Collaborative Reflection Under the Microscope: Using Conversation Analysis to Study the Transition From Case Presentation to Discussion in GP Residents' Experience Sharing Sessions

Abstract Phenomenon: In higher education, reflection sessions are often used when participants learn in the workplace. In the Netherlands, all General Practitioner training programs include regular meetings called Exchange of Experiences, in which General Practitioner trainees are expected to learn collaboratively from their own and one another's experiences. Despite this being common practice, we found little research into the structure and process of these sessions. The purpose of this study is to describe the structure and characteristics of group reflection by describing transitions in interactions. We aim to describe the tutor's role in some detail, as this could lead to faculty development. Approach: In medical education, reflection is often approached from a cognitive perspective. However, learning in a group is also an interactional achievement. It is therefore relevant to study the sequential nature of group interaction in collaborative reflective practice. We have used conversation analysis to study the reflection meetings, zooming in on the transition between case presentation and discussion, focusing on the role of each of the participants in these transitions. Findings: The transitions were conversationally complex. Three interactional aspects recurred in the meetings. First, the transitions can be characterized as ambiguous, as there is ambiguity about what will happen next and the floor is open. Second, transitions are an arena for negotiations between case presenter, participants, and tutors, in which knowledge and the right to take the floor (epistemics) play an important part. Third, the tutor can have different interactional roles, namely, that of teacher, expert, facilitator, and active participant. The role of the tutor is important as the tutor's interactional behavior is part of the hidden curriculum. Insights: Conversation analysis focuses on the interaction in group learning and shows how the interaction is part of what is learned and how learning takes place. Transitions are the “messy” moments in interaction yet can tell a lot about the way in which group participants relate to one another. Being conscious of how the floor is taken, the tutor's roles, and the way negotiations take place could help medical educators in the way they shape collaborative learning sessions.

KEYWORDS reflective practice; conversation analysis; general practice; discourse analysis; collaborative reflection The question

Collaborative reflection
Reflection is a current buzzword in medical education, and collaborative reflective practice (reflecting on practice in a group) has a place in many curricula. 1,2 Interestingly, most studies that examine reflection focus on the outcome of reflective processes, yet how these outcomes are accomplished through what actually takes place during collaborative reflection meetings remains a "black box." 3,4 Collaborative reflection meetings are theorized to be a part of an experiential learning cycle in which medical students have experiences in their residency; reflect on them with their supervisor, at the university, or by themselves; and then apply the fruits of their reflection to their next healthcare encounters. An important aspect in this process is the transition from experiencing to reflecting. In this article, we examine the interactional reality of reflection, and in particular how the transition is made from "the experience" to "reflecting on the experience." How is the aspect of the experience on which the group reflects determined in situ?

Exchange of experiences meetings
General Practitioner (GP) vocational training in the Netherlands is a combination of on-the-job training 4 days a week and 1 day of education at the university. In the 1970s, GP training became student centered, whereby residents' on-the-job experiences were the guiding force of what would be taught on the university release day. Each release day starts with an "Exchange of Experiences" (EoE). As the name suggests, one resident shares an "experience" from practice that is taken as the basis of a group discussion. The institutionally prescribed objective of EoE is "broadening and deepening the individual learning experience and feeling mutual support." 5(p9) This study focuses on EoE meetings for GP training in Rotterdam, the Netherlands.
The idea of using cases from practice as the basis for group reflection is not unique. Problem-based learning, for instance, does this too, and there are many methods of peer-to-peer coaching and collaborative reflective practice. EoE, however, is distinct from other approaches, mainly because of the way it is integrated into the curriculum. 6 EoE is usually the first item of the day, and it serves as a bridge between the on-the-job training and the rest of the education day. The group involved in the EoE stays together for the whole day, and for the entire year. This continuity and longevity is believed to create a degree of group cohesion and safety, providing an environment where gradually residents might share more than they initially did. 7,8 Spontaneity is another important distinguishing factor, as there is room for anything that arises-there is no lesson plan or strict format.
The EoE method was inspired by the Balint groups in which the GPs who founded vocational training participated. 5, 9 Balint groups, which still meet internationally, facilitate discussions of patient cases by peers (colleagues) under the guidance of a psychologist. Their purpose is to explore the "side effects" of the "most frequently prescribed drug," that is, the personality of the doctor. Although EoE is based on the Balint method, it has developed away from it in different ways. 10 In Balint groups, traditionally, only patient cases are discussed: case presenter-patient interactions. In EoE meetings, there are also resident-bound cases, such as their relationship with their residency supervisor, their progress as a student, and "first experiences" with particular situations with which seasoned GPs are already familiar. 11 Another difference from Balint groups is that, in Rotterdam, the EoE meetings are facilitated by a psychologist (PST) and a GP teacher (GPT).
Not a lot is known about the actual content and process of an EoE session and how tutors put its educational objective into practice. 5 On one hand, the sessions are characterized by a wide variety in structure, informal organization, and local agenda setting. On the other hand, formal assessment or measurable and specified learning goals are absent. This means that EoE resists traditional top-down coding efforts. 12 The resident provides the material for the discussion by recounting an experience, but it is unclear how the topic for the case discussion is derived from the case presentation and what the role of the tutor is in this transition.

