Skip to main content

Spoken English: Conversation Analysis

Lecture notes

Course: Spoken English

Week: 4

Lecturer: Keith Richards

Topic: Conversation analysis

These notes are extracts only and do not include the arguments developed in the lectures. Neither do they include handouts or workshop activities. They are an additional resource designed primarily for those who attended the relevant lectures.

Introduction

In the lecture I talked briefly about the ethnomethodological roots of conversation analysis, the idea of the design of talk and its joint construction. The following quotation is a reminder of this relationship:

Related to the accountability of behaviour in interaction is the ethnomethodological notion of the ‘architecture of intersubjectivity’, the means by which individuals participating in the same interaction can reach a shared interpretation of its constituent activities and of the rules to which they are designed to conform.

Taylor & Cameron 1987:103

I also pointed out that although the link between ethnomethodology and conversation analysis is a strong one, subsequent developments have seen differences emerge between conversation analysis and applied conversation analysis (usually synonymous with the analysis of institutional talk). Our interest is likely to be in the area of the latter, usefully summed up in Heritage’s list of ‘six basic places to probe the ‘institutionality’ of interaction’:

1. Turn-taking organization.

2. Overall structural organization of the interaction.

3. Sequence organization.

4. Turn design.

5. Lexical choice.

6. Epistemological and other forms of asymmetry.

(Heritage1997:164)

Even if you are not interested in CA as such, there seem to me to be two good reasons why you should at least familiarise yourself with its approach to analysis:

1. Many of the descriptive concepts used by CA (e.g. turn-taking, repair) have now entered the mainstream.

2. Its approach to analysis is the most painstaking and fine-grained of any available, so it provides an excellent means of developing and honing your analytical skills

Text for analysis

Many of the examples in the following sections will be drawn from the following text. You might like to attempt a preliminary analysis yourself before continuing. Details of transcription symbols used are available at:

01 A: Lot of rain last night. I was surprised.

02 (1.5)

03 B: We let the dog out and e:r (.) Tom came down as well so I

04 looked up and said, ‘(Tom,) the dog’s whining.’

05 A: Heheheh

06 B: ‘Where is she?’ .hhhh! She was absolutely soaked.

07 >Because we’d< closed the: er door which we don’t usually do.

08 A: Right.

09 (1.0)

10 A: And he normally runs into the garage, does he?

11 B: Yes. Yes. He has either the garage or the garden to

12 choose from. Heheheh!

13 (1m 3s)

14 C: Strange set up, isn’t it?

Note: Italics here are used to indicate utterances

delivered while laughing.

Turn-taking

In their seminal work on this subject, Sacks et al. (1974, 1978) begin by identifying certain ‘grossly apparent’ facts, such as the fact that speaker change does occur, that there are brief overlaps but transitions without gaps are common, and that turn allocation techniques are used. They then describe a turn-taking device that accounts for these. Part of the explanation involves the idea of a turn-constructional unit (TCU), such as a sentence, clause, phrase or even word (e.g. ‘Hello’), recognised as a meaningful unit by participants. At the end of this is a transition-relevance place (TRP), which marks the boundary of such a unit.

In the following extract, why is there only one TRP in B’s turn and how is it signalled?

03 B: We let the dog out and e:r (.) Tom came down as well so I

04 looked up and said, ‘(Tom,) the dog’s whining.’

05 A: Heheheh

What we notice here is how B constructs her turn so that the first opportunity for anyone else to respond comes at the end of her comment on the dog. Even the short hesitation and micropause, ‘e:r (.)’ do not represent a TRP because they are preceded by ‘and’, which indicates that more is to follow. Similarly, there is no intonational signal in ‘as well’ to indicate that this is a place where the turn might transfer. Where the falling intonation (indicated by ‘.’) signals a TRP, A’s laughter is timed to pick up on this. Such fine tuning, typical of conversation, is possible because the end of a turn-unit is predictable (projectable). A’s laughter is a response to B’s own laughter and does not interrupt the flow of the story.

Having pointed to what is ‘grossly apparent’, Sacks et al. go on to propose rules for the operation of turn-taking system, which can be summed up (crudely) as follows:

(a) First speaker (the speaker currently holding the floor) selects next. In this case, the person selected is obliged to speak and no others have the right. Transfer to the person selected occurs at the first TRP after selection.

(b) If the first speaker does not select next, then any other party may self-select at the TRP. The first speaker acquires the right to the turn and transfer takes place.

