Analysing Talk In Interaction
Lectures 1 2 3 4 5 6 7 8 9 10 11 12 13

Charles Antaki
Loughborough University

Lecture / Seminar 4: Conversation analysis I - the basics

The appendix to this lecture


As I mentioned in Lecture 3, my recommended text for all the CA lectures is Hutchby and Wooffitt. But the others are good too.

Hutchby, I. and Wooffitt, R (1997) Conversation Analysis. Oxford: Blackwell
Nofsinger, E. (1990) Everyday Conversation
ten Have, P. (1998) Doing Conversation Analysis. London: Sage
See also Chapter 6 in Levinson (qv) for an excellent, but dense, critical account.


The best way to start to explore CA is to see what it says about how people manage the (apparently simple) business of holding any sort of conversation. 


The ground floor of the local management system that CA identifies is the way people exchange the opportunity to speak - to 'turn-take'. This, on inspection, turns out to be more flexible and complex than our intuitions might suggest. People are fantastically good at it, though.

The term "adjacency pair" is a formalisation of our intuitive understanding that conversations proceed by participants taking up what the last person has said in a regular way - that the two utterances together form what is called an 'adjacency pair'.

First pair parts (FPPs) (like questions, invitations and orders) are regularly followed by certain kinds of responses, or second pair parts (SPPs) (which in these cases would be answers, acceptances, and compliances respectively).

[Apology: I shall use invented data for these very initial examples, which is wrong, but convenient. In the appendix to this lecture, and in future lectures, I'll stick to real extracts of talk]

A: How are you? [greeting]
B: How are you? [reciprocation]


A: Mum? [summons]
B: Yes? [acknowledgement]


A: Have you got any carrot-flavour crisps [request]
B: Sure (puts packet on the counter) [compliance]

And if these don't follow immediately, they follow eventually, after some other sequence has been inserted into its brackets.

A: Have you got any peanuts [request]
B: What, the roasted ones
A: Yes
B: Sure (puts packet on counter) [compliance]

Typical pairs are: greeting/reciprocation, summons/acknowledgement, request/compliance (all of which we saw above), assertion/agreement, question/answer, etcetera. These are all highly predictable.

But looking at conversations we often find that a first pair part (eg a question) is sometimes followed by something that is clearly not an 'answer' in the required sense - it might be a refusal to answer, a redirection to somebody else, a challenge to the questioner's right or competence to ask that question, and so on. It looks as if this destroys the predictiveness of the CA claim. But in fact it turns out that these non-expected SPPs do have some regular features. The clue is in how the 'answer' is given - in what CA calls the preference organisation.

Preferred and Dispreferred

Examination of unexpected responses (ie an SPP not typically normal) shows that they are marked as being different from 'answers'; so the idealisation of a proper answer is there, and oriented to by the speakers, even though a 'non-answer' is given. If the second part of the adjacency pair is not one of these normal responses, then it will be done differently: a bit less readily, and generally 'marked' in some way - by a hedge, or a request for clarification, or both.

A: "why don't you come to our housewarming party on Saturday?"
B: "Well I'd like to but it's Hannah's birthday" [marked rejection]

This latter is an example of what is called a 'dispreferred' response. The rejection is (it is empirically found) marked by hesitation and hedging and an account of why the preferred response wasn't given. The mark is so powerful that it alone will suffice as a rejection:

A: "why don't you come to our housewarming party on Sunday?"
B: "Well ..."

And A knows that B is declining the invitation.

Imagine what would happen if you gave the dispreferred reply without marking it:

A: "why don't you come to our housewarming party on Sunday?"
B: "No"

That would look strange and rude. You would infer something about what B was saying (e.g. that they were sulking). The informativeness of such deviation shows us that the substance of the dispreferred SPP (e.g. that it is a rejection) and its markers (e.g. a pause, a hedge) normally go together.

So you have four possibilities: (commonly) expected and unexpected answer which can be either marked or unmarked. Commonly expected answers tend strongly to be unmarked.

Psychological or statistical?

It is important to appreciate that 'dispreferredness' is not a psychological evaluation of the response. It's purely a frequency judgment. The more frequent response to a greeting enquiry about your health is 'fine, thanks, and you?'. But it's not meant to be an accurate report. It's just a feature of the system that it has 'standard' responses. It's useful because if you want to communicate that you're not fine, then all you need to do is hesitate and delay. Your listener will work out that you're giving the 'non-standard' response (and, in this case, are therefore not well).

E.g. compare (both these are invented)

A Hi, how are you?
B Fine and you? [even though B may be quite sick]


A Hi, how are you?
B (.3) Well....

Larger sequences

CA also moves up to describe conversational moves that need a more complex organisation of turns. For example, presequences.


A you know that book I lent you?
B the linguistics one?
A no, the sociology one
B yes?
A well have you finished with it it yet?

CA can also describe the overall organisation discrete chunks of conversations (e.g. stories) or even of (at least some kinds of) entire conversations. Probably the best studied of these are telephone conversations, which seem to fall into regular stages: the opening section; the first topic slot' next topic slot (often inaugurated by the 'hey!1 etcetera; and the closing section.

E.g. (stereotypical)

(phone rings) {summons
B Hello {acknowledgement, and offer of identification
A Hi, it's Charles {first greeting, and offer of identification
B Oh hi {reciprocation
A How was your weekend? {first topic
B Oh fine, fine
A Ok..listen, you know that book you wanted {termination of first topic and
prerequest signalling next topic
B yes? {permission to raise topic
well they don't have it at the bookshop {next topic

A Ok then, so I'll find out about Thursday {start of closing phase by future promise
B yeh
A and be in touch
B Right {termination of topic
A Right {acknowledgement of termination
B Ok then {preclosing: warning of closedown of turntaking
A Ok. See you {acknowledgement of imminent closedown
B Bye {closedown FPP
A Byw {closedown SPP
(puts down receiver)

Note that this invented dialogue offends against pure CA who would never do this other than as a convenience and under protest. They would always work with the actual transcript. But that's much less easy to read, as we will see.

Empirical discovery, not analytic rules

It is important to note that CA presents itself as discovering patterns empirically. It doesn't start with a theory about what conversationalists are up to; in fact, CA objects to other linguists' theorising before the facts. These facts are regularities, from which the CA will draw inferences about the rules to which the participants are playing. Once those rules are understood by the analysts, they can look at a particular play and make a guess at what function it served the speaker. A reasonable metaphor would be learning the rules of a game by watching it over a period. You pick up certain regularities (for example, the play is stopped if the ball is passed to a striker who is behind the opposition's players) and you suppose that the players knowingly orient to the rule that underpins the regularity. Once you know that, you see why, for example, a team tries to make sure it has all its players in front of the opposition's leading player so that the ball can't be passed to them without 'off-side' being called. 


For this seminar: Discussion of some conversational structures

It will be useful to do an exercise in which we examine some transcripts to spot some examples of some of the terms I have been using so far.

You'll find the data in this lecture's appendix file. Look through and see if you can find examples of the following:

adjacency pairs
first and second pair parts
preferreds and dispreferreds
insertion sequences

For the next seminar:

Read more of Hutchby and Wooffitt.

back to top

A course for the University of Southern Denmark, Odense 2003