Our goal
In studying what actually takes place during these meetings, we do not aim to check whether the learning goal has been achieved; rather, we want to describe how participants themselves display an understanding of this task and how to accomplish it. We pay specific attention to how the tutors and residents manage the transition from case presentation to case discussion. The case presentation topics are not formally determined beforehand but emerge from the interaction between tutors, case presenter, and other residents during the meeting itself. We want to find out how the agenda for the case discussion is negotiated and what the role of the tutor is.

Conversation analysis
We use Conversation Analysis (CA) in this study. CA is an ethnomethodological approach to the systematic analysis of recorded interaction. Its purpose is to "identify the actions that participants in interaction do and to describe the particular practices of conduct that they use to accomplish them." 13(p78) Its preference is for naturally occurring conversation, talk as it happens in routine, everyday interactions, rather than experimental settings. The starting point of analysis is how participants themselves demonstrate how they understand what is happening, rather than a process of data categorization based on the researcher's assessment. Finally, CA assumes that units of language (sentences, words) cannot be analyzed outside of the conversation of which they are a part, nor can they be understood outside of their social context. [14][15][16] In medical education, CA has been used to study settings 17 such as bedside teaching encounters, 18 doctorpatient communication, 19 and interactions with simulated patients. 20 Maynard and Heritage 19 identified five features of the CA approach, which we illustrate by means of a typical classroom interaction: Teacher: And five times five is, (2) Student: Twenty-five. Teacher: Very good.
First, utterances are seen as social activities. Even though grammatically the teacher's utterance is an unfinished sentence, it is recognized as questioning. Then the student is answering, and the teacher evaluating. We know this because of our knowledge of the world: our experiences with schools, our understanding of what teachers and students are expected to do in a classroom, and so forth. Analyzing what kinds of social activities participants perform provides insight into how they give shape to, in this case, learning in school. CA developed from ethnomethodology, a branch in sociology adhering to the idea that "we can only make sense of what is said in conversation of we know the social context, which does not appear in the words themselves." 21(p135) Second, patterns in conversations are of interest, such as turn taking 14 and sequentiality. 22 Different types of conversations might have their own characteristic pattern of sequences. The preceding extract is an example of a sequence that is the hallmark of classroom interaction: the Initiation-Response-Evaluation sequence. Here, the teacher initiates the sequence with a question, the student responds, and the teacher closes the sequence with an evaluation that indicates whether the student has given the required answer. 23,24 Third, interactional details are seen as the site of order and organization. The focus is not just on what people say but also on the way in which they deliver it. This leads the analyst to specifically focus on features of talk such as silences, overlap, and intonation, and this is why we use detailed transcripts. In the example, a rising intonation at the end of the teacher's sentence and the pause indicate that her turn is "designedly incomplete" so as to elicit a "knowledge display" by the student. 25,26 Fourth, the analysis is grounded in participant orientations. Because we cannot read one another's minds, participants in a conversation show whether and how they understand one another. Each turn is a here-andnow interpretation of the activity in which they are engaged. The fact that the student gives an answer treats the teacher's turn as a question, and the teacher's evaluation (instead of, e.g., "Let me finish!") confirms this understanding of the situation and shows that the teacher was already in possession of the requested information. This allows us to characterize it as a known information question. 24 This is the next-turn-proof procedure: 14 the understanding that participants display of one another's turns is available to the analyst as well.
Last, the CA procedure involves moving back and forth between close analysis of individual extracts; a broader view of the entire data set; and conversational phenomena, practices, and patterns described in the literature. One such conversational phenomenon that proved to be relevant in our analysis is that of epistemics. Epistemics has to do with "territories of knowledge": 27 what each speaker has come to know, wants to know, or already knows. For instance, if a patient tells a doctor about a headache he or she has, the patient has "privileged epistemic access" 28 to the topic at hand: The patient is the only one in that situation with the knowledge about experiencing pain. In education research, epistemics has been described extensively, typically in situations where the teacher prompts the student to give the "right" answer, such as in the preceding example. In EoE however, the epistemic landscape is complex.

