Though the idea that conversations are orderly is undoubtedly old, the systematic study of conversational interactions is somewhat new. Conversation Analysis (CA), which is just over thirty years old, describes the competencies and procedures involved in the production of any type of social interaction. In comparison to many sociological approaches, CA is an exact and empirical enterprise, avoiding immature theoretical speculations and informed by a set of theoretical propositions. In this chapter, I will discuss the basic principles of CA and its application, in particular, to the study of institutional interactions and practices (for overviews of CA, see Hutchby and Wooffitt 1998; ten Have 1999; Silverman 1998). In discussing the working principles of the discipline, I will also consider their potential relevance for the study of action in institutional settings, and will also address some disputes concerning the applicability of CA. I will then compare CA with other methodologies. At the end of this chapter, I will briefly return to the issues of how CA might be applied for practical purposes.
In institutional contexts, CA discerns the ways in which talk is specialized and reduced to accomplish the tasks at hand. CA studies do not generally rely on ethnographic knowledge, but the analysis of some institutional settings may require contextual knowledge in order to make sense of realms distinct from everyday life. CA uses inductive logic so its reliability is based on analytic induction (to be discussed more in Chapters 3), but in some specific cases statistical evidence plays a role.
The distinctiveness of CA as a social scientific approach derives from its object of analysis. CA studies conversational turns and interactional moves in their sequences.
1.1 Basic Ideas
The basic idea of CA is so simple that it is difficult to grasp: CA studies what an utterance does in relation to the preceding one(s) and what implications an utterance poses for the next one(s). As Hutchby and Wooffitt (1998, 15) put it, the nextâturn proof procedure is the most basic tool in CA (see Sacks et al. 1974). That is, the next turn provides evidence of the partyâs orientation to the prior turn, there and then. This methodic procedure is CAâs gateway to the participantsâ own understandings as they are revealed during actual interaction, thereby providing material for analytic explication. For example, consider this brief exchange between E and M (transcription simplified).
(1) [NB:VII:2] (Heritage 1984a, 236; ten Have 1999, 4)
E: eâthat Pa:t isnâ she a do:[:ll?]
M: [iYe]h isnât she pretty,
In using the nextâturn proof procedure, we should be able to say something about Eâs turn with the help of Mâs turn. Let us begin with the obvious. Mâs turn is designed as an answer, but a particular kind of an answer, an assessment. CA then proceeds with a comparative approach through which the specificity of the data instance is explicated. The comparison can be imaginary in the first place. If we wished to work empirically, we would collect parallel cases to find regularities through which similar kinds of actions are accomplished. Here, however, let us be content with an imaginary comparison, a kind of game where the analyst tries to locate an observation within his/her knowledge/imagination, thus sketching out the meaning of an actual course of interaction through comparison with imaginary cases. A proper demonstration would be based on an empirical collection of parallel/similar cases with whose help regularities would be spelled out. Let us proceed. Notice how M continues her answer after the response token âyeahâ and in so doing treats her âyeahâ as an insufficient response to Eâs action. At this point, you should be able to see a hermeneutical circle at work. The nextâturn proof procedure means that a reflexive relationship exists between adjacent turns: the next turn is used as an analytic resource for making sense of the prior turn, which, for its part, has provided the sequential implications that have made the next turn relevant.
Mâs turn suggests that E has invited M to produce a second assessment. In other words, despite its grammatical form, an assessment that is delivered through a yesâno question format does not work like an ordinary question. That is, M does not treat Eâs utterance as a straightforward question but as an invitation to assess the person E herself has described as a doll. Moreover, Mâs assessment is a specific kind of assessment compared to Eâs prior assessment: it is weaker and narrower, downgraded, which suggests that M does not agree that strongly with E. Now, just as we are about to close our analysis (at least for the moment), we are on the verge of sociological/sociopsychological findings. The situation is rather juicy: E and M are talking about the third party, Pat, and a particular type of relationship is emerging between E and M. E has invited M to participate in a joint appreciation of Pat, but M has declined the invitation with a mild response, and a gulf between their perspectives has been opened. E has provided an assessment, whose upgraded quality M has made plain through her mitigated second.
But, a reader may protest, is this all pure speculation? Can we say anything about the validity of this reading? Maybe Arminen got it all wrong? Is there any way to test and check the accuracy of the analysis? Actually, CA allows its findings to be tested through the very same nextâturn proof procedure (Heritage 1984a, 256â257). We can examine the turn following Mâs turn to see whether our explication of the interaction fits with the partiesâ sense of the ongoing interaction as they reveal it turnâbyâturn. We might even imagine the set of alternatives that E would use to counter Mâs downgraded assessment. In this way, even if we are not able to make infallible predictions of the next turns, we can give an accountable description of the course of the conversation and of its potential next moves. Further, proper empirical research would be based on a collection of cases, whose analysis should amount to invariable regularity.1 Here, we have the chance to check our skills simply by imagining how the exchange will continue, and then looking at the extended sequence of the exchange (see Heritage 1984a, 236; ten Have 1999,4).
