(DISCLAIMER: This is a slightly more academic-toned post than many I’ve been writing so far. These might become more frequent for a while as my doctoral writing goes on. Still I hope to make it all accessible and interesting even to those who aren’t necessarily up to speed with the latest in humanities academia and academic approaches to language.)
For a while now, I’ve been searching for a good academic methodological ‘hook’ for looking at language on Jordanian radio. One semi-popular approach to such issues is Conversation Analysis, an empirical-analytical method developed by a gaggle of U.S. sociologists from the 1960s onward with the goal of studying the finer details of human conversation and see how mutual understandings emerge in actual social interaction – rather than being forced to read these off as ‘interpretations’ from interviews or textual descriptions of events. On the surface, it looks workable: you transcribe people’s conversations from recordings, with a reasonable degree of accuracy, and then look for patterns and principles in the details of how they exchange their turns and communicate. But delving into actual Conversation Analysis (best practice is to capitalize it; though I’ll be using CA henceforth) as an methodological-analytical approach, it becomes clear it wouldn’t quite do. It is, on the one hand, a bit too narrow in the kind of data that it admits as valid for analysis; and also, once you look at it closely, not at all the methodologically rigorous end-all it claims to be.
CA takes as its starting point the idea that talk, as a prime component of human social interaction, should be studied in its own right – as an activity with its own particular features and principles, and not merely a neutral medium of transmission for human values and beliefs. Conversation analysts proceed from transcripts of conversation (any kind of conversation: live, phone, broadcast – though always recorded so that a proper transcript can be made from it), and look at features such as when and how people take turns, when they are silent, when and how they correct themselves or each other (what CA calls “repair”), when and how they display (or not) that they have understood their interlocutor in conversation, and so on. The goal is to see, in short, how talk is structured as a social activity, and what this can tell us about how people deal with and come to understand each other over the course of this activity.
One strong proponent of using CA in studies of media has been Ian Hutchby, especially in his book Media Talk, where he looks at several broadcasting contexts – such as TV interviews, radio call-ins, and TV ‘audience participation’ programmes – and shows how these are organized according to principles that conversation analysts have found holding for “ordinary conversation” (Hutchby’s terms not mine). At the same time, conversation in media also has certain features not found in “ordinary” talk, and Hutchby lists a number of such features that CA can identify – which I’m not going to recount in detail here, but they include practices such as generalising reference in giving advice and constructing power asymmetries between hosts and broadcasters.
Picture 1: Gail Jefferson, one of the three founding figures of Conversation Analysis (the other two being Harvey Sacks and Emmanuel Schegloff).
For this to work, though, one needs to assume that (a) broadcast talk is in fact different from “ordinary conversation,” and (b) we know what the principles of “ordinary conversation” actually are. Both are assumptions made by whoever is analysing the talk (I’ll discuss a bit more below why this is a problem for Hutchby in particular). The second one, especially, links to one pervasive weakness of CA: its ethnocentricity. Most work on CA has been done on English, with the result being that we simply know much more – in brute, numeric terms; I’m sure every conversation analyst would agree that it’s possible, and desirable, to run CA on languages other than English – about the details of Conversation in English than in any other language. (Linguistic anthropologists have begun to redress the balance a bit, but the skewing is still considerable.) It also has somewhat eclectic transcription practices which any purebred linguist would probably gag over (eye-dialect! Putting a lengthening sign on an English vowel without indicating its IPA value! Using a colon as a sign for lengthening sounds in the first place, rather than the two tiny triangles that IPA necessitates!! Etc.). It’s not, then, an analytical approach which would be easy to generalise beyond English (even if there’s nothing in the actual principles of CA that would exclude this possibility).
There are more problems that have to do with the way CA approaches data. The focus is always on transcripts – and transcripts alone. For a conversation analyst to find an interesting feature to talk about, this feature has to be noted in transcripts, and observable in them across a number of cases. What’s going on in the minds of the people talking to each other is irrelevant – for CA’s purposes, it’s what they say, and how they say it, that matters; and that alone. All the “shared understanding” that CA goes on about is in fact only what is demonstrated in conversation; the “understanding” here isn’t a cognitive understanding, necessarily, but rather enough common ground for the conversation to continue in order to make sense. (Obviously, there needs to be some prior common cognitive understanding – at the very least, of the language that is being spoken – before the understanding-emergent-in-Conversation can come into play; but that’s merely a background assumption, rather than a topic of analysis for CA.)
A corollary of this is that in CA the analyst can’t argue for the relevance of any issue unless an “orientation” to it on part of speakers is discernible from the transcript. Variables such as gender, class, age etc. are irrelevant, unless there is proof in the transcript that participants are making it relevant for each other. Which makes sense on the very basic level of constructing and exchanging turns in conversation – but any attempt to broaden your argument, and you soon run into problems.
Take, for example, gender. Unless the gender identity of a speaker is made relevant in Conversation – explicitly, or implicitly, e.g. by being systematically denied long turns because of their gender identity – whether a speaker is male or female is not a variable for CA. Of course, comparing across cases – of, for example, all-male and all-female conversations – one could find consistent differences linked to gender; but the absurd conclusion you are forced to take if you follow the CA line religiously is that gender is simply irrelevant as a variable within these interactions. Unless you assume that speakers are also making the same kinds of links, across contexts, that the analyst is… but then there is no transcript-based proof that this is the case. (I think – or hope – that nobody would be insane enough to question the conclusion that gender is relevant in such a hypothetical scenario on the basis of this latter point; but the point is that any such conclusion already moves us away from the strict ’empirical closeness’ badge that CA wears so proudly.)