(c) If nobody has been selected and nobody self selects, then the current speaker may continue, or may allow the conversation to lapse. If the current speaker does continue, then (a)-(c) reapply at the next TRP until speaker change occurs, or the conversation lapses.

Identify the TRPs in the following extract from Text A and decide which of (a)-(c) applies:

06 B: ‘Where is she?’ .hhhh! She was absolutely soaked. >Because

07 we’d< closed the: er door which we don’t usually do.

08 A: Right.

09 (1.0)

10 A: And he normally runs into the garage, does he?

11 B: Yes. ...

Since the questioning intonation in ‘Where is she?’ relates to an ‘internal’ question (hence my use of quotation marks), this is clearly not a TRP, and the sharp intake of breath which follows it serves as a signal that the speaker is about to continue. However, the falling intonation signals a TRP at the end of ‘soaked’ (06), and when nobody self-selects the speaker continues (c). The TRP at the end of line 07 is emphatic and (08) A self-selects (b). His short turn (an example of a TCU which consists of only one word) acknowledges receipt of the information and allows B the opportunity to self-select and provide further information. However, when she fails to do so (09), A continues (c) with a question (10) which obliges B to respond (a). We see, then, an application of each of the three rules which Sacks et al. identify in their work on turn-taking.

Gaps and lapses

It’s also very clear that conversation includes silences as well as speech. Three sorts of silence have been identified, two of which, gaps and lapses, are relatively straightforward.

A gap is simply a pause in the flow of talk, sometimes occurring within turns, sometimes between turns. Such silences are a natural part of any interaction and have no particular significance in themselves. Occasionally, though, the talk itself will stop only to be resumed later, in which case the resulting silence is defined as a lapse. Lapses tend to be longer than gaps (although this is not necessarily the case) and can be constructed in various ways. Often there is an obvious topic termination or fading out, followed by a long silence and then by the initiation of a new topic (occasionally by a re-initiation of the same topic). In fact, there are no specific criteria which would allow us to distinguish a gap from a lapse definitively, but a change of topic following a period of silence or an extremely long silence are good guides.

Identify the gap and the lapse in the following:

08 A: Right.

09 (1.0)

10 A: And he normally runs into the garage, does he?

11 B: Yes. Yes. He has either the garage or the garden to choose

12 from. Heheheh!

13 (1m 3s)

14 C: ((Referring to a recent proposal)) Strange set up, isn’t it?

In 09 the short silence is a gap in which another speaker might self-select in order to provide a turn oriented to the ongoing talk, whereas the long silence (1 minute, 3 seconds) in 13, followed by the introduction of a new topic (the ‘strange set up’ refers to something completely different) indicates that this is a lapse. Of course, had the ‘strange set up’ referred directly to the previous talk and been interpretable as an evaluation of the anecdote, it would be more difficult to be decisive about the nature of this silence despite its considerable length.

Attributable silence

There is, a third sort of silence, and it is one with particular interactional significance. Levinson (1983) quotes an excellent example of this, taken from Atkinson & Drew (1979:52):

A: Is there something bothering you or not?

(1.0)

A: Yes or no

(1.5)

A: Eh?

B: No.

According to the first rule in the turn-taking mechanism we have already considered, A has selected B (by addressing a question directly to him/her) and B is obliged to reply. When this doesn’t happen, A reframes her/his utterance as a polar question, requiring only a yes/no response, and when this still doesn’t produce a reply A uses a prompt, which finally elicits a negative answer. These silences are not simply gaps; B has noticeably (and accountably) failed to conform to recognised norms of interaction. A significant absence such as this is known as an attributable silence, which needs to be accounted for. Whether or not the account is present in the talk doesn’t really matter — in practice, such accounts are rarely requested or provided: the possibility of attribution is what’s important.

The adjacency pair

This idea of an attributable silence relates to situations where we can predict that a conversational slot should be filled, and this in turn raises the possibility that there is a category which describes the intimate linking of two turns, so that the first predicts the second and the second serves as a response to the first.

In fact, there is such a category, and the term adjacency pair captures the two main features of it: that it is designed as a two part sequence, and that the two parts are usually adjacent. In simple terms, these are paired utterances in which the production of a first part by one speaker (e.g. a question) invites the production of a corresponding second part by another speaker (e.g. answer). The first pair part is said to be conditionally relevant, because when it is uttered the second pair part is made relevant and predictable. If the second fails to appear, it is noticeably absent, and if another first pair takes its place this is assumed to be a preliminary to completing the second part. Schegloff & Sacks (1973:296) put it this way:

The basic rule of adjacency pair operation is: given the recognizable production of a first pair part, on its first possible completion its speaker should stop and the next speaker should start and produce a second pair part from the pair type of which the first is recognizably a member.