Conversation analysis and education
CA studies on education have described how educational activities are achieved through teacher-student interaction, considering this a collaborative accomplishment of both. 23,24 Most research on education, however, is on frontal teaching in primary education, showing, for instance, how the asymmetrical student-teacher relationship is constructed by both parties throughout the interaction. 29 Studies on university and postgraduate learning are still rare, and we discuss the exceptions here.
In her studies of university tutorial talk, Stokoe 30 identified segments of the talk that are normally considered "offtopic" as important sites for identity construction. Whereas educational research traditionally focuses on the parts of discussions where learning takes place, Stokoe's analysis revealed that task-setting sequences are an important site for students to define their relationship to one another and to the task at hand. These social considerations are relevant for group interaction, and for learning, as was also shown by Benwell and Stokoe, who did a number of studies on university tutorials. 31,32 They found that an important aspect of student peer groups is the attention to expected attitude or "saving face," for example, resisting a straightforward orientation toward academic identity. 31 Displays of enthusiasm toward the academic task, for instance, are censored by students. Although the tutor is shown to accommodate to this resistance, this study calls into question the view that tutorial talk is led by the tutor alone, rather than it being a group process.
Another line of study is the work on problem-based learning, which provides insight into how the group process in problem-based learning meetings contributes to the outcome. 3 Koschmann has examined how a learning issue is generated and how surgeons learn from their supervisors, by approaching "'the learnable in the lesson' [as] an interactional accomplishment of both the instructor and instructee." [33][34][35] He described how specifying the exact topic of the discussion takes up much of the conversational work. Any student can renegotiate the boundaries of a discussion topic, and the tutor provides implicit endorsement of certain topics as worthy of further exploration.

TEACHING AND LEARNING IN MEDICINE
Finally, Harris et al. and Waring examined reflection in teacher education. In a study on reflection in the professional experience, Harris showed that mentors drive interaction through the initiation and framing of topic sequences but that this can only be done with the collaboration of the mentees. 36 Waring explored how mentors can stimulate reflection 4 but also identified assessment and advice as two mentor practices that can trigger unsolicited reflection. 37 In conclusion, CA work in higher and vocational education shows that there is a degree of conversational asymmetry between teacher and student, in which group dynamics and face are very important and might be established when the interaction seems to be off-topic. As is usual in CA studies, we refer to further relevant literature in the analysis.

Data set
To capture the richness and the dynamical nature of educational practice, we used video data. In 2010-2011, we recorded 47 sessions of 13 groups, resulting in 76 hours of video recording. The sessions, comprising five to 14 residents per group, were recorded with two cameras. The researchers were not present during the recording. The study was approved by the Ethical Review Board of the Dutch Association for Medical Education. Anonymity was guaranteed, and participation was voluntary. Groups were recorded only if all participants consented.
This data set was recorded for the larger project of which this study is part, in which we describe the overall structural organization of EoE and make an inventory of the content of discussed cases. In the present study, we give an in-depth microanalytical account of one section of an EoE session. For this purpose, we selected fragments from two case discussions from different EoE sessions. The analysis we present, however, is informed by a view of the whole collection.

Transcription
The recordings were transcribed in Transana 38 software, which allows simultaneous viewing of video and transcript. Five different transcribers transcribed the recordings ad verbatim, resulting in approximately 2,000 pages of transcript. Two researchers checked the entire data set for transcription accuracy. In addition, 5% of the recordings were transcribed a second time and compared to the first version. Extracts selected for further analysis were transcribed in detail using Jeffersonian transcription. 39 (See Appendix for the transcription key.) These transcriptions include paralinguistic elements such as pauses, hesitations, and body language. Extracts discussed in this article were translated into English with the help of a native English speaker with fluent Dutch, trying to capture the literal meaning as closely as possible.

