Clarity in applied and interdisciplinary conversation analysis

Over the years, I have experienced the challenges of interdisciplinary and applied conversation analysis. In teaching a course on talk as social organization, among the most common issues that arise are whether one is ‘allowed’ to apply conversation analysis (CA) and whether one can combine CA with other methods. Likewise, when I was reporting research on women in meetings (Ford, 2009), conference audiences, scholars from inside and outside of CA, regularly questioned whether I could legitimately use CA methods to study participation by members of an a priori category: woman. My response is to ask: What better method than CA for understanding turn taking in meetings or in any study of interaction? Does one avoid the method just because one is not using it to ground social categories evidenced in participants’ practices?

As a form of ethnomethodology, CA uncovers systematic mechanisms through which social interaction is accomplished. With its aim of arriving at general methods for doing social structuring on a turn-by-turn basis, CA is unique in striving toward accounts of practices based upon what is visible, hearable, displayed, and responded to, by actors in real-time interaction. This standard of evidence is grounded in the fact that interactants themselves, those whose practices we aim to understand, rely on such conduct in making sense of one another, while simultaneously displaying the sense they are making. CA practitioners strive to be answerable to the question of whether interactants orient to the practices and resources we propose. In reporting our findings, we offer our best approximations. As ethical limitations permit, we make our data available for others to inspect, remaining open to better analyses, ones that bring us closer to general principles that hold on a case-by-case basis, as they must for the participants.

With regard to applying CA, combining it with other methods, or adapting it for non-CA research agendas, to me CA is simply too powerful an approach to be abandoned because one is not using it in service of its own primary ends. What I do find essential is that we maintain a standard of clarity in formulating relationships among findings based upon CA methods, how they articulate with other approaches, and what questions we are answering in combining methods. It is on this point that I will engage with Waring, Creider, Tarpey, and Black’s article. The key contributions of the piece are to continue a dialogue on CA and context, and to propose benefits of drawing upon participants’ video-stimulated reports, bringing such reports to bear on the results of CA analyses. The essay is successful in both regards, but it also involves what I have always found to be problematic formulations of how participants’ reports of their experiences and intentions are related to or enriching of CA-based analyses. Waring et al. provide an opportunity for dialogue with what I take to be contradictory positions and formulations that we tend to articulate in doing CA in conjunction with other methods.

In the conclusion to their article, the authors take a clear position regarding the status of stimulated interview data in relation to CA analysis:

CA uncovers what is done as displayed and oriented to by the participants themselves in the moment-by-moment interaction, and talk-extrinsic data (in the form of participant commentaries) reveal what is done as reported by the participants. In particular, these reports may represent a version of events that the participants would like to have their recipients see. (Waring et al., p. 489)

The authors warn that convergence between data from interviews and from CA ‘does not strengthen our finding of either, just as any disconfirmation would not weaken the findings of either. In other words, the findings from different data sources do not validate or invalidate one another’ (p. 489).

And they affirm that ‘CA and talk-extrinsic data do not answer exactly the same questions’ (p. 489). What I find problematic, in the article taken as a whole, is the mismatch between this position and the ways that the authors have formulated relationships between CA and interview findings in the case studies that make up the body of the article. The case studies present good CA findings and then consider whether video-stimulated interviews support new perspectives on what is revealed by CA alone. In these illustrative cases, the authors use language that suggests a directionality in relating interview findings to CA analyses. Such language seems to be at odds with the authors’ own clear position on the topic (quoted above). I will touch on the formulations that seem to go counter to the authors’ general position that interview data do not ‘strengthen’ or ‘weaken’ the findings of CA (Waring et al., p. 489).

For their first illustration, a sequence from a writing tutorial, the authors formulate a relationship between interview data and CA analysis as involving ‘specification’:

. . . the participant’s concern is broadly inferable but not specifically displayed in the CA data and only comes to light in a subsequent interview. In this case, then, the talk-extrinsic data serve to specify what is broadly inferable from the initial CA analysis. (p. 482)

It is not clear what ‘specify’ means here. We are offered a participant’s reflection on and report of her priorities in her tutorial work, and we gain access to her recollected concerns regarding a particular student she had several years before. This seems useful for arriving at a take-away message for tutors, tutor-trainers and writing centers. If CA is to be articulated with those reflections, then the video-stimulated interview data may help the researchers get at what is important to tutors in writing centers and how such concerns may be enacted in real moments of videotaped tutorials. This seems to support the value of combining interviews with CA-based findings. But if the claim is that the specifics from the post hoc interviews can now enrich what we know about how the interaction was structured by the participants as the original interaction unfolded, this seems counter to the authors’ position on relating interview data to CA. As they affirm in the conclusion, a participant’s report of concern is not the same as the enacted, visible, hearable concerns that she and her interlocutor displayed and responded to in real time, and it is the latter to which CA gives us access.

