The spotlight on Conversational Analysis

Its practical applications are increasingly recognised and, especially over the past five years or so, it has been used in applied contexts to address the real-world concerns of service-providers, clients, user groups, stakeholders and policy makers. Let us start by highlighting some of these.

Selected Practical Applications

Conversation Analysis can play a role in a wide range of areas, including:

• Interdisciplinary impact on sociology, linguistics and social psychology
• Doctor-patient interaction (Heritage et al, 2006); (Mangione-Smith et al., 2006); (Stivers, 2005)
• Police interactions
• Counselling interaction
• Political communication – the understanding of political rhetoric
• Human-computer interaction and software design
• Treatment of speech disorders
• Interviewing, focus groups and survey methodology and the optimisation of response
• Improving public speaking (Atkinson, 1984)
• Student participation in classroom interaction
• Second language acquisition

What is conversation analysis?

CA is a theoretically and methodologically distinctive approach to understanding social life. It emerged in California in the 1960s from the work of Harvey Sacks and his collaborators Emanuel Schegloff (Schegloff, 2007) and Gail Jefferson. Its intellectual roots lie in the sociological tradition of ethnomethodology, an approach primarily concerned with how ordinary activities get done and people’s ways of making sense of the everyday social world.

Sacks did not set out to study conversation, but as a Fellow at the Center for the Scientific Study of Suicide in Los Angeles in 1963, he had access to recorded telephone calls to the Suicide Prevention Center. His analysis shows that some activities in these calls are neither ordinary nor everyday: accounts of suicide threats getting laughed off, and descriptions of suicide attempts to ‘discover if anyone cares’.

Nevertheless, even these conversations were full of activities most of us do every day: complaining, requesting, agreeing, and disagreeing and so on. Sacks set about analysing in detail just how these activities get done. His focus, in other words, was not primarily on the topic of suicide. Rather, he saw talk (be it about suicide or doing the dishes) as a fundamental way in which social action gets done.

Two features of his initial dataset were crucial for CA’s development. First, the data were naturally occurring, rather than researcher-generated. Sacks was analysing the record of an actual piece of social life rather than a second-hand report of it. This meant the speakers were concerned with pursuing actions in their own lives – not simply answering research questions. Second, the data were recorded, allowing Sacks to listen to them repeatedly. This enabled a much more detailed and nuanced analysis than is achievable with observational field notes or recollection. These two features of data collection remain key.

Even today, the significance of this methodological shift should not be underestimated. Focus groups and interviews remain the method of choice for the majority of qualitative researchers despite the well-recognised problems posed by researcher-generated data.

Put simply, are our research participants telling us what really happened? People’s recall is notoriously partial, possibly faulty, and heavily influenced by social desirability effects. In telling about an event, people inevitably put a ‘slant’ on it. In interview or focus group research, then, researchers face the difficulty of how to theorise the relationship between participants’ talk and the things they’re talking about. There is a gap between the data (talk about an experience) and what researchers are usually most interested in (the experience itself) (see Kitzinger, 2004).

CA sidesteps this issue entirely by examining the experience itself. In Sacks’ early work, he did not ask suicidal people to talk about their experience of calling the helpline – as so often happens in customer satisfaction surveys – but analysed instead what actually happened during real-life calls. His recordings captured the silences, overlapping talk, and dynamics (emphatically, softly or loudly, with an upward or downward intonation, and so on) that are a feature of real-life interaction.

This is crucial because CA has shown that such minutiae can be consequential for an interaction. No researcher can make notes at that level of detail, and no participant can be expected to remember exactly what they said and how they said it during a past conversation. CA, however, offers the means to understand social action as it actually happens.

Key Assumptions

Sacks’ analytic approach was based on three theoretical assumptions, which remain central to CA today (Heritage, 1984):

[i] that talk is a form of action rather than simply a ‘flow’ of information from one person to another. For instance, people use talk to do things like complaining, requesting, agreeing;

[ii] that talk is structurally organised and orderly, rather than purely idiosyncratic or ‘messy’. So turns at talk are systematically related to one another, as, for example, when an acceptance or rejection follows an invitation, or a self-deprecation follows a compliment.