CA methods are also silent on how gender norms may make only certain kinds of conversations, rather than others, come to happen in the first place: it looks only at what actually happened (i.e., the transcript), not how this situation, with these participants, as opposed to any other, even came to pass – even though the fact that there are these participants present rather than others likely has an influence on how they converse with each other. (This is all relegated to the background, to “broader social context,” something that CA acknowledges but is essentially agnostic upon since it can’t (always) be reliably read off the transcript – even though it’s a crucial factor in how this transcript in particular is available.) Talking about media, if there’s an all-male panel broadcast somewhere, and none of the feisty males present produces a turn for which gender would be an issue… then it’s just not an issue! Gender inequalities aren’t brought up, so there are no power differentials at play, and it’s a perfect public sphere as long as any troublemakers (e.g., women) are kept out of the Conversation.
Picture 2: An example of an interactional context with no displayed power inequalities whatsoever. Hasselhoff-approved. (Via here)
In other words: for CA, unless the analyst can find an evident (to them) “orientation” to a concept such as gender in the transcript they are poring over, that concept is determined to be irrelevant to the interaction. Which is true only to the extent that it is irrelevant to the way the conversational exchange is unfolding at any particular moment. How this exchange is made possible in the first place, or where it might lead to, may well be influenced by “broader context,” but CA is not interested in this as much as in the micro-structures that make talk possible.
And this is the main issue I have with approaches like Hutchby’s. They seem to be convinced that CA is the only possible way in which talk on media can be analysed with an empirical basis for identifying asymmetries of power and other features of context, rather than simply saying such asymmetries are present. The analyst imposes their own understanding, the argument goes, rather than letting the data speak for itself.
But I’m not sure how identifying “orientations” in the micro-structures of talk involves any less imposition of the analyst’s own understanding compared to working on broader scales. In classic CA, at least, there’s the “next turn proof procedure”: the requirement that the other participant demonstrate, in their talk, that they have understood a certain orientation as such before this orientation can be said to be present. In other words, unless my interlocutor confirms (or plays on, or challenges) me making my own or anyone else’s gender relevant in my talk, there’s no basis for the analyst to claim that a certain understanding of the relevance of gender is shared between the two participants. But in media talk, where the interlocutors – the audience – are not present, there are no such subsequent confirmations. There is no way to prove that an orientation that a media analyst of Conversation identifies is actually shared by anyone else but the broadcaster – or, on the contrary, that orientations that the broadcaster does not bring up in their talk may not be relevant for some (or indeed all; there’s just no way to know) members of their audience. For all that, we have to fall back on the analyst’s intuition.
Picture 3: Conversation Analysis in action, as explained by linguistic anthropologists. From Charles and Marjorie Goodwin, “Participation,” p. 222-244 in A Companion to Linguistic Anthropology, ed. Alessandro Duranti, Oxford: Blackwell, 2004.
The fact that it’s the analyst’s intuition that counts causes problems even for classic CA. CA’s charter is to uncover general principles of human Conversation, but Jefferson, Sacks and Schegloff based most of their empirical conclusions and analytical guidelines on data from English. And though CA methods themselves are perfectly neutral in this respect, it’s difficult to see how somebody poring over transcripts in any language would be able to avoid bias without a good knowledge, or at least a vague intuitive sense, of what kind of communication ‘sounds normal’ for that language.
This is, of course, just the kind of work that many linguistic anthropologists do – but linguistic anthropologists also (usually) know the context of the conversations being produced fairly well, given that they’ve spent months or years with the people who produce them. But CA, in its traditional form, imposes no such requirement upon the analyst. One can (maybe even should) do CA based on transcripts and recordings alone, with no necessary knowledge as to where, how, for what purpose etc. such transcripts and recordings were made. But then one is limited to studying the principles of English (or Slovene, or Xhosa, or Yukaghir, or whatever language one can claim to know intuitively) Conversation; since even such simple things as long silences between turns – which might indicate some sort of problem, hesitation or disagreement, for (American) English speakers, but are perfectly normal in e.g. Apache – can vary considerably in what they mean in different cultural and linguistic contexts.
What I’m trying to argue here is not that using CA for analysing media is inappropriate. If you know how to do it, and do it consistently and carefully and with good knowledge of the broader context at hand, it can tell you a lot. It’s only that it is not inherently any less biased than any other kind of analysis of media texts (such as Critical Discourse Analysis, which Hutchby rather histerically seeks to demolish in his Media Talk book). So it’s probably not necessary to cling desperately to CA analytical methods and still produce some kind of sensible argument about the media text that you’re studying.
There are aspects of CA I still find attractive. It gives unrivaled access to the way in which conversation is organised, on a very minute level, and can reveal principles and asymmetries that more broad-based approaches can’t But in analysing talk on Jordanian radio (or any other media context for that matter), I don’t think it makes sense to sacrifice the broader arguments one can make – about power, politics, language, gender – simply to locate one’s analysis more firmly in a particular disciplinary tradition. Transcribing 50 calls from Muhammad al-Wakeel’s programmes according to strict CA conventions and looking at – say – sequences of turn-taking, or types of acknowledgment, or the way people put forward claims as to the veracity or relevance of their problems would of course be perfectly possible. But then I would be studying structures of turn-taking, or types of acknowledgment, or the way people put forward claims as to the veracity or relevance of their problems – and it would still be me deciding how relevant these are to power and inequality as it comes up in radio talk. (After, of course, a (hopefully) well-considered analysis.) So why not look more broadly as well?