Schegloff & Sacks 1973:296

It is important to remember, though, that this is not a determinant of a speaker’s behaviour, it simply provides a means of orientation: the speaker can choose to ignore the second part, aware of the particular interactional consequences of doing so. The importance of the adjacency pair lies in its interactional function:

What two utterances produced by different speakers can do that one utterance cannot do is this: by an adjacently produced second, a speaker can show that he understood what a prior aimed at, and that he is willing to go along with that. Also, by virtue of the occurrence of an adjacently produced second, the doer of a first can see that what he intended was indeed understood, and that it was or was not accepted. ... It is then through the use of adjacent positioning that appreciations, failures, correctings et cetera can themselves be understandably attempted

Schegloff & Sacks 1973:297-98

As a normative mechanism the adjacency pair is clearly very powerful, and some writers have gone so far as to claim that it lies at the very heart of CA. It features prominently in many of the seminal papers in the field, and the basis of such a claim as summed up by Taylor and Cameron is certainly instructive, if somewhat exaggerated:

The concept of the adjacency pair is, arguably, the linchpin of the ethnomethodological model of conversational structure. Not only, as we have seen, does the operation of the turn-taking system rely upon it, but also nearly every other structural feature so far identified by conversation analysts somehow incorporates the adjacency pair into its formulation (cf ‘openings’, ‘closings’, ‘repair’, ‘story-telling’ etc). And, at the same time, it is the notion of the adjacency pair that the ethnomethodological principles on which CA is based are most usefully and obviously employed. Without the concept of the adjacency pair, there would be no ethnomethodological model of conversation; and in turn, without the ethnomethodological principles of accountability and of the sequential architecture of intersubjectivity, there would be no concept of the adjacency pair.

Taylor and Cameron (1987: 109)

If this is interpreted as a claim that the adjacency pair is somehow an archetypal example, it is perhaps acceptable. However, as other writers (e.g. Levinson 1983:304) have pointed out, if it is treated as an attempt to reduce CA to a matter of adjacency pairs, it presents an impoverished view of what is possible.

Can you find an example of an adjacency pair in the text for analysis? What other sorts of ‘pairs’ might you expect to find in conversation?

The example I have in mind occurs on lines 10-11. Notice how A stops at the completion of the first pair part and B’s response is an immediate and direct response to this. The second pair part would be meaningless without the first, and the first would have been ‘left hanging’ had B not provided a response to it. Other examples of adjacency pairs would be greeting-greeting, invitation-acceptance, accusation-denial, assessment-agreement etc.

Note

Although adjacency pairs are normally adjacent, sometimes further information may be required in order to produce the second pair part. In such a case, there may be an insertion sequence consisting of a further adjacency pair. The following example (Merrit 1976:33, quoted in Levinson 1983:304) illustrates this:

A: May I have a bottle of Mich?

B: Are you twenty one?

A: No

B: No

Preference

You may have noticed that in some adjacency pairs the second pair part may take two forms: positive and negative. So if I invite you to join me for a drink, you have the option of accepting or declining. The terms used for these options are preferred and dispreferred responses. Unfortunately, the terms suggest a psychological dimension that has no place in CA, and a fair amount of heat has been generated on the subject of their status. This is a debate that you do not need to pursue, and the best way of thinking about it in terms of speaker design: usually, the first speaker will design their turn to elicit a particular response and if the response is as predicted it is preferred, if not it is dispreferred. Levinson (1983) has also noted that while preferred responses tend to be unmarked:

A: Why don’t you come up and see me some┌times

B: └I would like to.

Atkinson & Drew 1979:58

dispreferred ones are often marked (by hesitations, fillers, accounts, etc.):

A: Uh if you’d care to come over and visit a little while this

morning I’ll give you a cup of coffee.

B: hehh Well that’s awfully sweet of you, I don’t think I can make

it this morning ⋅hh uhm I’m running an ad in the paper and- and

uh I have to stay near the phone.

Atkinson & Drew 1979:58

Repair

More interesting from our point of view is the concept of repair. When considering this concept, it’s worth bearing in mind the idea that conversation is built up (designed) by the participants. When that process of building goes wrong, it’s necessary to put things right, to get back on track. Hence the idea of ‘repair’, rather than ‘correction’. For example, if you’ve misunderstood something, or I’ve given incorrect information, it may well be that we start talking at cross purposes, and it will be necessary to ‘repair’ this. (If you’re interested in this topic, Schegloff et al. 1977 offer a detailed analysis of repair, the results of which are succinctly summed up by Levinson 1983:340-42. For a focus on the ELT classroom, see chapter7 of van Lier 1988 or chapter 4 of Seedhouse 2004).)