Our focus in the analysis
We observed a wide variation in how the meetings are structured. There are, however, a few common factors: Each EoE meeting starts by finding out who has something to share. Once the case presenters are selected (residents are not required to prepare their case), the cases are discussed one by one. Many aspects of an EoE session are locally determined: how the cases are selected, the number of cases discussed, amount of time per case, and when a case is finished. Generally, one resident presents a case, which is followed by a group discussion, and after that, another case is presented.
The task for the discussion is set in the transition from case presentation to case discussion. However, this is not done formally. From the outset, it is unclear how exactly a discussion topic is selected from a case presentation. Even if the resident has a clear question, the discussion topic is rarely limited to answering this question. We deliberately chose extracts in which a transition is made, rather than the case discussion itself, because we are interested in the tutor's role in structuring an EoE session. This relates mainly to deciding what is going to happen next. In our study, we did not look specifically at differences in the tutors' backgrounds, because CA only looks at what emerges from the data. We note that the conversational roles that we discuss here are not exclusive to either the GPT or the PST. GPTs, for instance, make therapeutic interventions just as PSTs do on medical matters.

The analysis
Descriptions Case A description: Sibil tells about her encounters with a 21-year-old patient she describes as "a borderliner," who came to see her about a medical issue but then asked two additional questions. Sibil answered the first of these. The patient then asked for a referral to a psychologist; Sibil responded by saying she would need to see him again before giving a referral. Instead of seeing Sibil again, the patient tried to get the referral over the phone and got upset when Sibil refused. Sibil later found out that the patient's mother had already received approval for the referral from Sibil's supervisor. Case B description: Lars had been ill the previous days. Last night, his supervisor called him and questioned if he was actually sick enough not to come into work. Visibly upset about his integrity being called into question, Lars shared his feelings about this phone call with the group.
Case A: From case presentation to case presenter During the case presentation, the presenter is always the primary speaker. The presentation takes the form of a narrative that is mostly, but not entirely, a monologue. In the beginning, as the presenter is sharing the case, we see a lot of clarifying questions like this one: Extract 1, 1:11:03-1:11:21 SIB I thought like ok? I-it is just stated after all. I am just n-after the next. I will just see him. so I am just going to really clear // it up. PST //and how many days was that after that first visit? how-SIB well quite a while, a few weeks after. GPT so weeks. SIB yeah PST ok after SIB so eh he comes in.
In line 3, the PST's question requests information about a particular detail of the story. It is a question from a listener to a narrator. The latter knows the whole story; the former learns it in parts. Another way of saying this is that, in the current interaction, Sibil is in a more knowledgeable state (KC), whereas the GPT, along with the rest of the group, is less knowledgeable (K-). 28 The question is treated as an intermezzo in the story: Sibil is narrating in lines 1-2, and continues narrating in line 8: "so he comes in." As Sibil reveals more details of the story, the part that has been revealed can be used as a resource by the group, and so the presenter's access becomes less privileged: it becomes shared knowledge. This allows the tutors to intervene in a different way (lines 3-8): Extract 2, 1:12:21-1:12:39 SIB it was kind of a bronchitis situation so it was right that I thought like #ok PST what makes it / that you a border, it is just a normal patient isn't it // so what makes it suddenly SIB //it is not a normal ((group laughs)) SIB it is not a normal patient // and then PST // what makes this a borderliner ((group chats)) SIB you could // GPT //third question SIB yes and then that "third question arose PST "ok #ok. SIB there "there it GOES right? So far/ it is just the () but now that third question PST /yes In this extract, the tutor contests that, on the basis of the descriptions that Sibil has given so far, the patient should be characterized as a "borderliner," which Sibil had said earlier was the topic of her story. The tutor makes use of his expert knowledge of how a borderliner, or at least a patient with a personality disorder, typically behaves, and he diagnoses that what he has heard so far suggests that the case is about a normal patient. The tutor claims to be knowledgeable to some extent ("it is just a normal patient isn't it"; line 3), perhaps more knowledgeable than Sibil, who is telling the story.