The final assumption, [iii], is that talk creates and maintains intersubjectivity. A first speaker understands, by what a second speaker does, how the second speaker heard their first turn. So if I ask “why don’t you come and see me sometime” and you say “yes,” I understand that you heard my turn as an invitation, whereas if you say “sorry, I’ve been meaning to but I’ve been so busy,” I understand that you heard it as a complaint. CA is about understanding these small slices of everyday life. It is an empirical programme with few theoretical manifestoes, and it offers a set of methods for the systematic identification of the patterns, structures and practices of interaction. Our central concern is to use these tools to demonstrate what is going on in the interaction for the participants themselves.

Data Collection

Conversation analysts typically work with recordings of naturally occurring interactions. The limitations of early recording technologies meant that CA was founded largely on the analysis of audio-recordings.

Video-recordings were introduced relatively slowly from the mid-70s on. There is now a substantial set of published work based on video analysis of interaction. CA has increasingly developed an interest in bodily behaviour (gesture, gaze, pointing, posture, etc.) and the relationship between this and accompanying speech.

A comprehensive transcription notation representing various characteristics of the timing and delivery of talk was invented by Gail Jefferson and is continually being refined (Jefferson 2004). This key shows only the notation used in the data displayed in this article

We show some transcripts that use this notation in this article. However, it is the recordings themselves that constitute CA’s primary data, and transcripts are regularly revised during the analytic process.

Provided that appropriate ethical clearance is obtained, the researcher’s colleagues, clients, user-groups, stakeholders and others can also be given access to the recorded data. Best practice in contemporary CA is to make the data extracts displayed in published articles available on the web. So, for example, you can listen to many extracts from Schegloff’s articles at:

www.sscnet.ucla.edu/soc/faculty/schegloff

Data Analysis

Each interaction is unique. It occurs in particular circumstances between people who bring to the encounter their own experiences, beliefs, ambitions, and the history of their relationship.

But conversation analysis shows that each interaction also follows some familiar patterns. CA is concerned with identifying and explaining those patterns. Data analysis draws on a large and growing body of research which offers technical specifications of recurrent patterns in talk.

Such patterns are evident in the two data extracts featured. When people have bad news to tell, there is a recurring feature: they create conditions under which it is the recipient of the bad news who actually says it. (We are drawing here on Wilkinson & Kitzinger, 2008, who summarise the earlier findings of Schegloff, 1988, Drew, 1984 and Maynard, 1997.) These transcripts have been slightly simplified (to make them easier for non-specialists to read). You may also want to refer to the Transcription Key box on the previous page.

News of a death

The bad news in the first data fragment (taken from a telephone conversation between two friends, Belle and Fanny) is the death of someone they both know.

01 Bel: ... I, I-I had something (.) terrible t'tell you.=
02      =So [uh:  ]
03 Fan:     [How t]errible [is it.]
04 Bel:                    [.hhhhh]
05      (.)
06 Bel: Uh: ez worse it could be:.
07      (0.7)
08 Fan: W'y'mean Ida?
09      (.)
10 Bel: Uh yah.hh=
11 Fan: Wud she do die:?=
12 Bel: =Mm:hm,

Notice that Belle, the bearer of the bad news, does not actually tell it herself: it is Fanny who eventually announces the news (in question format): 'Wud she do die' (line 11). When Belle says 'I had something terrible t'tell you' (line 1) she indicates to Fanny that she has bad news, but she doesn't produce it. Belle's protracted delay (look at the long silence on line 7) in producing the 'terrible' news she has projected leads Fanny to derive - and articulate - the news herself: first, who it is about (line 8), and then what has happened to her (line 11). Belle simply has to confirm that Fanny has understood correctly (lines 10 and 12).

