a situation that interpreters have to deal with.The principle of relevance includes the notion that, to be relevant, an utterance must have some contextual effect, by strengthening, modifying or deleting previously held assumptions (Sperber & Wilson 1995: 109).The contextual effect thus modifies the context (set of premises used) for the processing of the next utterance in a constant process of re-contextualisation, as suggested by CA analyses.The peculiarity of the sequence in Sequence 1 is that contextual effects appear to be strictly local.They create the conditions for the next utterance in the sequence (e.g. the adjacency pair of question/answer) but they do not appear to modify context in the sense of "a subset of the hearer's assumptions about the world" (cf.Sperber & Wilson 1995: 15).For example, "I wasn't given that chance [to be assisted by a lawyer]" (line 3) entails that the claimant was not so assisted but does not entail that the claimant requested assistance.Hence, the immediately following question: "But did you ask it?"(line 5).Conversely, there appears to be no significant recontextualisation as each turn at talk follows the preceding one.The possible inference that involving outsiders (lawyers or family) in any way is futile and dangerous does not seem to be a new assumption entertained by the questioner.Thus, actual contextual effects are dependent on pretexts -in the sense of "the entire set of contexts people have access to before they enter the interaction" (Maryns 2006: 6).Into this uncertain terrain steps the interpreter.The ways in which successful communication is/is not established and the way interpreters respond to this lack of common ground, their struggle to establish a mutual cognitive environment call out for analysis on a wide scale.

Competing discourses: ostension, inference and response
Adopting both a conversation-analytic and a social systems theory approach, Baraldi (2006) 6 provides extensive evidence of the ways in which interpreters cope with the lack of common ground in the context of immigration and medical encounters.He analyses a particularly telling sequence, reproduced here as Sequence 2, in which a judge seeks to inform an immigrant that the latter's failure to renew a stay permit cannot be condoned on the grounds that the document had been lost.
Sequence 2  Baraldi (2006: 239) acutely observes that, in this sequence, the judge and the interpreter are performing different actions and thus producing different meanings: "the judge is asserting a legal norm, which has to be correctly understood by the defendant (…) while the interpreter is more interested in a way to create effective communication between the judge and the defendant, doubting that a correct translation is sufficient for this".For our purposes in this article, we can draw on Baraldi's case study to illustrate ostension and inference at work -as a complement to the kinds of insights into structure and organization that conversation analysis offers.In line 1, the Judge adopts a marked expression (da questo giudice) by referring to herself in the third person.The salience of such an expression suggests, according to an ostensive-inferential model, that the linguistic expression offers something more than a straightforward interpretation of the propositional meaning of the utterance.For example, a distancing of the self of the speaker from the view being expressed might readily be inferred, leading to an interpretation that what is said is to be taken as a legal judgement rather than a personal opinion.What we are not claiming here is that this is the inference that will necessarily be made by actual participants -whose pretextuality (see above) may lead them in a number of variant directions.Simply, on the assumption that -unless there is evidence to the contraryspeakers seek to assist comprehension of their own utterances, marked ex-

Ian Mason
pression points to the need for inferencing and, in a given context, some inferences are likely to be more plausible than others.
Similarly the Judge selects the expression forza maggiore (force majeure) three times (lines 2, 12 and 15) in three successive turns at talk, an ostension that makes this particular lexicalisation salient within the exchange.Here too an inference that might well be drawn is that what is referred to is a legal category that applies only under certain conditions, the loss of documents (cf.line 1) not being one of these.Now, it is noteworthy that, in respect of these two examples of ostensive behaviour, the interpreter's ostension does not offer either of the inferences suggested above."The judge doesn't believe…" (line 5) is more likely to suggest a personal opinion while "[not] good enough reason" (line 6) also suggests a personal assessment rather than a legally defined category.From a methodological point of view, it is important to note that we cannot reliably state either (a) that the interpreter has not made the inferences or (b) that the interpreter has made the inferences but has other communicative priorities which override these.The latter is contextually plausible, likely even, but in the absence of access to the interpreter's thought processes, we do not have conclusive evidence.In other words, we can show evidence of ostensive behaviour but we can only suggest possible inferences, except where succeeding turns at talk provide evidence of actual take-up of particular meanings by particular participants.The notion of take-up can be illustrated from the same sequence.The judge's subsequent ostensive behaviour suggests actual response by her to the interpreter's versions of her talk.Cerca di tradurre… ("Try to translate as faithfully as possible"), spoken (line 13) at this particular juncture, implicates dissatisfaction with immediately preceding translations while Gliel'ha spiegato?("Have you explained it to him?" -line 17) suggests doubt.Conversely, Non è una raggione sufficiente ("it is not sufficient reason" -line 21) places on record the judge's final acceptance of the interpreter's renditions: the insistence on forza maggiore has been dropped.The sequence as a whole illustrates a series of re-contextualisations (cf.Pérez González 2006) as each participant responds to each other's conversational moves, co-constructing the meanings that are exchanged.It also shows how far reality is from the stable communicative transaction suggested by code models of interpreter behaviour.