Instead of answering the question, Sibil challenges the tutor's claim by resorting to her privileged epistemic position as the narrator. It is interesting that she incorporates the tutor's challenge in her narration, making another epistemic claim (line 14): "so far," that is, on the basis of what I have told you up to now, you might think that this is just a normal patient, "but now" the third question will reveal that he is really not. This marks the GPT's position as a listener again. Sibil continues the story and tells how, in his "third question," the patient asks for a psychologist referral. After 2 minutes, the GPT uses his role as a discussion leader to intervene: Extract 3, 1:14:33-1:14:55 SIB so his mother was then eh so that morning. That was eh the day after he had been to see me, been to see doctor {name GP supervisor} PST wait a second ho ho ho this is going a step too // far SIB //yes yes yes, no // he said PST //but I want to just structure something// SIB //because "no because it is important for what he/ will bring up later, so he calls the doctor's "assistants. for the neurologist's letter. and did I write the letter for the psychologist yet, and on that morning his mother went to see my GP trainer. PST /yeah In line 1, Sibil starts a new segment of the story ("so that morning"). In line 3, the PST displays three different ways that this new activity is "going a step too far." Sibil protests and tries to continue her story ("he said"). Now the PST announces that he wants to "structure" the talk, that is, interrupt the story and start a new activity, using his role as a discussion leader. However, the PST does not succeed as Sibil shows that she does understand what the PST wants to do (stop her narrating) but resists by characterizing her activity as "important." She does this by referring to something that will still happen in the story and is as yet unknown by the audience. She uses the fact that she knows the whole story and the group only knows what she has told so far, which makes her, and not the PST, the authority on whether all the significant details of the case have been shared. In addition to knowing what happens in the story, she also claims authority on the best way to structure her narrative. Because she is then allowed to go on uninterrupted by the PST, this negotiation between resident and tutor has resulted in consensus on what is more authoritative for the matter at hand: epistemic authority 28 trumps procedural authority.
Finally, the PST manages to take over control of the interaction by making the transition from narrative to narrator explicit (line 4): SIB that is the bottom // line PST // butyes SIB but // he keeps pushing it to the next level. PST // but what do you think do you think () yeah but just to you now like // SIB //to me to //me PST //where is / your, what would you like to discuss with us// SIB / that I SIB //that I / ehm kind of with these people, with people who are so:: demanding, let me call it that / who keep // PST /yes /hm PST //demanding manipulating SIB exactly that with them at one point I just say like no I am not going to do it, and what they "do next is that they walk away eh angrily Although it takes several attempts, the PST is finally the one to move the interaction to the next activity and identifies what that activity is about and what the key question is for the narrator and the group.
This shifts the focus of the discussion from the case presentation to the case presenter, and why she took the floor to narrate this case. This shift involves negotiation of epistemic positions: which person has the sort of knowledge that is required about the topic at hand, and which is the one lacking knowledge. In the first extract, there is agreement that Sibil is in a knowledgeable position with regard to how much time has passed between the first and second patient visit. In the second and third extract, she manages to continue her narration despite the tutor asserting first his expert authority (Extract 2) and then his procedural authority (Extract 3) to complete the story. She does this by using her privileged epistemic status as a narrator, thereby reflexively constructing her tutor's role as that of a-less knowledgeable-listener. The tutor finally manages to bring the storytelling activity to a close by constructing the presenter as being in a less knowledgeable position by shifting her epistemic status from one who is divulging information to an audience to someone who is doing so only to learn from the group.

