Discourse analysis: what is it and why is it relevant to family practice?

This paper aims to illustrate what discourse analysis is and how it can contribute to our understanding of family practice. Firstly, we describe what 'discourse analysis' is, mapping the discourse analysis terrain by discussing four studies relevant to primary care to illustrate different methodological approaches and key concepts. We then address the practicalities of how to actually do discourse analysis, providing readers with a worked example using one particular approach. Thirdly, we touch on some common debates about discursive research. We conclude by advocating that researchers and practitioners take up the challenge of understanding, utilizing and extending the field of discourse studies within family practice.

Introduction

Discourse analysis is gradually becoming more established in family practice. Using rigorous methods and techniques, discourse analysis can offer a sophisticated insight into the complex world of family practice. But what do we actually mean when we talk about ‘discourse analysis’ and how is it done? To answer this question, the first section of this paper focuses on some basic theoretical ideas and concepts, drawing on studies relevant to primary care to provide readers with an understanding of the features of discourse analysis and the different approaches available. In the second part of this paper, we explore how to do discourse analysis (using data from a recent study in family practice) and consider common debates. We conclude by showing how discursive studies might add a ‘new’ methodological dimension to family practice research.

What is discourse analysis?

Discourse analysis is the study of social life, understood through analysis of language in its widest sense (including face-to-face talk, non-verbal interaction, images, symbols and documents).1 It offers ways of investigating meaning, whether in conversation or in culture.2 Discourse analytic studies encompass a broad range of theories, topics and analytic approaches for explaining language in use. They ask ‘What is social life like?’ and ‘What are the implications for individuals and/or wider society?’

Approaches to discourse analysis

To help explain what discourse analysis is we now describe four discourse studies relevant to family practice which range from micro-level study of face-to-face talk through to macro-level study of institutions in society (see Table 1). We have chosen these four studies as they allow us to demonstrate some of the diversity within discourse analysis; however, there is overlap between studies in terms of underlying theories and approaches.

T

Examples of discourse analytic research relevant to family practice

Study	Research question	Methods	Key findings	Study implications
Study 1: Misunderstandings in consultations, Roberts et al.3	How do misunderstandings occur in consultations between patients with limited English and GPs?	Sociolinguistic discourse analysis of 232 video-recorded consultations from inner city general practices. These were viewed independently by two discourse analysts with 37 selected and transcribed for detailed analysis. These reflected a wide range of English language ability and were analysed to explore why and how misunderstandings occur.	Patients’ different communication styles (including varied pronunciation, word stress, presentation styles, use of grammar and/or intonation) contribute to misunderstandings in GP consultations.	Culturally different styles of communication can result in misunderstandings. Practitioner training is needed to identify these problems, and prevent and repair misunderstandings.
Study 2: Coping with cancer, Wilkinson and Kitzinger4	How do women with breast cancer talk about positive thinking in their lives?	Discursive psychological analysis of focus group and interview discussions with women with breast cancer. Discussions explore women's feelings at first diagnosis, coping and support systems and the effects of their cancer on their lives and relationships. Data were audiotaped and transcribed, with analysis focused on occurrence of the words ‘positive’ or ‘positively’, paying detailed attention to the meaning and function of statements in the contexts in which they occur.	‘Thinking positive’ is a socially sanctioned way of thinking and talking about coping with cancer. Talk about thinking positive also serves a variety of interactional functions, e.g. moving conversation on after awkward topics (e.g. following talk about illness or death) and bonding the group together.	Statements need to be understood in context. Therapy and counselling services which focus on strategies for thinking positive are contributing to a moral obligation to ‘think positive’ in the face of cancer.
Study 3: Health professionals’ views of ECT, Stevens and Harper5	How do health professionals talk about ECT?	Drawing on discursive psychology and Foucauldian traditions to analyse transcriptions of eight semi-structured interviews with health professionals involved in administering ECT. Attention was given to the way that professionals’ talk about ECT is persuasive and justifies particular courses of action.	Health professionals’ accounts describe ECT recipients as ‘severely ill’, with urgent physical psychiatric treatments as the most obvious intervention (in spite of conflicting clinical evidence). This description acts to rationalize ECT and restrict choices about other possible interventions (e.g. psychotherapy).	Training packages could help health professionals become aware of the ways in which the language they use can restrict treatment choices for patients. Empowerment strategies might help patients to be more involved in decisions about their mental health care.
Study 4: The development of primary care research policy, Shaw and Greenhalgh6	What are the discourses that have dominated primary care research policy and which have been suppressed or excluded?	A Foucauldian approach recognizing the historical, social and ideological origins of policy and the role of power. Data included 29 key policy documents, 16 narrative interviews with policy stakeholders and additional historical documents. Analysis involved detailed deconstruction and linking across texts to reveal prevailing storylines, ideologies, power relations and tensions.	Recent policy has been shaped by discourse associated with the ‘knowledge-based economy’. This has meant that health research has become more important to the UK economy with microscopic ‘discovery’ and technological approaches being privileged for commercial advantage. As a result, primary care research has been reoriented as a ‘population laboratory’ for clinical trials research.	Health research policy is not value-free and serves particular interests. There are implications for the kind of research that is valued and funded, as well as who undertakes it and how. Researchers can use findings to make active decisions about how to position their work.

