Providing Opportunities for Patients to Say More about Their Pain without Overtly Asking: A Conversation Analysis of Doctors Repeating Patient Answers in Palliative Care Pain Assessment

: As the main symptom in palliative care, pain requires careful assessment. Repeating patient answers is one recommended communication technique for helping convey to patients that they have been heard, and to encourage them to say more. We examined 23 episodes where experienced doctors repeat patients’ answers with mirrored rhythm and downward-ﬁnal intonation, captured in pain assessments video-recorded in 37 consultations in a large UK hospice. Using conversation analysis, our aim was to determine whether or not the repeats invite additional talk, and if so, how they do so. Our ﬁndings reveal lexical and prosodic features of doctors’ repeated pain answers that signal completion of the sequence. At the same time, because the patient has greater epistemic access to their own pain, a repeat can also invite conﬁrmation or discon-ﬁrmation. The patients in our data sometimes—but not always—respond to the repeat with conﬁrmation or further talk. We conclude that repeating patient answers with mirrored rhythm and downward-ﬁnal intonation provides a no-obligation opportunity for patient-led conﬁrmation, disconﬁrmation, or expansion of pain descriptions, particularly when the pain matter is new, revised, or has been problematic to report.


INTRODUCTION
As the main symptom in palliative care, pain requires careful assessment. Repeating patient answers is one recommended technique for making patients feel understood and encouraging them to say more. In this article we explore what occasions a doctor's use of a repeat, whether or not it invites additional talk, and if so, how it does so, questions that have not previously been investigated. In conversation analytic terms, repeated answers occur in thirdposition: after a speaker has asked a question and the recipient has answered, the first speaker then repeats the recipient's answer. We examine repeated V C The Author(s) (2021). Published by Oxford University Press. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/ 4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. answers in one particular context: pain assessment during palliative care consultations. Palliative care is an interdisciplinary approach to specialist care of patients with life-limiting disease, and their families, when the disease is no longer responsive to curative treatments (Twycross 2003). It can be provided in various settings including in hospitals, at home, or as part of community programmes. It is the bedrock of hospice care, wherein the ultimate goal is to improve the quality of life of patients and their families (Faull 2012). Practitioners work with patients and their families to: discuss end-of-life care; manage psycho-social aspects of degenerative disease; support spiritual needs; and prevent and relieve symptoms such as nausea, constipation and especially pain-one of the foremost presenting symptoms in palliative consultations (Potter et al. 2003;Faull 2012;Lovell and Yates 2014). Impeccable assessment of pain is required to address patient suffering and monitor illness progression (Faull 2012;NICE 2015).

PAIN ASSESSMENT AND PALLIATIVE CARE
Pain is a highly subjective experience, understood within palliative care to be influenced by physical, emotional, social and spiritual factors (Nicholson 2012). Pain assessment relies on a careful history taken from the patient, allowing evaluation of the nuanced character of pain: something which can only be achieved in a joint enterprise of co-construction (Nicholson 2012;Noble et al. 2013). Clinical guidance advocates using established pain inventories and assessment tools, as well as asking questions about a range of characteristics, influences on, and implications of pain (NICE 2015). Palliative care textbooks provide descriptions of communication skills or 'active listening' techniques. These are said to help the patient feel understood and encourage them to say more; they include nods, tolerating silences, verbalizations such as 'Mmhm', and repeating parts of the patient's own words (Buckman 2000;Twycross 2003). This guidance is generally based on indirect evidence in the form of post-hoc reports, rather than on in situ analysis of communicative practices (Bostrom et al. 2004;Lucket et al. 2013). Studies using conversation analysis to examine (recordings of) interactions in detail have demonstrated that health communication practices recommended or discouraged in guidance do not always function in the way that guidance suggests; in real-life, linguistic practices are more complex with a range of interactional consequences (Parry 2005;Perä kylä and Vehvilä inen 2003). Conversation analysis is uniquely positioned to provide concrete descriptions of known practices and to interrogate (and, where appropriate, correct) the assumptions of communication practice recommendations (Perä kylä and Vehvilä inen 2003).
To provide new understandings about the functioning of one recurrent form of post-answer repeat, and its particular use as a communicative technique used by experienced palliative care clinicians undertaking pain assessments in a hospice setting, we present a systematic analysis of recorded real-life episodes. Having observed several occurrences of doctors repeating patients' answers with downward intonation during pain assessments, we wanted to explore the implications this had for what a patient could and should say next, and why it is used in this environment.