The place in a conversation where the problem initially arises is known as the trouble source, the point where the repair begins is referred to as the repair initiation, and the putting right is referred to as the repair itself. The movement through this is known as the repair trajectory. As we shall see, some repair trajectories can be very short, but it is possible to have extended trajectories, where the repair is carefully negotiated.

There are also distinctions made on the basis of whether the person responsible for the trouble source (‘self’) or someone else (‘other’) is responsible for the initiation and/or repair. To illustrate this, and other aspects of the repair process, I’ll take examples from my own data, all from the same discussion.

Self-initiated self-repair

Self-initiated self-repair (SISR) occurs in one of two places: either within the speaker’s own turn and before a TRP, or following a TRP. The obvious difference here is that the ‘first opportunity’ (before the TRP) example is an option available only to the speaker, whereas at the TRP it is possible for some other speaker to initiate repair. This sort of repair is what Schegloff et al. (1977) refer to as the ‘most preferred’ form. In the following example we see SISR operating in the case of ‘don- don’t’ and again after ‘produce’, where the speaker replaces ‘produce’, which has a specific meaning in the context of broadcasting, with ‘write things’, which is less ambiguous:

‘Well don- don’t you produce- don’t you write things for radio?’

Had the speaker continued after the question mark (a TRP) and repaired, for example, ‘radio’, replacing it with ‘television’, this would still have been SISR, but at the second opportunity.

Other-initiated self-repair

Sometimes a speaker says something which is not clear to another participant, or which seems in some other way problematic, and is given the opportunity to repair this. Here we see K referring to Harry as though the referent were clear, when this is not the case (trouble source). However, J initiates repair by asking which Harry is involved, and K immediately makes this clear. J confirms this and the discussion continues:

K: =liaised with Harry as well to see what-=

J: =Your Harry or ┌our Harry?┐

K: └our Harry ┘=

J: =Your Harry=

Other-initiated other-repair

Although this is regarded by Schegloff et al. as the least preferred form of repair, it does occur. In this form, someone other than the speaker initiates and someone other than the speaker repairs. Usually, as in the example below, they are the same person. J has made a mistake about the nature of the board (it is not ‘wrong’), and K initiates repair. J attempts to self-repair, commenting on J’s ownership of the board, but this too is inaccurate and is followed by other-initiated other repair from K. J’s ‘A:h’ acknowledges the repair:

K: God knows what they’ll pick up. (1.0) Tom Smith will love the examining.

(0.5)

J: Yeah. Well (.) it’s the wrong (.) board, but still.

K: No it isn’t. He em

J: Oh yes he owns it (.) doesn’t he.

K: No he masterminded e:m (1.0) e:::m (.) ((Name of exam))

J: A:h

Self-initiated other-repair

This is a form of repair not considered important by Schegloff et al., but it is quite common in the classroom and occasionally occurs in ordinary conversation. In this example, A is taking notes and has announced that she was attending to writing rather than the discussion, so when she explains the problem, we see a slowing down as she announces the name, which is produced with emphasis and sound stretching, inviting repair. Instead, K puts her on the spot, and she completes the name only after hesitation (“em”) and a pause. This then produces other repair of the name she has offered from both K and P:

A: >It was something about- (.) w-where you were talking about< Ke:n=

K: =What?=

A: =em (.) Roberts.=

K: =No ┌Ken Foster.

P: └Ken Foster.

A: Ken Foster right.




Sometimes it can be difficult to decide whether something counts as a repair. For example, the garage has not been mentioned in the text for analysis when A introduces this in line 10, and it's hard to judge from the interaction whether it's designed simply to elicit more detail or whether it invites clarification of the 'door' referred to. Similarly, not all opportunities to repair are taken up. Can you find in the text for analysis above an example where the opportunity to repair arises but it is not taken?

This can be found in the extract below. Although B refers to her dog as ‘she’, A in 10 fails to pick this up and uses ‘he’ instead, repeating this in the question tag. B could respond with ‘embedded’ repair (Jefferson 1987) by repeating ‘she’ in her next turn (11), but the preference for agreement overrides this, and she uses ‘he’, aligning with A’s formulation:

06 B: ‘Where is she?’ .hhhh! She was absolutely soaked. >Because

07 we’d< closed the: er door which we don’t usually do.