In the evaluation of how mixed methods combine in the second case, an educational encounter between an adult and a child, Waring et al. find that CA reveals misalignment as the participants jostle for the terms of a learning game. The post hoc interview with the adult participant provides insight into her experience of the original interaction and what purposes she may have had for acting as she did. Very reasonably, the authors highlight convergence in results of the two methods: the participant’s report ‘matches what was revealed in the initial CA analysis’ (p. 484). What is less convincing is their formulation of the convergence as ‘what is displayed is also what is intended’. A different research method would be required to operationalize ‘intention’, to identify it in self-reports and to identify its enactment in naturally occurring interaction. As it is, we do not know that participants have clear access to their intentions either in their moment-to-moment participation in interaction or their subsequent interviews, stimulated by access to recordings of their prior interactions. ¹ This is not to deny the participants’ experiences or their reflections on their intentions. Rather, it is to call for clarity about what is afforded us by different methods and data sources: What do we know and how do we know it? For the novice CA researcher encountering the language of Waring et al.’s summary of this case, my concern is about the potential misconceived ideas such formulations may support regarding what can, with methodological integrity, be concluded when self-reports match CA findings. We cannot straightforwardly discover the intentions behind the practices engaged in in the recorded interaction.

Also in their second case, Waring et al. use the notion of confirmation to formulate the connection between interview data and CA analysis: interview data, in their words, ‘confirm the initial CA analysis in terms of what was going on’. CA findings can only be ‘confirmed’ or, to use the paired scientific term, ‘falsified’ by other CA-derived findings. Indeed, one regular CA enterprise is the search for deviant cases that test the power of a generalization. When a claim is found untenable, a broader more nuanced account is called for (Schegloff, 1993). What is ‘confirmed’ in Waring et al.’s second case is not the CA analysis but something else, a claim or hypothesis not articulated in the article. The CA analysis and the interview findings for this child–adult interaction could be combined to address other sorts of questions but not CA questions. If we are after an understanding of individual and institutional values and goals (accessed through interviews) and how those fit with current interactional practices (accessed through CA), then we can usefully draw on convergence and divergence in findings from different methods (e.g. Kitzinger, 2011; Peräkylä and Vehviläinen, 2003). We might communicate in non-CA terms about ‘the agenda that motivates’ participants (Waring et al., p. 484). But by using the language of confirmation to articulate what the interview data does in relation to CA findings, the authors invite others to treat interview data as a confirmatory resource for CA, when in fact, as the authors themselves affirm, the methods answer different questions.

The language of confirmation is also deployed in the third case study, where interview data ‘serve again to . . . confirm the initial CA analysis with regard to what transpired’ (p. 486). For the third case, the authors also present interview data ‘serving as a corrective’ to inferences from CA (p. 486). I agree that the interview corrects ‘an inference grounded in the initial CA analysis’ (pp. 486), but the notion of inferring from CA requires reflection. Among the ways that we might explore inferences in relation to interactional data, at least three come to mind: 1) evidence of possible inferences displayed by participants in their emerging interaction; 2) inferences made by analysts based on CA; and 3) inferences made by and/or reported by participants after the interaction. What Waring et al.’s interview data provide is a view of what one participant, in retrospect and stimulated by the videorecord, comes to understand: that he may have misunderstood something in the moment of interaction. But this does not change what CA reveals as made available, demonstrated to be misunderstood or not, in the original interaction. Practices of repair were available but not used, and that is simply how the interaction ran off. In short, the CA findings themselves are not corrected. This does not mean nothing is to be learned by combining CA and interview data in this instance. What is learned is not that the CA missed a key aspect of how that social moment was co-constructed. A lesson that might be learned, for the participants and in retrospect, is something useful about how they, in this case close friends, achieve or fail to achieve understanding in a vernacular sense. That could be beneficial for research on interpersonal communication and the possible benefits of stimulated recall in relation to longtime friendships and the sedimented practices they involve. Friends and intimate partners in some societies regularly work to gain insight into how they enact their friendships, and CA has much to recommend itself for such enterprises.