WHAT IS ALREADY KNOWN ABOUT REPEATS IN INTERACTION?
Before presenting our analysis, we overview current conversation analytic (henceforth CA) evidence about the various actions that people attempt and accomplish through repeats within a range of interactional contexts and languages. In this section, we overview various features of repeats described in the literature-including where they happen within the turn-by-turn back and forth of interaction, and other factors known to be relevant to how they work, including intonation, gaze, and knowledge differentials between the speakers involved. We examine these characteristics and functions to investigate what a patient could and should say following a repeat, narrowing our focus to repeats in third position-these include post-answer repeats.
In general, repeats can be used for a wide range of actions. They can: display disaffiliation or disagreement; do complaining and challenging; be a first step in dealing with talk that has not been fully understood; show that prior talk has been heard; and highlight a particular section of what was said as a matter for further talk (Schegloff 1997;Betz et al. 2013). Working out what a speaker is attempting to do via their repeat, and whether/how to respond to it, is a task for the recipient (Schegloff 1997) through attention to the various features of the repeat. One of those features is where within talk the repeat occurs: its sequential placement (Schegloff and Sacks 1973). CA makes a distinction between repeats according to the repeat's 'position'. 'Second-position repeats' occur after a first speaker has initiated something (e.g. by asking a question, asserting something, extending an invitation). Such repeats can, amongst other things, convey a problem with what has been said (Greer et al. 2009;Bolden 2009).
On the other hand, the repeat can happen after a responsive turn-termed 'Third-position repeats'. This is the kind of repeat we examine: first, a speaker (in our cases the doctor) requests information, often in the form of a question; second, the next speaker (the patient) responds; and thirdly, the doctor repeats the patient's response. 1 In all our examples the repeats mirror the patient's downward-final intonation and the prosodic rhythm of the turn. In Extract 1, which we will examine in more detail later, the doctor asks whether the patient's shoulder is painful. The patient answers 'Just a bi'.', which the doctor receipts with a repeat. Receipting an answer is not the only function of third-position repeats; previous research has clearly established that repeats with similar compositions and in similar sequential positions can implement radically different social actions (e.g. Kim 2002). Third-position repeats may initiate repair, register receipt of what has been said, or identify talk that is rendered preliminary to what follows (Schegloff 1997). A recipient's response options are different for each of these types of repeat. That is, even narrowing down to third-position repeats, we already know that these can attempt or accomplish a variety of actions that have different implications in terms of what the recipient could or should do next. Furthermore, many third-position repeats appear designedly equivocal in terms of whether they are, or are not, produced in the service of sequence closure (Schegloff 2007). In addition to sequential position, there are several turn-design characteristics that affect whether and how the hearer should respond, including: lexical features, prosody, whether in epistemic terms the repeat pertains to something that is best known by (in our context) the patient, and gaze. The next section of our literature review focuses on these characteristics of third-position repeats.

Lexical features of third-position repeats
It is commonly assumed that repeating an answer is one way for a speaker to show they have listened to the first speaker (Buckman 2000). Empirical analyses support this to some extent; repeats demonstrate strong access to the preceding talk, allowing speakers to display rather than merely claim what they have heard, and to demonstrate that it is noteworthy (Schegloff 1997;Betz et al. 2013). This has been observed in everyday conversation when a speaker repeats information such as a phone number or baking instructions, to register that they have heard (Goldberg 1975). It has also been observed in institutional settings such as British English and German speaking focus groups, where moderators' repeats can acknowledge what one participant has said (Puchta et al. 2004), and in police interviews in the U.K., where police officers confirm a suspect's answer via a repeat (Stokoe and Edwards 2008). This sort of neutral acknowledgement does not suggest further talk is relevant. By contrast, empirical research in psychotherapeutic environments in the U.K. has shown that lexical repeats can elicit elaboration and encourage the person who answered to say more (Muntigl et al. 2013). It follows then that lexical features in isolation do not account for the interactional functioning of third-position repeats.

Prosodic features of third-position repeats
Having considered lexical features, we now examine what is known about ways in which recipients understand prosodic features of third-position repeats. Firstly, specific intonational patterns have been associated with particular actions: a repeat that matches pitch in absolute terms, rather than relative register, is treated as mimicry (Couper-Kuhlen 1996); a high rise-fall intonation can be used to initiate repair, specifically, to claim that the repeated talk is 'wrong' or in need of correction (Benjamin and Walker 2013); high global pitch and increased volume is linked to astonishment or surprise (Selting 1996); final-rising intonation makes a repeat likely to be understood as a question (Robinson 2013); and 'list item like' intonation has been shown to elicit further responses (Puchta et al. 2004). Consequently, each of these actions makes relevant a particular response or set of responses. Repeats with downward-final intonation, the subject of our analysis, have been found to be more neutral, avoiding explicit endorsement of the prior information; they avoid claiming to have understood what has been said (Betz et al. 2013). Downward-final intonation is prominent in repeats that register receipt of what has been said (Schegloff 1997) and such repeats have been said to advance the progressivity of the talk rather than assess prior talk (Puchta et al. 2004).
Rhythm and stress can also affect the meaning of a repeat: when a speaker repeats talk with added stress this gets understood as signalling problems with the preceding talk (Stivers 2005). Conversely, mirroring the prior speaker's prosody is recurrently treated as neither endorsing nor rejecting what has been said (Betz et al. 2013). Thus, a range of prosodic features can give recipients clues as to the function of a repeat and how to respond. However, in considering prosody, it is important to note Walker's (2014) general argument on describing linguistic form independently of interactional functions; she notes that prosodic features can be systematically associated with an interactional practice whilst warning against assigning functional labels to prosodic features. Although prosodic cues have been found convincingly to be associated with specific actions, Walker's (2014) caution highlights the danger of ascribing too much interpretive weight to features of prosody on their own. The analytic consequence of this is that repeats with lexically similar composition, prosodic features and sequential positioning still need to be individually examined for what action they are being used to do in their local, interactional context, through examining how recipients respond to them (Schegloff 1997;Robinson 2013).
Whilst we have noted that downward-final intonation tends to bring talk to a close, and that the lexical features of a repeat display that the prior talk has been heard and further talk is not required, Schegloff (1997) notes that many downward-intoned third-position repeats are nevertheless followed by some form of agreement. The epistemic properties of the repeats and the role of gaze shed light on why this is.