Case B: From case presentation to case discussion
The previous analysis shows that there can be disagreement about what the next activity should be: presenting the case, or focusing on the reason for sharing. This shift from the story to the storyteller is not always accomplished by the tutor. Often, the case presenters make this transition themselves. In the following extracts, however, we see that even then it is not self-evident which activity the group should focus on next.
Lars is telling the group about a phone call with his supervisor: Extract 5: Lines 4-5 appendix LAR ((quoting supervisor)) you should be aware that you have to be really sick, not to come into work.
Although he does not explicitly state it, in the delivery of his report of the phone call Lars shows that he was, and still is, taken aback by what he takes to be an accusation. He stammers, leaves long silences, and often corrects himself: Another contrast with the previous case is that, whereas there the tutor attempted to cut the story short, in this one the presenter is the one who marks, by his silence, that the story is complete. 10  In line 5, Lars makes another closing ("so yeah"). This time, a fellow resident starts a new turn. But it is a question directed at him, which requests more information, thus putting him in a KC position: Does he know from an external source why his supervisor responded in that way? In Case A, we saw an example of clarifying questions during the story. Odette's (ODE) question, in which she requests specific information before responding to the presenter's request, is something we see more often in the data set. Her question shows an understanding of why Lars is sharing the case, responding to his emphatic "I don't know what to do" (Extract 7, line 4). Lars does not know what to do. Odette wonders whether Lars knows why his supervisor reacted that way. Odette's question is a response in the sense that she treats knowing why as a requirement for knowing what to do. In this sense, both Lars and Odette are negotiating epistemics: What do you need to know? When Lars continues speaking, one of the tutors intervenes for the first time: Extract 9: Lines 39-52 appendix LAR he just kept on ((stammers)) I have I ehh I answered his questions//and I PST //you were overwhelmed actually LAR yes I did not expect it at all, and then ehm only later after the call I thought like well, he is just trying to make me feel guilty, but it doesn't work. I think if I had done something #like he's right. then it is possible to make me feel guilty. but now it did not work at all. ((group laughs)) and I thought #like I I was just sick.
The tutor's remark may seem like just a summary of what Lars has already said, but this type of conversational object has been widely studied in CA and shown to perform more actions than just summarizing what has been said. It is used in, for instance, psychotherapy to "offer the therapist's version of the client's description and focus on subjective experiences." 40 It is a statement in which Person A makes an assertion about Person B's domain of experience, and formulates something that the other "implied but did not articulate." 41,42 This type of statement has two effects: It warrants confirmation (or disconfirmation), which Lars provides in the next turn (line 3). Second, a statement about another's domain of experience is almost always taken as a request for more information. Whether intentionally or not, in the EoE context, we see that tutors' use of this type of formulation almost always initiates a topic shift-in this case, Lars speaking about being "overwhelmed." Note that, here, this tutor is assuming a different, more therapeutic role than that assumed by the tutor in Case A, who took on a procedural role.
In In combination with the times that Lars has mentioned earlier, there is now a distinction in the story between 1. the phone call yesterday evening; 2. the moment right after the phone call; 3. this morning, when Lars is sharing the experience, and; 4. next week, when he will next speak to his supervisor. At the same time, there is a lot going on with regard to reflection: 1. Lars constructs himself as incapable of reflecting-inaction due to being overwhelmed, but reports on reflection-on-action after the phone call; 2. a hesitant delivery, full of self-repair and pauses, might indicate that Lars is reflecting while telling the story, looking for the right words, and so forth; and 3. a reason for needing to reflect, a next action (should I speak about this to my supervisor?) So Lars tells the story, and reports on his reflective process. Lars makes another closing "that's where we are now," followed by a silence. This time, another resident responds: Extract 11: Lines 59-65 appendix ILO but I would bring that up. say-say that you were overwhelmed, and taken a back by the conversation, LAR hm ILO and that it kept you worrying, This is the first outright response to Lars's reason for sharing. It is formulated as advice from a peer ("I would"), combining both the presenter's question (whether to "bring it up") and the tutor's formulation ("overwhelmed").
Lars responds to Ilone's (ILO) advice by adding new details about the situation with his supervisor, treating her advice as requiring more sharing of what happened. He switches to storytelling mode again ("and then"). While Lars is still doing this, the PST interrupts once again: Extract 12: Lines 77-85 appendix LAR and then it is half past five and then I have time now for (1.0) an eh (1.0) supervision meeting, PST but how did it how did that "actually affect you?
because it overwhelmed you at the #moment of the conversation, and "afterwards? LAR yes PST it started to kind of settle, what ((stammers in a questioning way)).
The rising intonation at the end of line 2 indicates that Lars is not yet finished. "But" and "actually" mark that she is asking about something Lars has not yet shared. Notice how the PST distinguishes what is already known (line 3/4) from what is unknown. One topic is closed (being overwhelmed), and the appropriate activity is defined not as advice giving, or storytelling, but as exploring how Lars felt "afterwards." Despite this, Lars continues to speak about his being overwhelmed, giving an account for not having the wherewithal to respond to his supervisor with a question: PST I get that but, LAR but PST how how how how did it when you had put the phone down,((imitates hanging up with her hands)) //because then ((starts to laugh)) LAR //yes I was. yes I was I was really very angry. GPT yes LAR I was actually very angry.
She successfully closes the "during the call" topic by specifying what Lars did-put the phone down-and making a parallel gesture. Of interest, when Lars says that he was "really very angry," the GPT's response is one of approval-of Lars having given the "right" answer.
With this intervention by the PST, the case presentation topic has been set, as other residents make contributions to this topic: Extract 15: Lines 102-119 appendix LAR eh look when I call in sick every month or something ok. but this was eh-n-n day two of ehh of this entire year. () angry YVO but apart from the fact that it is day two thatthat is not important of course. I think. / I think that the the fact that "he doubted your story, that that is a a a sore spot, and and you think I did not deserve this, right I / always work and I work hard and I am always honest, and / I think that that is the (1.0) sore spot. LAR / hm /yes / yes GPT / yes LAR yes exactly In lines 5-6, Yvonne (YVO) claims having access to what Lars really thinks, and why he is really angry (KC). 41 In doing so, she mimics the "therapeutic" role of the tutor took on earlier (extract 9, line 2). 42 Both Lars and the GPT seem to go along with this continuation of the topic in lines 8-10. And so the appropriate activity after the case presentation has been redefined from asking Lars questions, to giving advice, to discussing Lars's feelings after the phone call. Only after this discussion has been completed do the tutors suggest that Lars roleplay the conversation with his supervisor with one of the group members.