08 A: Right.

09 (1.0)

10 A: And he normally runs into the garage, does he?

11 B: Yes. Yes. He has either the garage or the garden to choose

12 from. Heheheh!

Sometimes it’s best just to leave things alone.

Four basic rules for CA

As we’ll see in the Friday workshop, CA analysis is a painstaking business and it’s not easy to sum up the process in a few well chosen points. However, the following four basic rules should always be adhered to:

1. Use naturally occurring data

This is the most basic condition of research: invented data is never used, even for the purposes of illustration. Sacks emphasises the point repeatedly: ‘the kind of phenomena I deal with are always transcriptions of actual occurrences in their actual sequence’ (1984:25; see also 1985:13). Atkinson and Heritage (1984:2) are more emphatic: ‘Within conversation analysis there is an insistence on the use of materials collected from naturally occurring occasions of everyday interaction by means of audio- and video-recording equipment or film.’

2. Move from observation to hypothesis

Conversation analysis is not hypothesis testing. The analyst’s aim is to treat the talk as something fresh, something to be approached in its own terms. Sacks argued that ‘using observation for theorizing’ has certain advantages:

Treating some actual conversation in an unmotivated way, that is, giving some consideration to whatever can be found in any particular conversation we happen to have our hands on, subjecting it to investigation in any direction that can be produced from it, can have strong payoffs … Recurrently, what stands as a solution to some problem emerges from unmotivated examination of some piece of data, where, had we started out with a specific interest in the problem, it would not have been supposed in the first instance that this piece of data was a resource with which to consider, and come up with a solution for, that particular problem.

Sacks 1984:27

3. Rule nothing out

This derives directly from the first two points and might therefore be subsumed under them. Atkinson and Heritage (1984:4) make the point well: ‘nothing that occurs in interaction can be ruled out, a priori, as random, insignificant, or irrelevant.’

4. Focus on sequences

Conversation is jointly constructed, and if we are to understand the ‘architecture of intersubjectivity’ we must treat each utterance in the context of its response to what has gone before and its relevance to what follows. ‘For conversation analysts, therefore, it is sequences and turns within sequences, rather than isolated sentences or utterances, that have become the primary units of analysis.’ (Atkinson and Heritage 1984:5)

A note on transcription

This is an aspect that has received a fair amount of attention in the literature over the years. Without wishing to trivialise the issues, it seems to me that there are a couple of essential points that need to be borne in mind:

  • A transcription is no more than a best effort to represent spoken interaction in written form. There is no such thing as a final or definitive transcript.
  • The transcription system used should relate to the purpose of the research. For example, when transcribing interviews as part of a field study there is no need to use the same level of delicacy as in transcriptions of casual conversations for the purposes of CA.

The following observation should help you to set these points in their broader research perspective:

Transcribing... is a political act that reflects a discipline’s conventions as well as a researcher’s conceptualization of a phenomenon, purposes for the research, theories guiding data collection and analysis, and programmatic goals

Green et al. 1997:172

It’s also worth bearing in mind that when you become involved in the process of transcription, you will inevitably hear things in different ways at different times. As Stubbs (1983:228) puts it, ‘auditory hallucinations are real problems, both practically and theoretically.’

One of the ongoing debates in transcription centres on whether to use standard orthography or what has been called folk transcription (Stubbs 1983) or eye dialect (Roberts 1997). The following are worth examining with this in mind:

Examine the following two transcriptions of the same exchange and decide what differences the transcription makes to our interpretation of them:

Text A

A: Should have put one up somewhere shouldn’t we really? We

need a social organiser.

B: Yes you should. Oh, is that your responsibility?

Text B

A: Should have put one up somewhere shouldn’t we really. We

need a social ┌owganisah

B: └Yeah you should- (.) Oh is that your

responsibility?

(Note: Words in italics are spoken while laughing.)

If we compare Text A with Text B, we can see that all words are represented in the first text, but this text reads very differently from the version which follows. The inclusion of a ‘grammatical’ question mark and the period after ‘should’ lead to an interpretation along the following lines: A is addressing a question to B, who responds by challenging A directly (‘you should’) and then has second thoughts and checks that getting things put up is A’s responsibility. This ‘tidied up’ transcript is easy to read but a travesty of what actually occurred. In fact, A’s statement sets up an ironic comment on his own position as social organiser, and B’s initial response is cut off short when she realises that A is the social organiser. She then responds to the joke. Part of the misunderstanding arises because B has interrupted before A has completed the ironic comment his first utterance.