Epistemic characteristics of third-position repeats
Epistemics concerns matters of what, and how, speakers claim to know and understand (Heritage 2012). Robinson (2013) argues that how much the producer of the repeat knows about the repeated item affects what the repeat is understood to be doing. For example, whereas in everyday German conversations repeats are not hearable as endorsing the prior information (Betz et al. 2013), in educational settings in the U.S. and Korea, where the teacher asks a question whose answer they already know, repeating the student's answer is seen as evaluating the prior response in some way (Hellerman 2003;Park 2012). Robinson (2013) argues that epistemic considerations are crucial in determining what a repeat is doing. He demonstrates that in everyday conversations in U.S. English, if a recipient of some talk is understood to accurately comprehend what has been said, repeating it with rising intonation is hearable as disagreement. In contrast, if the recipient is understood to have inadequate or inaccurate knowledge of what has been said, a repeat with rising intonation is hearable as indicating trouble with hearing or understanding. Robinson contends that it is through a combination of prosody and epistemic factors that repeat-recipients infer what the repeat is doing, and consequently, deduce whether a response, and what type of response, is expected.
In line with this, Muntigl et al. (2013) used epistemic factors to explain why repeats of clients' talk with downward intonation in a U.K. based psychotherapeutic setting prompted clients to elaborate. Using Labov and Fanshel's (1977) work on B-event statements, Muntigl et al. (2013) argue that the recipient (i.e. the client) has greater rights and access to the referent of the repeat than does the producer; as a result, this kind of repeat is generally interpreted as a request for confirmation. Thus, epistemic characteristics contribute to whether and how a repeat recipient could or should respond.

Speaker gaze and third-position repeats
Finally, we consider gaze. Gaze has been identified as a critical interactional tool in mobilizing response in everyday conversations in U.S. English and Italian: withdrawing gaze can signal talk as complete, whereas gazing at the recipient can elicit a response (Stivers and Rossano 2010;Rossano 2013). Rossano (2013) demonstrates that in everyday conversations participants work to withdraw their gaze as they approach sequence closure, and that sustained gaze overwhelmingly leads to further talk. However, we note that interactants' gaze behaviour has been identified as highly sensitive to what social actions are being implemented (Rossano 2012). For example, recipients regularly look towards speakers as soon as they understand the speaker to be telling an extended story, but they do not necessarily do so when the speaker is asking a question (Rossano 2012). Further, differences in gaze behaviour have been identified across language types (Egbert 1996;Rossano et al. 2009). We do not assume the universality of the response-mobilising properties of gaze as identified in everyday conversations and in specific sequential positions (e.g. initiating questions); however, we reiterate the importance of considering non-verbal behaviours when interpreting the function of a turn. We note that in addition to the lexical and prosodic features of third-position downward-intoned repeats that indicate that no further talk is relevant, there are also verbal and embodied aspects of such repeats that may prompt a response.
In summary, the various characteristics of any specific repeat, including sequential position, lexical construction, prosody, relative access and rights to knowledge of the referent, and gaze, each contribute to the recipient's understanding of what the repeat is doing, and to whether they are expected to respond and, if so, how.

RESEARCH AIMS AND QUESTIONS
Our aim is to better understand the practice of doctors repeating patient answers during pain assessments, including the interactional implications for what the patient may or may not do next. We aim to describe and explicate the linguistic and interactional design features that produce this ambiguity in whether further talk is relevant. One reason for doing so is to contribute more granular understandings of repeats, with one of the aims being to inform healthcare communication advice in general and palliative care communication in particular. We consider the following questions: • How do the linguistic and interactional features of third-position repeats with downward-final intonation invite or discourage a response? • Do these repeats promote confirmation/elaboration from patients? Or do patients treat them as closing the local sequence? • What can this analysis tell us about why such repeats are used in pain assessment sequences?

