CA
Conversation Analysis (CA) is the study of interaction using a fine grained analysis of spoken conversations. The researcher moves between detailed examination of individual cases (specific segments of transcriptions) and a general view of a set of related cases. By collecting multiple cases the researcher can get context independence. Audio recordings are important in CA because there is a great deal of attention to not just to the words that are spoken, but also to their timing and intonation.
CA is about actions and practices. Practices are distinctive, turn-based and notable for the consequences that they have. The way that prior turns talk is understood is key to CA. This is known as next-turn-proof-procedure. Establishing patterns of behavior is important, but deviant cases, where a pattern is broken are extremely important because they provide insight into the normative structures that the participants are engaged with.
Generally speaking CA draws on regularities and co-occurrences in talk. Some examples of these regularities outlined by Paltridge (2012) are:
- opening conversations: how conversations are initiated or started
- closing conversations: how conversations end
- turn taking: the ways in which participants signal the end of their turn
- adjacency pairs: regularities in two successive speakers that form expected sequences of behavior
- stage of conversation: examining the way different adjacency pairs can behave at different points in the conversation
- preference organization: examining the ways in which adjacency pairs can be examined using the preferred and dispreferred response
- feedback: how speakers show that they are listening and understanding with their words
- repair: how speakers correct themselves or others in their speech
- discourse markers: items in spoken discourse which act as signposts of discourse coherence: oh, now y’know
The focus on the spoken interaction, and the transcript is not to make a philosophical claim about the world–it is simply a method for analysis. It is an argument for evidence based analysis.
Often CA starts as undirected research, or immersion in data, where the analyst notices interesting outcomes in the text, and tries to identify what conversation practices are involved. Or it may be that the analyst notices features of the talking, that invite them to take a closer look at what outcomes they might be associated with. This initial noticing can lead to identifying the same pattern in other text. Data collection can take a while–years in some cases. An analytical unit (turn, sequence, etc) can be part of multiple collections of related collections.
In another reading this week Sidnell & Stivers (2012) uses an analysis of the use of Oh to show how it used not to indiciated surprise, but a transition from unknowing to knowing. While I understand that Sidnell is introducing CA and how it is performed, I seemed to miss the connection between the detailed understanding of Oh and a research question. How does mapping the terrain (to use Sidnell’s metaphor) of the use of Oh in this particular discourse help answer a research question? Or is it meant to simply be descriptive of a particular phenomenon? What is the community of practice that is being understood here?
Schegloff, Jefferson, & Sacks (1977) ties the idea of self and other from sociology to turn-taking systems, or conversation. They are making the point that self-correction is a preferred form of correction than other-correction, and want to understand how that preference is deployed in conversation. I almost wrote why it is used, but I’m not sure at the outset if that is in fact their goal. Why seems like more of a psychological question.
I kind of like the way they start out by stating their base assumptions about self-correction in plain language. It adds clarity. They are intentionally scoping the idea of repair to include the already established notion of correction. So it will include situations where there isn’t necessarily an error that is being replaced. They use examples to illustrate this, which is very helpful.
It struck me while doing the readings this week that the transcription notation in CA offers greater detail but also has the effect of making the cases really stand out from the regular text of the article.
Repair initiation and repair outcome are useful tools for looking at the repair sequence. They basically show that the preference for self-correction can be explained by looking at how repair initiation operates to give the self an opportunity to correct rather than doing the correction themselves.
The number of examples in Schegloff et al. (1977), and the categories that they were taken to represent was a bit mind numbing and difficult to keep in my head at one time. I felt like a diagram would have been useful. Perhaps this complexity is because he was carving out a new area of research, and the options for publishing diagrams at the time may have been limited?
Wilkinson & Weatherall (2011) take a deep dive into a particular type of repair that Schegloff et al. (1977) cataloged: insertion repair. They look at 500 examples of insertion repair, to see how it is used to intensify or specify talk. It’s interesting to note that the study drew on multiple datasets of spoken English from England, the United States and New Zealand.
Insertion repair can be used to differentiate between possible referents, which is either explicitly stated or inferred. If there isn’t another referent involved, then the insertion is being used to clarify the type of referent it is. Their choices of examples are easier for me to understand, even with all the detail they offer. Also I like how they show how observation of insertion is used, for example in this segment:
01 DR: Cause I mean they even: (0.8) they’ve got
02
street gangs rou:nd in Cuba Ma:ll no:w.
03 IV: Ye:s.
04 DR: for [ cry] in’ out lou[:d. ]
05 IV: [(◦Mm◦)] [Yea]:h.
06 DV: It’s- (0.5) (pretty) ◦ba:[d.◦]
07 IV: [O:h] it’s getting ba:d.
08 It’s really- They [’ve gotta do something.]
09 DR: [ I mean you can’t even ] wa:lk
10 (0.2) thr(h)ough Cuba Ma:ll now [huh]=
11 IV: [No:]
12 DR: =all those people ma:n.
13 (.)
14 IV: Yea:h.
15 DR: They steal your bloody sh: Doc Mar:tens shoe:s
16 an:[:::wh::]whatever you’ve got o:n it’s-
17 IV: [ Ri:ght. ]
18 IV: Mmm.
19 (0.5)
20 DR: pretty ba:d but [- ]
21 IV: [Ri]:ght.
22 (0.8)
23 DR: They won’t be stea(h)ling the co(h)rolla
24 anyw(hh)ay s(h)o: huh [huh huh] huh huh
25 IV: [ Yea:h. ]
The authors connect the dots between the insertion of a type of shoes (Doc Martens) and the subsequent use of the Toyota Corolla to draw attention the way in which brands are used to qualify the type of thieving that’s going on. Examples like this get me to see how CA could actually be useful when trying to analyze, understand and contextualize a particular type of speech, scenario or environment. It takes it out of the realm of exploring the universe of language, which is interesting, but also feels like a bit of a navel-gaze. The analysis needs to be put to use in some way: having a paper that just finds some new feature of language use doesn’t in itself seem that interesting to me. Although I grant that it is interesting to other people.
For my own research I think CA could be a useful tool. I recently conducted 30 interviews with archivists of web content to see how they decide ascribe value to content and enact appraisal in web archives. I have audio recordings and transcripts of these interviews, and have spent some time coding the transcripts and my own field notes from the interviews. I can see that CA could be a useful way to identify patterns in the conversations, that could possibly help me uncover some buried meanings, or insights into what is going on. For example I could examine the way that collection development policies were talked about.
Even though these interviews were unstructured ethnographic interviews they aren’t really natural conversations about appraisal. I worry that my participation in all of them as the interviewer would really distort the analysis. But perhaps that could be accounted for. Also, I have 30 hours of transcriptions, which is a lot of material to sift through. How can I narrow down to individual cases in that content? I have definitely been immersed in the data to do coding, but I haven’t examined them from a conversational perspective. I guess this is something that makes it appealing, to triangulate on the interviews using a different method, to see what I can learn from them.