Notice also, in terms of the key within which the exchange is conducted, the importance of the way ‘social organiser’ is transcribed. The ‘eye dialect’ n Text B captures the humorous frame which Text A misses. And what about ‘Yeah’? This is not a trivial point, because the ‘Yes’ in Text A looks much more formal, perhaps more of a challenge than an agreement. In fact, an examination of the database from which this is drawn reveals that speaker B usually uses ‘yes’ rather than ‘yeah’, so the latter is important here.

Conclusion

I’d like to conclude with a final reminder about the nature of CA. It would be all too easy to treat the brief summary of conversational features in this lecture as what CA is really about, but this would be a serious mistake. Analysis involves much, much more than the identification of such features; it is, in essence, an attempt to understand the ways in which interactants use such resources to construct a particular interaction, one which reflects their shared understanding of ‘what is going on here’.

The aim of the workshop is to offer you the chance to practise some basic analysis in which you should be able to draw on many of the things covered above. However, in order to understand the sequential construction of the relevant talk, you’ll need to go far beyond a mere identification of specific features.

References

Atkinson, J.M. and Drew, P. 1979. Order in Court. London: Methuen.

Atkinson, J.M. and Heritage, J. 1984. Introduction. In J.M. Atkinson and J. Heritage (eds), Structures of Social Action, pp.1-15. Cambridge: Cambridge University Press.

Green, J., Franquiz, M. and Dixon, C. 1977. The myth of the objective transcript: Transcribing as a situated act. TESOL Quarterly, 31(1):172-176.

Heritage, J. 1997. Conversation analysis and institutional talk. In D. Silverman (ed), Qualitative Research: Theory, Method, Practice, pp.161-82. London: Sage.

Jefferson, G. 1987. On exposed and embedded correction in conversation. In G. Button and J.R E. Lee (eds), Talk and Social Organisation, pp.86-100. Avon: Multilingual Matters.

Levinson, S.C. 1983. Pragmatics. Cambridge: Cambridge University Press.

Merrit, M. 1976. On questions following questions (in service encounters). Language in Society, 5(3):315-57.

Roberts, C. 1997. Transcribing talk: issues of representation. TESOL Quarterly, 31(1):167-72.

Sacks, H. 1984. Notes on Methodology. In J.M. Atkinson and J. Heritage (eds), Structures of Social Action, pp.21-7. Cambridge: Cambridge University Press.

Sacks, H. 1985. The inference-making machine: notes on observability. In T. Van Dijk (ed.), Handbook of Discourse Analysis Vol.3: Discourse and Dialogue, pp.13-24. London: Academic Press.

Sacks, H., Schegloff, E.A. and Jefferson, G. 1974. A simplest systematics for the organization of turn-taking for conversation. Language, 50(4): 696-735. Later, variant version (1978) published in J.N. Schenkein (ed), Studies in the Organization of Conversational Interaction, pp.7-55. New York: Academic Press.

Schegloff, E.A. and Sacks, H. 1973. Opening up closings. Semiotica, 7:289-327.

Schegloff, E.A. Jefferson, G. and Sacks, H. 1977. The preference for self-correction in the organization of repair in conversation. Language, 53:361-82.

Seedhouse, P. 2004. The Interactional Architecture of the Language Classroom: A Conversation Analysis Perspective. Oxford: Blackwell.

Stubbs, M. 1983. Discourse Analysis: The Sociolinguistic Analysis of Natural Language. Oxford: Basil Blackwell.

Taylor, T.J. and Cameron, D. 1987. Analysing Conversation: Rules and Units in the Structure of Talk. Oxford: Pergamon Press.

van Lier, L. 1988. The Classroom and the Language Learner. London: Longman

Recommended reading

The introductory literature on CA is growing all the time and the relevant section in Cameron, but I also strongly recommend two other books. If you’re interested in this area or in classroom talk, Seedhouse combines an excellent introduction to CA with an ELT focus, which makes his book essential reading. Ten Have provides what has become the standard basic introduction to CA and its approach to analysis, so I would also want this on my shelves.

Cameron, D. 2001. Working with Spoken Discourse. London: Sage.

Seedhouse, P. 2004. The Interactional Architecture of the Language Classroom: A Conversation Analysis Perspective. Oxford: Blackwell.

ten Have, P. 1999. Doing Conversation Analysis: A Practical Guide. London: Sage.