DATA AND METHODOLOGY
The data come from 33 video and four audio recorded consultations between five experienced palliative care doctors, 37 patients, and-in 16 of the consultations -companions who were usually relatives, recorded at a large UK hospice in 2014. Speakers predominantly had British English as a first language. To capture consultations, two video recorders were placed either side of the room, and a voice recorder on the desk. Ethical approval was obtained from National Research Ethics Service Committee Coventry & Warwickshire, UK. All patients had a life-limiting diagnosis; all participants granted written informed consent. A total of 21 hours and 48 minutes of recordings were made. The consultations lasted between nine and 77 minutes, with a mean of 35 minutes. The recorded consultations were initially transcribed verbatim. All identifying information (person and place names) were removed from the transcripts reproduced in this paper; specifically, in the transcripts we refer to the participants as patients ('Pat'), companions ('Com'), and doctors ('Doc'). The data were analysed using Conversation Analysis (CA) (see Sidnell 2009). This research approach examines how participants do things through talk and non-verbal behaviours e.g. gestures, gaze (Schegloff 2007(Schegloff , 1995Sidnell 2009). It focuses specifically on participants' actions e.g. greetings, telling a story, asking questions, and considers the different ways in which people can implement those actions, and the differential consequences that implementing actions in different ways can have on the unfolding interaction (Schegloff 1996). Due to CA's detailed and highly empirical approach to explaining interaction, it is particularly suitable for the study of communication practices-and provides understandings that cannot be gleaned from post hoc accounts and reports about communication.
We made this particular collection by: (1) identifying all episodes involving reference to pain and related sensations; (2) isolating sequences in which the doctors solicit information (we found 180); (3) identifying each episode where the doctor repeated the patient's response; (4) we excluded paraphrased repeats, repeats later than next turn, those with upward/questioning prosody (because these are known to function differently, e.g. Kim 2002;Robinson 2013), and repeats where the doctor continued their turn immediately after the repeat-these repeats are known to target next action and are treated as a preliminary to what follows. This resulted in a collection of 23 repeats.
We sought to explicate the linguistic features of the third-turn repeats we analysed, and to gain evidence on whether or not and why they were treated by the recipients as response worthy. This involved closely watching/listening to the recordings alongside verbatim transcripts to identify episodes in which pain was talked about. We homed in on the doctors asking about pain: a coherent collection where all were observably similar in their action. For this analysis, we focused on one particular practice: doctors repeating the patient's answer. We examined each extract on its own merits, using internal evidence for how participants understood each other. We relied on previous study findings across a range of languages and contexts but did not automatically assume that they applied to the present data. As is conventional in CA, we examined in detail the properties of these repeats through repeated watching of the recordings, supported by producing detailed transcripts including aspects of speech delivery such as intonation, emphasis, silences, and overlaps (Jefferson 2004) and through group data analysis. ANALYSIS Our collection comprises full or partial repeats of the patient's pain description, taking place within sequences of talk where the doctors seek information about pain. These repeats match the rhythm of the patient's turn in terms of prosodic delivery, and imitate the turn's downward-final prosody (with relative register, rather than in the absolute terms of mimicry; Couper-Kuhlen 1996).
In the analysis that follows, we demonstrate that patients sometimes do, and sometimes do not respond to these doctor repeats. We demonstrate how these repeats provide neutral acknowledgement and work to register what the patient has said. We also argue that these repeats give the patients the chance to say more, because they refer to the patient's pain, something to which they clearly have the greater epistemic access. In addition, we show features of the doctors' nonverbal conduct that can encourage the patient to respond. Therefore, the repeats are equivocal in terms of whether they invite a response. We argue that this ambiguity makes these repeats clinically useful because they provide opportunities for the patients to say more but also make it easy for the patients not to do so, without having to directly decline.