News of a medical diagnosis

Our second data fragment is taken from a medical interaction between a doctor at a developmental disabilities clinic and the mother of a child referred for assessment. The task facing the doctor is to deliver a (bad news) diagnosis. Here's how he does it:

01 Dr: What do you see? as- as his (0.5) difficulty.
02     (1.2)
03 Mo: Mainly his uhm: (1.2) the fact that he
04     doesn't understand everything. (0.6) and
05     also the fact that his speech (0.7) is very
06     hard to understand what he's saying (0.3)
07     lot[s of ti]me
08 Dr:    [ right ]
09     (0.2)
10 Dr: Do you have any ideas wh:y it is: are you:
11     d[o yo]u?h
12 Mo:  [ No ]
13     (2.1)
14 Dr: .h okay I (0.2) you know I think we basically
15     (.) in some ways agree with you: (0.6) .hh
16     insofar as we think that (0.3) Dan's main
17     problem (0.4) .h you know does: involve you
18     know language.
19     (0.4)
20 Mo: Mm hmm
21     (0.3)
22 Dr: you know both (0.2) you know his- (0.4) being
23     able to understand you know what is said to 
24     him (0.4) .h and also certainly also to be
25     able to express:: (1.3) you know his uh his
26     thoughts
27     (1.1)
28 Dr: .hh uh:m (0.6) .hhh in general his
29     development ...

Although it is the doctor who has information about the diagnosis, he delays telling it – just as we saw Belle and Charlie do – thereby creating an opportunity for the recipient to tell the bad news herself. He explicitly asks the mother for her understanding of the child's problem (line 1) and its underlying cause (lines 10-11). And although, in this case, the mother does not actually produce the bad news herself, the doctor acts as if she had. In saying (at lines 14-15) 'we basically ... agree with you', he treats her as if she had produced the bad news he is now simply confirming. This is an institutional adaptation of the ordinary conversational action of breaking bad news – and it illustrates how patterns we use in ordinary conversation typically underpin workplace interactions. CA has discovered many conversational patterns like this (e.g. strategies for bringing a topic to a close, normative ways in which people do disagreement as opposed to agreement, and practices for holding the floor at points where another person might start to speak), and it is the cumulative set of these and many other discoveries that constitutes the ‘tool set’ we bring to the analysis of new data.

Applications

Communication in medicine, counselling, and healthcare is perhaps the largest area of ‘applied’ CA (see Heritage and Maynard, 2006), with significant training possibilities for healthcare practitioners. For example, in a project commissioned by a helpline for women in trauma after childbirth, recorded extracts are now used when training midwives and other care-providers to give more effective support to women post-natally (Kitzinger & Kitzinger, 2007).

We selected recordings that – our analysis showed – offered good illustrations of the appropriate use of interactional features, such as story prompts, or silence (and also some that illustrate more problematic interactions). Compared with role-play – used in earlier workshops for the helpline – the recorded data make for a much more vivid and immediate understanding of interactional issues, and encourage self-reflective practice.

Along with medical and helpline interaction (Shaw and Kitzinger, 2007; Kitzinger & Kitzinger, 2007), key applied research areas within CA include counselling and psychotherapy (Peräkylä, 1995); classroom interaction (Koshik, 2001), news interviews (Clayman and Heritage, 2002) and calls to the emergency services (Zimmerman, 1992). Findings from all these domains are potentially of key interest both to practitioners and to policy makers. For example, the Department for Work and Pensions has recently commissioned a group of us at the University of York to investigate the interaction between personal advisers (PAs) and benefit claimants in Jobcentre Plus interviews (Drew, Toerien, Irvine and Sainsbury, ongoing research), to identify what constitutes effective practice.

Direct commercial applications of CA are still in their infancy, but are advancing most rapidly in California (the birthplace of CA) with the work of researchers at the Palo Alto Research Center (formerly Xerox PARC). They have used conversation analysis in designing an electronic guidebook, Sotto Voce, for visitors to historic homes, museums and other attractions (Woodruff et al, 2002). This group has also described practical strategies for integrating a conversation analyst into a research design team (Woodruff et al, 2004).

Opportunities and Challenges

CA is a valid and reliable method of qualitative research (see (Peräkylä, 1997), in that it sidesteps problems associated with the gap between what people say they do and what they actually do.

By capturing what people actually do, on video, we have an indisputable record of actual behaviour. The method of analysis enables the researcher to find patterns in these behaviours, and it is these patterns that constitute the key findings. Results can often be communicated effectively to clients for whom the recorded data offer the opportunity for reflective practice.