Discussion
We have focused on transitions and tutors' roles. Although the observed interactional practices vary from case to case, the basic interactional issues identified are ambiguity regarding the next appropriate activity, negotiating epistemics, and tutors' roles. We now elaborate on these three aspects.

Finding 1: The ambiguity of transitions
The interactional environment of transitions from case presentation to discussion turns out to be one of undefined rules and opens the floor for anything that might come up. We found many situations in which noticeable silence marks a context in which any participant can start a next activity, or, if no one does, the previous speaker can either make a closing or continue with the same activity. There seems to be ambiguity about the next appropriate activity; for example, is it advice giving or interrogating the case presenter? Is it answering questions about the story or is more case presentation needed?
Transitions require the closure of one activity and the opening of the next one. Our findings about transitions and off-topic sequences seem to echo Stokoe's findings on tutorial discourses 25 in that they prove pivotal for the way participants relate to one another and to the task at hand.
In this respect, it could be valuable to incorporate studies that have looked at sharing experiences in contexts other than education. Work on sharing experiences in therapy, 43 self-help groups, [44][45][46] and patient support groups 47 could be relevant. The extensive body of CA research on ordinary talk can also be used as a resource to pinpoint how EoE differs from, for instance, informally sharing experiences.

Finding 2: Collaboration and negotiation
The cases presented in this article are illustrative of how topic selection is truly a collaborative accomplishment. It takes a lot to run a reflection group-a lot is going on, different agendas and goals. The tutor has an important role in negotiating transitions. Managing transitions for the tutor means negotiating epistemics and shifting between different roles. As Harris et al. noted in their study of reflection by professionals, the tutors drive the interaction by framing and initiating topic sequences but are dependent on the residents' collaboration. 36 In Case A, we see that this collaboration is not obvious: It takes three attempts and 3 minutes before the tutor and the presenter align on the next topic. In Case B, it also takes a few attempts for the tutor to initiate a topic change, but this is not so much due to a lack of collaboration as to a breakdown in intersubjectivity. 48 In Case A, Sibil understands what the tutor wants yet does not comply with his plans, whereas in Case B Lars does not seem to understand what the tutor wants to achieve.
A crucial and recurring issue in transitions seems to be epistemics: what each speaker has come to know, wants to know, or already knows about something. Simply said, participants who present a case are in a powerful position and can control the topic, take the floor, and interrupt. However, in general, the tutors have a role in opening and closing activities and have institutional power. They can also extend an activity that others want to close, or close an activity that others want to continue. This power is sometimes trumped by a case presenter, who "owns" the case. "Outside of very specialized contexts such as psychoanalysis, the thoughts, experiences, hopes, and expectations of individuals are treated as theirs to know and describe." 49(p6) In reflection, epistemics also involves constructing the other (Lars, who by "I don't know what to do" constructs himself as being in K-), as being in some kind of KC state (ODE: Do you have other sources for knowing? PST: How did you feel after the meeting?). In Case A, it involves managing the transition from a resident who is in KC mode (I know what happened, you don't) to Kmode (What do you want to know from the group?).