WHEN PATIENTS DO NOT RESPOND TO A DOCTOR'S REPEAT: TREATING DOWNWARD-INTONED THIRD-POSITION REPEATS AS CLOSURE RELEVANT
Our first example is one of the eight repeats within our collection of 23 wherein the patient does not treat the repeat as response worthy, either verbally or with an embodied action. It comes from an outpatient consultation with a male patient with cancer. He has restricted movement in his right shoulder.
During an initial discussion about the patient's work situation, the patient puts his left hand inside the sling which holds his apparently painful right arm. We join the talk as the patient nods towards the sling issuing an apology (presumably for attending to his arm rather than solely the doctor). The doctor comments on the patient's holding of his arm, and the patient provides a reason: that it is throbbing, and the doctor repeats this. .HH Just to find out a bit more about you:, ((The doctor asks about practical support available to the patient)) At line 4, after accepting the patient's apology, the doctor observes that the patient is 'holding your arm at the moment.' (lines 4-6). The patient explains in line 7, ''Cause it's throbbing.'. This pair of turns could stand alone. However, in line 8, in third position, the doctor repeats the patient's response.
We begin by describing the features of the repeat that indicate that the sequence is closed. The repeat is lexically identical, mirrors the patient's downward-final intonation, and has similar prosodic emphasis, or rhythm. As noted in our literature review, prosodically matched repeats are neutral; that is, they are recurrently treated as not taking a position on what has been said (such as endorsing or rejecting it), and they are also not treated as claiming understanding of its import (Betz et al. 2013). The doctor's repeat instead conveys that she has registered new knowledge (which she invited the patient to provide) and can be heard as advancing the progressivity of the talk (Puchta et al. 2004).
Following the doctor's neutral acknowledgement of the patient's answer, the patient neither responds verbally nor non-verbally, such as with a nod (line 9). Next, the doctor and nurse speak in overlap; the nurse acknowledges the patient's response with a confirmation 'Mm.' (line 10) as the doctor says 'Mkay' (line 11), a word that treats the prior response as adequate, but also has a forward facing character, prefiguring movement towards what will be talked about next (Beach 1995). In the 0.9 seconds of silence that follow, the patient does not treat the repeat or the 'okay' as response worthy.
In this extract, the patient has several opportunities to respond to the repeat. Although there is only a brief silence on line 9, the patient could have projected the doctor's turn completion on line 8 and initiated a turn either in the small silence or in overlap with the nurse and the doctor's competing turns (and in fact, we will show that this does occur in other examples in our collection). Also, the patient does not talk further at line 12. Extract two has thus illustrated that these third-position repeats can be, and are, treated as not response worthy. Noticeably, the doctor does not overtly pursue a response from the patient, nor does she treat its absence as 'missing' or problematic.
On the other hand, a close look at extract two reveals that there are nevertheless features of the repeat that could be taken by the patient as making confirmation or elaboration relevant (whilst not obligatory). The doctor and nurse both opt to take short turns (lines 10 and 11), but these leave the floor open again at line 12. If the doctor's repeat was simply receipting the patient's talk, it would make sense for the doctor to carry on with the activity in progress, without leaving this expanded transition space. Instead, a 0.9-second silence develops as none of the parties (doctor, patient, nurse) opts to take a turn. What is more, during this silence, the doctor maintains her body position and eye contact-behaviours known to encourage the gazed-at interlocutor to take a turn in everyday interactions (see Stivers and Rossano 2010). Thus, the doctor's bodily and verbal behaviours provide an interactional opportunity to the patient to respond to her repeat, even though he opts not to.
Turning to further features that can be taken to encourage the patient to respond to the repeat, we consider the difference between the patient's and the doctor's knowledge about the patient's pain. The doctor is attending to the patient's pain, and the patient clearly knows more about the way his arm is throbbing; it is a domain over which he maintains primary epistemic rights. In therapeutic settings, statements about a client's experience or sensation tend to be treated as requests for confirmation and are used to solicit elaboration. There are also contextual ways in which the repeat emphasises something worthy of elaboration: the patient is highlighting arm pain for the first time. In the consultation, the patient initially raised the issue of reduced function in his shoulder and arm. When asked further, he talked about pain in the shoulder, which the doctor promises to revisit later in the consultation. In the extract, throbbing in the arm is first highlighted, but the patient provides little detail about this new sensation. The repeat explicitly acknowledges the throbbing, and it does so in a way that provides the patient with an opportunity to give further information.
Extract two has illustrated the cases in our collection in which the patient does not respond to the doctor's repeat. However, there is a feature of extract two that is atypical in our collection; unlike all the other cases, it begins with the doctor reporting an observation, rather than the doctor asking the patient a grammatically formatted question. For that reason, we now turn to an example of a patient not responding to a repeat where the sequence begins with a grammatical question. In extract three, the patient is visiting the outpatient clinic with stomach pains. After discussing medications, the doctor begins more detailed questioning regarding the nature of the pain, beginning with its location. So, the doctor seeks information about the location of the patient's pain with an interrogatively formatted question (lines 1-5). At the same time, he conveys an assumption about this location, by placing his hands on his own stomach. The patient responds by conveying the location of the pain with his hands, whilst saying 'It's about there' (line 7). The doctor repeats 'About there.' (line 10), mirroring the patient's downward-final intonation with the same prosodic emphasis and moving his hand to demonstrate a revised understanding of the patient's pain location-that is, he indicates a change in his knowledge as a result of the patient's response. A 0.5-second silence builds (line 12) in which the doctor maintains gaze towards the patient. The patient does not take the opportunity to respond, and nor does the doctor pursue further information about the location of the pain. The doctor then continues his line of questions regarding the nature of the sensation (line 14) prefaced with 'and'-a common practice in medical encounters for tying together a series of related questions (Heritage and Sorjonen 1994).
So far, we have pointed out the contribution of the doctors' bodily behaviour, and the fact that the doctor allows a silence to develop after the repeat. We now turn to epistemic and contextual factors. Like extract two, the doctor is topicalizing the patient's pain, a domain over which the patient maintains primary (and greater) epistemic rights. As such the repeat may be treated as a request for confirmation. There is also a contextual sense in which further talk is made relevant: the doctor repeats the patient's response in a way that demonstrates that the doctor has a new or amended understanding of the location of the patient's stomach pain, and by demonstrating his revised interpretation he makes confirmation from the patient relevant.
In both extracts two and three, the doctors' post-answer repeats mirror both the rhythm and lexical format of the patients' responses. In this way they receipt what has been said without marking it as problematic and requiring clarification or elaboration. This makes it hearable as sequence-ending. On the other hand, both examples entail features that could be taken by the patient as making a response relevant: sustained gaze, allowing a silence to build, plus the epistemic nature of the repeat and local contextual factors that make confirmation or elaboration relevant. In this sense, the repeat provides an opportunity-as opposed to an overt invitation-to (dis)confirm or elaborate, and the silence the doctors leave in both examples expands that opportunity space. However, by not overtly pursuing an elaboration, the doctor avoids putting the patient in a position of having to decline providing further detail. In this local moment, the doctor bestows control to the patient to take initiate in adding or adjusting the pain description. Both the receipting character of the repeats, embodied by the mirrored prosody and the acknowledging character of the lexical repetition, work to make the patient's lack of response unproblematic, and the consultation progresses with the doctor's ongoing series of questions.

WHEN PATIENTS DO RESPOND TO A DOCTOR'S REPEAT: TREATING DOWNWARD-INTONED THIRD-POSITION REPEATS AS RESPONSE WORTHY
Like the two extracts in the previous section, all 23 instances in our collection are repeats of the pain description occurring in third position, with downward-final intonation and mirrored prosody, and they relate to something to which the patient has primary epistemic access (e.g., their pain experience). Despite these commonalities, there are 15 cases in which the patients do treat the doctor's repeat as response worthy. We found two overall types of patient response: simple confirmations, such as 'yeah', 'yes', and 'mm' (9/15); and more than minimal talk that includes modifications or expansions of the original pain answer (6/15).