Using recordings to provide feedback to clients can, however, pose ethical concerns. This is particularly true in large organisations, where clients may include front-line service providers, their managers, and even high-level policy makers. Ensuring that appropriate measures of anonymity are in place (e.g. pixelating video data, deleting personal references) can be time-consuming and technically demanding; this can also make the recordings more difficult to watch. It is also common for organisations to expect feedback more quickly than is possible using CA.

CA is a specialised and resource-intensive methodology and should be used only when appropriate, and when the research question merits it. Questions about how people behave are appropriate to CA; questions about how they think or feel are not. It can be difficult for clients to appreciate the appropriate range of questions to which CA can provide answers. Sometimes CA may be combined appropriately with other methods in order to address different kinds of question. In the research on electronic guidebooks, for example, CA research was used to establish how people behaved when using guidebooks while looking around a historic attraction. Interviews, meanwhile, were used to establish their subjective impression of how much they enjoyed the experience.

Some primary CA studies have been supplemented, very effectively, with follow-up quantitative research. For example, conversation analysts noticed that patients seemed more likely to reveal an additional health concern when asked at the end of a GP consultation whether they had “something else” they wanted to address, than whether they were asked if there was “anything else”. Subsequently a randomised controlled trial was conducted and showed that these two formats were in fact statistically significantly different in terms of eliciting patients’ concerns.

Another research project offering a good example of a multi-method approach explored how doctors might resist pressure for antibiotic prescriptions in paediatric consultations. A statistically significant association was shown between what parents said and the likelihood of the doctor prescribing antibiotics inappropriately (i.e. inappropriate prescriptions were more likely if the parent questioned the doctor’s initial treatment plan for the child).

In this kind of applied CA research, the use of statistical measures is important because they allow us to demonstrate a significant relationship between an interactional practice and a relevant outcome (e.g. soliciting unmet concerns or appropriate prescribing). However, it is important to recognise that these statistics do not replace the fine-grained qualitative approach offered by CA but are based on it.

CA’s specific focus on interaction is, then, not simply a limitation, but also a unique strength. After all, in so many instances, it is talk that gets the job done. Taking a medical history, cross-examining a witness, lecturing students, counselling a new mother, advising Jobcentre Plus claimants – these are all tasks that are done through talk. In using CA’s empirical findings to understand what is happening in talk-in-interaction we eliminate the gap between what people say they do and what they actually do, and gain access to the heart of human action.

References and Further Reading

Atkinson, J.M., (1984), Public speaking and audience responses: Some techniques for inviting applause. in J.M. Atkinson and J. Heritage, Structures of Social Action: Studies in Conversation Analysis. Cambridge: Cambridge University Press. pp. 370-410.

Clayman, S., and Heritage, J., (2002), The News Interview. Cambridge: Cambridge University Press.

Drew, P., (1984), Speakers' reportings in invitation sequences. In J.M. Atkinson and J. Heritage (eds), Structures of Social Action: Studies in Conversation Analysis (pp. 152-164). Cambridge: Cambridge University Press.

Heritage, J., (1984), Garfinkel and Ethnomethodology. Cambridge: Cambridge University Press.

Heritage, J., and Maynard, D.W., (eds) (2006), Communication in Medical Care. Cambridge: Cambridge University Press.

Heritage, J., Robinson, J.D., Elliott, M.N. Beckett, M. and Wilkes, M., (2006), Reducing patients’ unmet concerns in primary care: The difference one word can make, Journal of General Internal Medicine 22(10): 1429-33.

Jefferson, G., (2004), Glossary of transcript symbols with an introduction. In G.H. Lerner (Ed.), Conversation Analysis: Studies from the First Generation (pp. 13-31). Amsterdam: John Benjamins.

Kitzinger, Celia (2004), Feminist approaches. In Clive Seale, Giampietro Gobo, Jaber F. Gubrium, & David Silverman (Eds.), Qualitative research practice (pp. 125-140). London: Sage.

Kitzinger, C., and Kitzinger, S., (2007), Birth trauma: Talking with women and the value of conversation analysis, British Journal of Midwifery 15(5): 256-264.

Koshik, I., (2001), Designedly incomplete utterance: A pedagogical practice for eliciting knowledge displays in error correction sequences. Research on Language and Social Interaction, 35(3): 277-309.