Miscommunication repair and interactive positioning
The interpreter's attention to potential miscommunications was among the topics listed in Section 2.1 above on interpreting as interaction.In our final example, we examine the consequences of interpreters' decisions in such cases, at the same time raising the issue of the distribution of power within the exchange.In this sequence a Polish interviewee in a UK immigration interview is asked why she is not in possession of her passport and why she cannot retrieve it from the person who has it.A superficial account of this sequence would simply note the interruptions of the routine turn-taking sequence (Official > Interpreter > Immigrant > Interpreter > Official, etc.) by the interpreter (lines 3 and 5) for the purpose of clarification.In order to provide a maximally informative reply to the official's question, the interpreter seeks to establish whether the person who has the interviewee's passport has taken it away with him.The Polish woman's negative reply, however, is problematic, seeming to imply that she still has her passport after all (line 4) 7 .This then requires further clarification before the interpreter is able to give a full and coherent account of the situation (lines 8-10).Such a limited account would miss what is at stake here, however, by failing to link the turn-by-turn analysis of talk (CA) to the macro-contextual factors governing the progression of the encounter towards its resolution and the fulfilment of participants' goals.CDA-inspired accounts of immigration and asylum encounters inform us that it is crucial to the immigrant in such encounters to provide a coherent story and, above all, to avoid self-contradiction or inconsistency (see, e.g.Barsky 1994: 225 and1996: 46;Maryns 2006: 1 and passim).At the same time, interpreters are generally enjoined to translate accurately what is said, without embellishment or comment.Thus, a strict adherence to approved procedure would require the interpreter simply to translate the Polish woman's reply in line 4 as "No, I have the passport at my place".Had the interpreter done this, the internal inconsistency in the woman's account could not fail to be noted by the official.Yet as far as the immigrant is concerned, nothing inconsistent has been said since it appears from co-text that she intended: "No, I have the passport at his place".The methodological point here is that, in our de-

Ian Mason
scription of interpreter behaviour, we must consider what is said by interpreters in the light of what might have been said but was not.Only thus does it become possible to fully appreciate the influence of the interpreter's turn on the subsequent direction of the talk exchange.The interpreter's repair of the miscommunication at line 4 is likely to have a determining effect on the unfolding of the exchange (i.e.what happens next) and the overall account that is made available to the immigration official.The actual move positions the interviewee as both consistent and coherent; an alternative move would have positioned her in a far more negative way.The example is a clear illustration of Pérez González's (2006) point about recontextualisation (see above) and suggests that experimental research on the effects of the resolution or non-resolution by interpreters of miscommunications is needed.Beyond that, the example also points to the need, in addition to the local recontextualisation, to access relevant aspects of the wider context.

Aligning macro-and micro-contexts: ethnolinguistic and ethnographic accounts
By way of synthesis of the various arguments put forward in this paper, we can draw together some of the key concepts and distinguish different levels of analysis.Widdowson (2002: 157) distinguishes two major senses in which the term 'discourse' is used (cf.section 2.2 above).The first sense of the term (D1) is the pragmatic process of negotiation of meaning in communication, based on (but by no means reduced to) what is said.RT and Gricean pragmatics deal with this level.Discourse (D2), on the other hand, as "socially constituted… conventions of belief, established values which constrain the way people think and use their language to achieve meaning", echoes Foucault's sense of the term and is the sense widely used in Critical Discourse Analysis and in Cultural Studies.The study of D2 presents us, as we have seen, with some unresolved methodological problems but that should not lead us to abandon the concept.The purpose of discussing Sequences 1 and 2 above was, in part, to suggest the psychological plausibility of D2 and its genuine relevance to the community interpreter's task.
In the same way, it will be apparent from research reported above that we need to distinguish two kinds of context: • C1 "the set of premises used in interpreting an utterance/ a subset of the [user's] assumptions about the world" (Sperber and Wilson 1995: 15), subject to each user's pretextuality and constantly evolving within the exchange (re-contextualisation); • C2 relevant aspects of the socio-cultural/historical context, including especially institutional constraints.
By the same token, it is particularly important in our field to distinguish two kinds of power: • P1 Power within the exchange (gate-keeping rights, the power to interrupt, question, etc. often invested in the interpreter, who occupies a "position[] which control[s] scarce resources" - Anderson (1976: 218).The investigation of D2, C2, P2 is crucial if we are to make sense of community interpreting events but our investigations also need to be informed by evidence of the evolution within actual exchanges of D1, C1, P1.If we limit our consideration of power to P2 alone, we do not necessarily notice what is actually transacted in particular encounters.By conducting pragmatics-sensitive discourse/conversation analysis of talk in community interpreted interaction, we can fill the gap between the turn-by-turn analysis of talk in its narrowest sense and the ethnographic study of interpreters as social beings and of the events in which they participate, as exemplified by Cronin (2002), Inghilleri (2003;2005a/b) and others.