Finding 3: The role of the tutor
The transition from case presentation to case discussion means shifting the focus of the discussion to the case presenter. The tutor can have a role in closing the case presenter's storytelling activity and starting the activity of exploring the reasons why he or she is presenting the case to the group. When the case presentation is closed, the tutor has a role in monitoring the discussion and negotiating the topic flow to a "learnable" position. Both of these are complex interactional accomplishments, especially when transitions are marked by competing agendas (e.g., the case presenter continuing the story, the tutor focusing on why the story is being told).
Tutors switch between different roles, such as audience member, discussion leader, expert, and facilitator of the reflective process. These roles have different interactional consequences. The first role that we saw is that of the teacher and GP training representative. The tutor exemplifies this role by asking questions that are intended to teach the group something, but also by asking questions or monitoring the assessment of the trainees and their learning environment. Second, the tutor is an expert on the content that is being discussed and has a lot of experience with EoE sessions. Third, the tutor is a facilitator of the discussion and the reflective process, ensuring that trainees learn by interacting with one another and asking one another questions. One of the ways tutors do this is by not reacting immediately, leaving space for participants to take the lead. In addition to these three roles, the tutor is also an active participant in the EoE sessions, listening to the case and responding much like any other participant (e.g., sounds of outrage, shock, disappointment, empathy).
The tutor is part of the group interaction and, in the moments of transition, the tutor's interactional behavior is relevant as it implicitly shows the group what is expected in the sessions. In other words, in the transition phases, the hidden curriculum 50 of the experience-sharing sessions comes into view. "Lessons learned but not openly intended" 51 do not show up in research approaches that focus on what participants say they do in teaching practice, but they become observable when we examine what they actually do.
CA has confirmed again and again how "much of the interactional organization that Conversation Analysts study, while robust and perfectly amenable to formal description, exists below the level of ordinary awareness of the ordinary person." 13(p79) Being more aware of this interactional organization could make tutors more conscious of what they are already doing and what they can do. In guiding an EoE session, tutors need to be aware of their own role as teacher, expert, facilitator, and participant-and when to take on which role. However, it is not just up to the tutor to choose a role; the group can also steer the interaction a certain way, restricting or facilitating certain behaviors by the tutor.

Strengths, limitations, and further research
We have conducted a thorough microanalysis and unveiled characteristics of EoE that were previously unknown. The richness of the data might inspire the reader and encourage deeper thought about interaction in education. We see this study as an empirical starting point: A description of how sense-making practices in learning settings are accomplished in situ can be the basis for examining what makes a particular practice or intervention effective. 30 As there is so little research on this particular genre of interaction, there are many possible future projects. Future research could focus on different tutoring styles, collaboration between the two tutors, or ways in which participants structure and present their case. Another line of research would be to ask participants and tutors to rate very useful and educational sessions or cases, and a few that are less so. We could then identify features that make an EoE session valuable. We feel that, despite all these possible research avenues, microanalysis such as conducted in this study is a suitable way to generate hypotheses and change current ways of thinking about how reflection or learning is achieved in group sessions.
A discursive approach to group reflection There are many theoretical models of reflection, yet "beyond perceptions and abstractions … systematic inquiry into how reflection is pursued and produced in the details of actual … interaction, however, remains rare." 4(p99) CA is interesting for educators because it focuses on the practical business of education. In the case of collaborative small-group learning, this involves group dynamics. In our analysis, we show that precisely these "messy" moments in the interaction, where the question of what activity is to be the focus of the discussion is subject to negotiation, are rich in implicit information about the way in which group participants relate to one another.
We took into account the reality of the classroom, in which reflective practice is an in situ and interactive accomplishment of teacher and student, and which is supposed to mediate these outcomes. 3,36,37 Many studies on reflection are outcome based and do not take the "messy" interactional reality into account. Insight into the tutor's role in EoE meetings (and similar educational ventures) can put flesh on the bones of the myriad of abstract reflective models that teachers are expected to translate into practice.