Simple confirmations
The female patient in extract four is a hospice inpatient with cancer, experiencing severe pain. She has confirmed that morphine helps and describes having had two injections the previous night. As the extract starts, the doctor asks about the duration of pain relief, then repeats the patient's description of it: (.) 08 Doc: And did it come ba:ck again then.
Before the doctor has completed her question about the longevity of the pain relief, the patient begins to answer (line 3). She responds in terms of how long it 'seemed to be okay for' and shifts from heading towards 'quite a while' to the more specific 'a couple of hours'. At this point the doctor repeats: 'a couple of hours' (line 5). This repeat targets the component that was repaired. Like the two extracts in the previous section, the repeat here matches the downward-final intonation of the prior turn, and the rhythm is mirrored. In this way it enacts a neutral acknowledgement. As in extracts two and three, these features register receipt in a way that does not require further response from the patient. However, as in earlier extracts, it has characteristics which make some confirmation by the patient relevant. The repeat (and the repeated talk) is about a matter within the patient's epistemic domain: reduced or alleviated pain sensations, to which the patient has primary access, and the doctor maintains gaze towards the patient.
The question at lines 1-2 is not the first time the doctor has tried to ascertain how long the morphine's effects lasts. In a previous attempt (before the transcript above), the patient's response did not fit the question: she described procedural matters-that she had two injections the previous night, about 45 minutes apart. Thus, the effectiveness of the pain medication was not at that point known to the doctor. In extract four, the patient's response in lines 3-4 better fits the agenda of the doctor's question: she provides information about how long the pain relief lasted for. She thereby gives the doctor previously missing information, and the repeat can be hearable as acknowledging or marking the patient's response as clinically relevant/noteworthy (Puchta et al. 2004). In receipting the patient's answer, the repeat contributes to ongoing intersubjectivity (that is, conveying mutual understanding between the patient and the doctor) about the nature of the pain and its properties. The patient's immediate 'Yeah' response conveys an accepting of what the doctor has said, and a re-confirming of her own response. This signals the particular sequence as complete (Schegloff 2007;Rizwan-Ul and Alia 2015). Indeed, what happens next is something new but related: following a brief silence (line 7) the doctor issues a next question (line 8), as we saw in extract three, linking it with a series of related questions with the 'and' preface.
Our next example is a further case where the doctor is asking a question about pain relief, but this time relating to side effects the patient experienced. Like extract four, we see the patient responding to the repeat with confirmation. We join the consultation 11 minutes in, as the doctor asks the patient about the effects of recent reduction in pain medications from a dosage of twelve down to nine. The patient reports that she had been experiencing side effects on the higher dosage, and the doctor seeks more information about these. The patient describes sickness, headaches and sleeping more, but then asserts that she was in less pain on the higher dose. The doctor repeats this. As the patient lists several negative side effects (lines 7 to 11), the doctor acknowledges with 'Right.' and 'Right. Okay.' (lines 10 and 12). After a silence, the patient expands her response by asserting that she was in less pain on the previous, higher, dose of pain killer. The doctor repeats the aspect of line 14 that is new information: 'In less pain.' This repeat mirrors the patient's prosody and downward-final intonation, embodying an acknowledgement of what has been said. Like the previous example, the turn that the doctor repeats is a patient response that conveys new and relevant information: here, that the patient's pain levels have changed. She was in less pain on the higher dose of pain medication, and by consequence, is currently experiencing more pain. The repeat is hearable as acknowledging or marking the reported pain level as clinically noteworthy (Puchta, Potter and Wolff 2004). The patient responds with a simple confirmation 'Mm.' (line 17). The doctor begins her next turn on line 18, pursuing further details about the patient's current pain.
In the two examples in this section, the patient responds to the doctor's repeat with a simple, brief confirmation, and the doctor moves to a new question. This contrasts with the first two examples which, despite having identical lexical, prosodic, epistemic and non-verbal properties, do not lead to the patient responding. There is no evidence that either form is treated by the doctor as problematic.
More than minimal talk after a repeat: modifying or extending a response Extracts four and five illustrated that the patients can respond to these repeats with a simple 'yeah' or 'mm' confirmation. We now move to those cases (6/ 15) where the patient produces a more than minimal response following the doctor's repeat. In Extract six we return and expand upon the excerpt we examined in our introduction. It occurs within the same consultation as extract two, wherein the patient reported arm throbbing pain. The patient is visiting as an outpatient and has cancer. At the outset of the consultation, he talks about reduced movement in his right shoulder. We join the talk as the doctor asks whether it is painful and repeats the patient's response, following which the patient expands on his answer. The doctor enquires about whether 'it' is painful at all (both context and gesture make it clear that 'it' is the shoulder; lines 1-3). After a brief silence, the patient responds: 'Just a bi'.' (line 5). The doctor repeats this, with identical lexical formulation, mirrored prosody and downward-final intonation (line 6). As with the other repeats, this one could be taken as sequence-closing in that it acknowledges the patient's response. Its matched prosody does not highlight any aspect of the response as in need of clarification or adjustment. Again, the repeat is also hearable as making confirmation or elaboration relevant by virtue of the doctor's gaze towards the patient, and by referring to something that the patient knows most about-his pain experience.
Although the patient has conveyed that impaired shoulder function is his main problem, he has not, until this point, mentioned pain. The doctor's repeat, singling out a matter and treating it as worth elaborating-indicates a doctorable concern has arisen. By repeating it, the doctor provides an opportunity for the patient to say more. The patient immediately treats the repeat as response worthy; after initial confirmation ('Yeah', line 7), he quickly moves to elaborate on the nature of the pain experience with an analogy describing the pain as 'like somebody pullin' yeh arm out a socket.' (lines 7-9). This implies a significant level of pain. The patient's initial response reported a somewhat moderated/minimised presence of pain. The repeat prompts the patient to detail the nature of that sensation, using an analogy which conveys greater severity. Thus, the patients may treat the doctor's repeat as worthy not only of a confirmation, but in some of our six instances of more-thanminimal-talk post repeat, as a prompt to tweak or add to their description of the pain. Modifications or expansions of the prior turn can take several forms, and one of these is to adjust the intensity of the pain. Because we had only six examples though, it is beyond the scope of this article to explore in further detail what the expansions do; a larger collection of more-than-minimal-talk following a repeat would allow us to pursue such question more confidently 2 .
So far, we have analysed examples of patients providing no response, simple confirmations, and more than minimal talk following doctors' repeats. In each example, and the majority of repeats in our collection, the doctor does not introduce anything additional to what the patient has said. However, we observed six occasions across the full collection of 23 when a repeat was coupled with another practice 'yes'/'yeah' or 'okay'. In CA terms, a turn in third position that combines practices is termed a composite (Schegloff 2007). We examined whether the patients treat composite turns differently to standalone ones.