(1)((continuation)) [NB:VII:2] (Heritage 1984a, 236; ten Have 1999, 4)
3 (.)
4 E: Oh: sheâs a beautiful girl.=
5 M: =Yeh I think sheâs a pretty gir[l.
6 E: [Enâ that Reinamân::
Any time you feel that you have a better account of the sequence than one that has been given, please feel free to develop it further, and check it against sequence of data. The adequacy of this kind of analysis is not primarily theoryâbound. The analysis is not supposed to be measured against any theoretical account of interaction, but against the reality of recorded interactions and their transcriptions.
Hutchby and Wooffitt (1998, 38â39) make a useful analytic distinction between âsequential orderâ and what they call âinferential orderâ, though sequential and inferential orders do presuppose one another. That is, the partiesâ inferential work â the kinds of implications and inferences participants draw about each otherâs talk and conduct to make sense and to hold each other morally accountable â allows them to build sequences of action upon which this inferential work rests. Sequential order means the âdescribable ways in which turns are linked together into definite sequencesâ (ibid.), and its analysis provides the backbone of CA. However, this sequential order is tied to the inferential order, hence the sequential analysis touches also upon the inferential order. In the final instance, the inferential order is the basis for everyday semiotics. This becomes plain in everyday life, but can also be seen in literature. For example, in Shakespeareâs Twelfth Night, Viola, disguised in male clothes, interprets Oliviaâs way of speaking as a sign of her psychological state:
She made good view of me; indeed, so much That, as methought, her eyes had lost her tongue,
For she did speak in starts distractedly.
She loves me, sure ⊠(2.2.18â21)
The distinction between different âordersâ opens up the multidimensionality of the CA research object. Essentially, CA is about the organization of interaction, about the syntactic, semantic, and prosodic qualities through which turns are designed, and about the pragmatic connections between turns.2 Furthermore, as Hutchby and Wooffitt (1998, 39) stress, these concerns interplay with normative and inferential properties of talk through which participants orient to the sense and implications of their interaction. The multilayered orderliness of talk makes it a âdeepâ object, so that even a seemingly innocent or insignificant property of talk may become relevant when looked at from another angle. CAâs programmatic stance suggests that we should not a priori assume the irrelevance of any detail of talk; instead, we should try to find order at all points, as Sacks said (1992a, 484; Hutchby and Wooffitt 1998, 17â22). This methodological canon creates the possibility of unlimited new findings, but also makes the research process a neverâending quest. We may think of CA as the reverse engineering of an immense complex of intersubjectivity. Deciphering this enigmatic structure requires that the analyst be highlyâskilled in observing, detailing, describing and systematicizing this fractalâlike multitude.
On an analytical level it may be helpful to distinguish between different styles of doing CA. The analysis may focus on the sequential order, paying only minimal attention to the inferential properties of talk. For instance, we could have concentrated on the properties of Eâs and Mâs turnâdesign and on the relationship between turns, passing over the potential social implications of their exchange; or, we could have analyzed the properties of Oliviaâs turnâbeginnings, such as breaths and other aspirations including laughter and laugh tokens, recognizable contextedâsilences, coughs, âyâknowsâ, âuhâ in all its varieties, cutâoffs, reâbeginnings, reâdirections, etc. (Schegloff 1996a, 103). CA demands a disciplined approach, with the analyst not jumping to sociological or psychological conclusions, falling into immature theoretical speculations, or relying on everyday assumptions. Paul ten Have (1999, 107) goes so far as to make a distinction between âpureâ and âappliedâ CA, arguing the former should concentrate âon talk âitselfâ, rather than its âcontextââ. To my mind, however, this strict division and the whole notion of âpureâ CA is misleading and inadvisable. Moreover, separating talk from its context goes against all the basic ideas of CA, according to which the contextârenewing properties of talk amount to the endogenous construction of context, as parties orient to the âcontextâ through the management of talkâinâinteraction as an observable part of doing social actions in the context (to be discussed more in the next chapter). A more sensible way to address the issue of the applicability of CA is to stress that CA allows, and even necessitates, the selection of the focus of analysis, which may be more closely connected to the sequential or inferential properties of talk and action.
As a whole, CA is a technology to access the orientations of the members of a culture, and to avoid implausible constructive theorizing. CA is a program of âreverse engineeringâ which analyzes interactional practices in order to articulate and respecify the generic building blocks of social interaction. The results obtained illuminate ways in which social and institutional realities are occasioned, maintained, and managed with the help of the organization of talkâinâinteraction (Heritage 1984a, 233â292; Drew and Heritage 1992b; Pomerantz and Fehr 1997).3 The findings of CA are both âuniquely adequateâ in that they provide a contextâsensitive understanding of a specific instance of interaction (Psathas 1995, 45â53), and are âgenerically in...