Mangione-Smith, R., Elliott, M.N., Stivers, T., McDonald, L.L., & Heritage, J., (2006), Ruling out the need for antibiotics Are we sending the right message? Archives of Pediatrics and Adolescent Medicine, 160 945-952.

Maynard, D.W., (1997), The news delivery sequence: Bad news and good news in conversational interaction. Research on Language and Social Interaction, 30: 93-130.

Peräkylä, A., (1995), AIDS Counselling: Institution Interaction and Clinical Practice. Cambridge: Cambridge University Press.

Peräkylä, A., (1997), Reliability and validity in research based on transcripts, pp. 201-220 in Silverman, D. (ed.) Qualitative Research: Theory, Method and Practice. London: Sage.

Peräkylä, A., (2004), Conversation analysis. In C. Seale et al (Eds), Qualitative Research Practice (pp. 165-179). London: Sage.

Schegloff, E.A., (1988), On an actual virtual servo-mechanism for guessing bad news: A single case conjecture. Social Problems, 35: 442-457.

Schegloff, Emanuel, A., (1997), Whose text? Whose context? Discourse & Society 8(2), 165-187.

Schegloff, E.A., (2007), Sequence Organisation in Interaction: A Primer in Conversation Analysis. Cambridge: Cambridge University Press.

Shaw, R., and Kitzinger, C., (2007), Memory in interaction: An analysis of repeat calls to a Home Birth helpline. Research on Language and Social Interaction 40(1): 117-144.

Stivers, T., (2005), Non-antibiotic treatment recommendations: Delivery formats and implications for parent resistance, Social Science and Medicine 60: 949-964.

Toerien, M., and Kitzinger, C., (2007), Emotional labour in action: Navigating multiple involvements in the beauty salon. Sociology 41(4): 645-662.

Waitzkin, H., (1985), Information giving in medical care, Journal of Health and Social Behavior 26: 81-101

Wilkinson, S., and Kitzinger, C., (2008), Conversation analysis, in Willig, C., and Stainton Rogers, W., The Sage Handbook of Qualitative Research in Psychology. London: Sage. pp. 55-71.

Woodruff, A., Aoki, P., Grinter, R., Hurst, A., Szymanski, M., & Thornton, J., (2002), Eavesdropping on electronic guidebooks, pp. 21-30 in D. Bearman and J. Trant (Eds) Museums and the Web 2002: Selected Papers. Pittsburg PA: Archives & Museum Informatics. Online at: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.18.4535

Woodruff, A., Szymanski, M., Grinter, R. & Aoki, P., (2002), Practical strategies for integrating a conversation analyst into an iterative design process, Proc.DIS, ACM: 19-28.. Online at: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.19.3682

Zimmerman, D.H., (1992), The interactional organization of calls for emergency assistance. In P. Drew and J. Heritage (eds), Talk at Work. Cambridge: Cambridge University Press.

Foundational texts

Sacks, H., Schegloff, E.A., & Jefferson, G., (1974), "A Simplest Systematics for the Organization of Turn-Taking for Conversation." Language 50: 696-735.

Sacks, H., (1992) Lectures on Conversation. Oxford: Blackwells..

Schegloff, E.A., (1968), Sequencing in conversational openings, American Anthropologist, 70:1075-95

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

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

Introductions to Conversation Analysis

Drew, P., (2005), Conversation analysis, pp. 71-102 in K.L. Fitch and R. Saunders (eds.) Handbook of Language and Social Interaction. Mahwah, NJ: Lawrence Erlbaum.

Liddicoat, A.J., (2007), An Introduction of Conversation Analysis. London: Continuum. (accessible introduction to a broad range of key concepts in CA).

Schegloff, E.A., (2007), Sequence Organisation in Interaction: A Primer in Conversation Analysis. Cambridge: Cambridge University Press. (A clear and authoritative introduction, by one of the founders of the field, to one of the key concepts of CA; other primers will follow on turn-taking, repair, and word selection).

Learning more about CA

Workshops and short courses on conversation analysis are regularly taught as part of the Continuing Professional Development portfolio at the University of York. For information contact Celia Kitzinger cck1@york.ac.uk