A NOTE ON COMPOSITE TURNS THAT INCLUDE A REPEAT
We discuss our analysis of composite turns via two extracts. In extract seven below, the doctor is asking an outpatient with head and neck cancer a series of questions about his sore, sticky eye. The extract begins as the doctor asks about the occasional sharp pains the patient has reported. Note that the patient has cancer-related difficulty articulating. In response to the doctor's: 'ho:w much of a problem's the . . . shooting pain' (lines 1-2) the patient initially suggests that the pain is not a problem (lines 3-4), then describes the speed at which it travels ('qui'e fas'', line 8), and following the doctor's continuer, he provides a numerical description of 'six or seven' (line 11), which revises and upgrades his earlier (lines 3-4) indication that the shooting pain was not really much of a problem. The doctor repeats the patient's quantified response in line 13, prosodically mirroring the patient's downward-final intonation, followed immediately with 'yeah'-this acknowledges the patient's on-going talk at line 14. Noticeably, the patient produces a response to the repeat as soon as the standalone repeat is completed (and thus in overlap with the doctor's 'Yeah'). That is, he evidently treats the repeat itself as worthy of a response (line 14), and he goes on to provide both confirmation and additional information-that the scores of six and seven relate to a duration of 'seconds', after which the pain subsides.
In our final extract, we examine a composite that includes 'okay'. 'Okay' more clearly signals a closing of what is being talked about; it marks both acknowledgement of the prior utterance and a move to next topics (Beach 1995). Prior to extract eight the doctor has suggested a numerical pain scale, and the patient has responded that he has never 'got [his] head round that one'. We join them as the doctor asks the patient to scale his pain intensity as mild, moderate or severe. The patient reports 'moderate', a response which the doctor repeats, followed by 'Okay'. The doctor's repeat of the pain descriptor 'Moderate' mirrors the patient's downward intonation, and the doctor follows immediately with 'okay'. The repeat acknowledges receipt of the pain intensity description that was not forthcoming via the numerical scale, before shifting to a new sequence. Whilst the addition of 'okay' could be treated as closing and moving to next matters, the patient treats the repeat as response worthy, issuing a revised version 'upper moderate'. This amends the list of three options provided by the doctor, modifying his pain descriptor to a more severe level.
We found that composite turns that included a repeat-those produced together with 'yeah' or 'okay' embody a more complex and multi-faceted character and response relevancy. The repeat itself has some characteristics that complete a sequence, and others that offer the patient opportunity to say more. These repeats can be coupled with words that more or less strongly invite more, or that point to a move to a next matter. Reflecting this complexity, in our collection we found that the patients sometimes treat repeats as response-relevant even when coupled with a word that is closure relevant.