In their final case, from classroom interaction, the authors astutely observe, ‘not displaying recognition is not the same as not recognizing’. They report that differential access to cultural knowledge among the analyst, the participants, and the participants whose stimulated recall they accessed led to reconsideration of who understood what and when. Accumulating layers of knowledge led to an ever richer perspective on the possible interpretations available to the multiple participants present in the classroom as a student offered the term ‘Californication’, as a complete turn and action. The authors’ consideration of multiple meanings and associations with the culturally rich term raises important issues for large group interactions. For example, in addition to the interview data and the analyst knowledge that the authors cite, how many other perspectives were possible among those in the classroom regarding what action this word and turn were doing? With regard to combining CA and other methods, for this case Waring et al. (p. 487) find that the talk-extrinsic sources ‘correct an initial CA analysis with regard to what is being done [. . .] or what might be inferred’. Again, CA provides an analysis of how a sequence of action actually ran off; and the other sources, here interviews and the analyst’s reflections on her own cultural knowledge, provide multiple views of what work a particular word, one turn and action in that sequence, might possibly be doing. CA attends to the way(s) the turn was responded to in the moment, and from that methodological perspective there are no grounds for correcting the analysis of what was made relevant through the participants’ practices. On the productive side, by combining CA and talk-extrinsic data, the authors’ work may suggest new directions for exploration focusing, for example, on how jokes are done and taken up. ² What makes something come off as being done as a laughable and how might we understand when and why such constructions are affiliated with by subsequent laughable in response and when they are essentially deleted in favor of other matters? For research on pedagogical interaction, we might propose that teachers’ priorities are staying with the teaching agenda rather than getting into playful cultural references. So playfully inserting such a turn might be a way of ‘doing being’ a student in this class, and not fully aligning with the play may be part of ‘doing being’ a teacher in that moment.

In sum, and as Waring et al. themselves assert, CA answers questions distinct from those answered by video stimulated interviews. We are not warranted in claiming that the participants’ post hoc comments are capable of verification or correction of CA findings per se. Waring et al. are not alone among CA practitioners who have used such terms as ‘verify’ and ‘confirm’ in formulating relationships between participant reports and CA analyses (e.g. Maynard, 2003; Pomerantz, 2005). ³ However, the language we use in conducting our research and reporting our findings is consequential. Language guides us in our analyses, and it models for others the norms and standards for formulating relationships between the results of these different methods in mixed method projects. My particular concern is for those new to CA and to its interdisciplinary applications. I want to encourage interdisciplinarity and application, but I want to support it, as much as possible, by reporting my mixed method research with clarity about what the different approaches afford and how they can be related to one another.

So what is to be made of participants’ self-reports of their interactional practices? And what of their perceptions and experiences as they view video records of their actions? In terms of CA’s goals, such moments are new opportunities to inspect the joint construction of social activity; they are interactions between an interviewer and an interviewee, and results can be studied as interactions in their own right to help us understand practices for doing interviews. For non-CA research agendas in which CA is used as one method, self-reports are sources for understanding participants’ concerns, ideologies, and the potential relationships between retrospective recall reports of motivations and the real-time structuring of social action, among other matters. For some sorts of research questions, gathering data on such reflections makes absolute sense. ⁴

As an applied and discourse-functional linguist and a conversation analyst with a long-standing commitment to interdisciplinary approaches to language, I share the motivation to do multi-method studies such as that reported in Waring et al. At the same time, I am convinced that it is essential we be clear in articulating the difference between doing CA for CA’s sake and drawing upon CA to answer non-CA questions. Both enterprises are valuable, and in no way am I suggesting that one should only engage CA if one is aiming to contribute directly to its core agenda. But the power of CA derives from what the method uniquely affords: CA focuses on how, in real time and for one another, humans jointly construct the local social orders that make up their daily lives. The success of projects bringing CA together with other methods depends entirely on the integrity with which CA’s core methods are used and reported. Without such integrity, there would be nothing coherent to combine with other forms of data gathering and analysis or to apply in addressing other research priorities.