DISCUSSION
We set out to examine the practice of repeating patient answers when assessing pain in palliative care, in order to investigate when such repeats are used, and whether/how they invite additional talk. The repeats we examined occurred in a particular sequence of talk-in third position following the doctor's question and the patient's answer, where the repeat mirrored the patient's prosodic rhythm and was uttered with matching downward-final intonation and rhythm. Our first objective was to identify whether or not they encourage the patients to say more. We found that some repeats in our collection were treated by the patients as closure relevant: in some cases, the patients did not respond to a repeat, and the sequence continued without either party providing any indication that this was problematic. However, elsewhere, the patients treated third-position repeats as response worthy and provided confirmation of, and/or expansion or modification of their pain descriptions. Patients sometimes, but not routinely, responded to the doctor's repeat of their answer. Examining how the doctors treat the patients' responses to repeats provides evidence as to whether the doctors are deploying repeats as a means to elicit response. What we found was that neither responses to repeats, nor absence of responses to repeats, were oriented to by the doctors as problematic or insufficient. Thus, while these repeats provide opportunities for the patient to confirm or say more, they neither oblige nor rule out responses from the patient.
In setting out to understand the functioning of repeats in the context of pain assessments typically dominated by doctor-initiated questions, we note that they are produced in response to new or revised pain concerns, or when the doctor has had some difficulty eliciting pain descriptions. In these contexts, the built-in ambiguity as to whether a patient should respond brings a particular benefit-it conveys an interactional sensitivity in terms of whether the patient has anything further to add. It acknowledges the patient's answer whilst simultaneously presenting an opportunity for confirmation or revision. It holds the door open a little longer, but with no obligation, before the doctor initiates something new. The no-obligation nature of the repeat means that the patients with nothing further to add need not account for saying nothing more. The repeat cedes agency to the patient to contribute further information in the otherwise doctor-driven series of questions, in circumstances where the patient may, or may not, have further things to add regarding a pain matter that is new, revised, or has been problematic in terms of achieving shared understanding.
In our analysis, we also sought to identify how the linguistic and interactional features of third-position repeats with downward-final intonation function to invite or discourage a response. Drawing upon prior research on repeats in a range of languages and settings, we examined characteristics previously found to be associated with sequence closure or inviting further talk. We did not assume these characteristics would function in the same way in our context; rather, we used internal evidence (specifically: what the doctor did after the slot wherein the patient responded or not to the doctor's repeat) to establish whether this form of repeat is built to invite a response. As we have noted, our findings in this specific clinical context contribute to a wider understanding of third-position repeats and are available to be explored more broadly across different settings, and in other languages.
Our rigorously generated observations show that these repeats entail a configuration of some features that encourage response, and others that discourage it. Certain turn-design features found consistently across all our cases worked to convey completion of the sequence. First, by lexically displaying strong access to the preceding talk, these repeats indicate acknowledgement and can thereby convey sequence closure. In some of our cases, the doctors and the patients collaboratively treated the repeat as just that. This confirms some prior CA findings in which repeats work to confirm an answer, both in everyday talk (Betz et al. 2013), and settings where the institutional business is to elicit and record recipient views or descriptions for the purposes of focus group research (Puchta et al. 2004) or as part of police testimony (Stokoe and Edwards 2008). Second, prosodically mirroring the downward-final intonation and rhythm of the repeat works in such a way as to indicate the patient's response as unproblematic-not requiring further clarification or expansion and thereby conveying the prior talk as neither endorsed nor problematic (Schegloff 1997;Betz et al. 2013).
Other turn-design features found in all our cases worked conversely: towards mobilizing response-as opposed to closing the sequence. In particular, the epistemic characteristic of the repeats in our data make a response potentially relevant; the doctor repeats something about a patient's own experience-in the patient's own epistemic domain (like Muntigl et al.'s 2013 therapeutic data)-and in this context, both prior research (Muntigl et al. 2013) and our analyses show that recipients can treat this as seeking confirmation. Sustaining gaze has also been associated with further talk rather than sequence closure in everyday talk (Rossano 2013). By repeating the patient's pain description whilst maintaining gaze at them, the doctor makes confirmation or expansion relevant, and in many examples the floor is left open allowing a silence to develop. This further provides an opportunity for the patients to respond. If repeats were being used specifically to prompt further elaboration, we would see interactional trouble where the patient does not respond. Instead, what we see is that if patients do not respond, the doctors do not pursue response; if patients do respond, the doctors do not treat this as inapposite. Schegloff (2007) argued that third-position repeats can be designedly equivocal as to whether they are, or are not, produced in the service of sequence closure; our findings illuminate how the repeat's constituent components function to create this equivocality, generating what Betz et al. (2013, p.157) describe as an 'intermediate practice' between information receipt and repair initiation.
This leaves us with a question about variability: how do we account for the unpredictability of how the patients respond to repeats of their answers, if the doctors' repeats have an identical sequential position, and identical lexical, prosodic, epistemic and non-verbal characteristics? Schegloff (2010) argues that deconstructing the constituent parts of a turn only offers part of the picture, and to fully understand variability in response it is necessary to examine the occasions in which the turns inhabit. This leads us to consider how the medical consultation environment, unlike everyday talk, is one in which there is continually sustained task-oriented talk (Schegloff and Sacks 1973). Within pain assessment sequences, a key task is to elicit a description of the patient's pain. This is what has been called the overall course of action, or activity (Heritage and Sorjonen 1994), and is achieved collaboratively over a series of sequences. Repeating a patient's answer opens up a space for the patients to offer more, or adjust their description, or for the doctors to initiate the next (related or new) information request. On the one hand, the response mobilizing features provide for the patients to confirm or expand upon their answers, on the other, the ultimately doctor-led structure of the medical activity (eliciting a pain description) provides for the patients to wait for the doctor to continue with the questioning. Thanks to the co-occurring closure relevant features, and the relevance of a next doctor-initiated question in the overall medical activity, the patient is not obliged to produce any further talk. Thus, the slot following the kind of doctor repeat we examined stands as an agentive moment for the patient in which a response is relevant, but not accountably so.
Our findings have implications for how practitioners can repeat patient answers in a way that acknowledges what the patient says, and prompts, but does not obligate, the patient to elaborate. Firstly, the way the repeat is delivered is important. Repeating the patient's answer in a way that mirrors the patient's downward-final intonation (rather than adopting 'questioning' intonation), and also mirrors the patient's prosodic rhythm, enable the repeat to be heard as displaying that the patient's description has been heard, and not problematically. Secondly, repeating patient answers is valuable at particular moments, such as following the introduction of new/revised pain matters, or when descriptions have been difficult to accomplish. In these moments it is particularly salient to encourage patients to check, or add to, their descriptions. We propose that recommendations to repeat patient answers include this more layered understanding of how repeats function and when they are used.
We have interrogated communication guidance that promotes repeating patient talk and provided a more detailed description of how repeats function. Previous research has shown that repeats have a wide range of functions, and we have highlighted the prosodic, lexical and epistemic features of a repeat which enable the doctor to simultaneously acknowledge what a patient has said, whilst creating an opportunity but not an obligation for the patient to expand upon or adjust their descriptions. Examining recordings of real-life clinical practice provides nuanced and situated understandings of how a linguistic technique, such as repeating a patient answer, actually works in clinical practice.

NOTES
1 See Schegloff (2007) for a fuller discussion on the ways in which sequences can be expanded following a second pair part; this includes minimal sequence-closing thirds, and non-minimal turns including repair (and disagreement-implicated repair), topicalization, rejecting the second pair part, and so forth. Degree signs enclose hearably quieter speech bu-Hyphens mark a cut-off of the preceding sound Yeh, 'Continuation' marker, speaker has not finished; marked by fall-rise or weak rising intonation, as when delivering a list Yeh.
Full stops mark falling, stopping intonation (final contour), irrespective of grammar, and not necessarily followed by a pause Yeh?
Question marks signal stronger, questioning intonation, irrespective of grammar