Approaches to discourse analysis are not easy to pin down. Different studies focus on different types of data (including spoken and written) and different types of discourse: for instance, Study 1 in Table 1 explores culturally specific styles of communication in general practice consultations3; Study 2 explores the ways that topics can be meaningfully talked about among women with breast cancer4 and Studies 3 and 4 uncover taken-for-granted ideas and ideologies in society in interviews with health professionals and in research policy documents.5,6 Discourse analytic approaches are influenced by a wide range of disciplines including anthropology, linguistics, cultural studies, gender studies, social psychology and philosophy.1 To gain a more in-depth appreciation of discourse analysis, we encourage readers to access the papers in Table 1.

Micro-level studies—such as the sociolinguistic discourse analysis in Study 1—involve the detailed study of language in use. They tend to be concerned with the techniques and competencies involved in successful and unsuccessful conversation, allowing researchers to build up a model of social life from an empirical understanding of actual linguistic events.7 Micro-level approaches owe a great deal to conversation analysis that provides a conceptual framework for systematically analysing face-to-face talk.8,9 The starting point for micro-level studies is the participant's perspective, allowing researchers to appreciate the cultural and communicative patterns which inform his or her behaviour and perceptions.10 Analysis explores how interactions are organized moment by moment through subtle yet taken-for-granted processes. For instance, Study 1 looks at patterns of misunderstandings in consultations between doctors and patients with limited English, identifying misunderstandings resulting from culturally specific styles of communication (such as how personal or impersonal to be, how direct to be in self-presentation or how literally to interpret a question).

Meso-level studies—such as the discursive psychology approaches in Studies 2 and 3—may also look at face-to-face talk. However, there is less emphasis on micro-level interaction and more on the connections with broader social and cultural contexts. The starting point for such studies is that discourse guides certain ways of talking about a topic, defining ‘acceptable’ ways to talk, write or conduct oneself and that this can serve a range of social functions.11 For example, Study 2 looks at how women with breast cancer talk about how they cope with their illness. Analysis reveals how ‘thinking positive’ is a powerful discourse, governing socially acceptable ways of thinking and talking about coping with cancer and placing a moral obligation on women with breast cancer to conduct themselves in particular ways. Talk about thinking positive serves a variety of social functions: it is used as a device to move conversation on in awkward moments (e.g. following talk about illness or death) and also acts to bind the group together through establishing a shared identity as breast cancer sufferers.

Discursive studies may take a critical perspective, for instance, exploring how different groups achieve and maintain their status through their control of conversational encounters and ‘systems of knowledge’. For example, Study 3 looks at how health professionals talk about electro-convulsive therapy (ECT). There are variations in the diagnostic labelling of people with mental distress and variations in decisions to treat with ECT. Analysis of health professionals’ talk about psychiatric treatments draws attention to the ways in which it creates a boundary between ‘severely mentally ill’ and ‘not severely mentally ill’. Health professionals’ discourse about the severely mentally ill acts to rationalize particular courses of action (i.e. ECT) and restricts choices about other possible interventions (e.g. social support or psychotherapy).

Macro-level approaches—such as the Foucauldian approach adopted by Study 4—tend to involve the study of language and ideology in society.2 The starting point is a concern with the role of power and knowledge in society, identifying patterns of language, demonstrating how they constitute aspects of society and establishing how and why the language available to us sets limits on what it is (and is not) possible to think, say and do.12 Analytic approaches can ‘deconstruct’ or unravel taken-for-granted assumptions, understand what these assumptions might mean for individuals and wider society and explore possible alternatives to accepted ways of doing things.13,14 For instance, Study 4 looks at the discourses that have dominated research policy and how these have shaped primary care research. Analysis reveals how the UK economy has been influenced by the drive towards a ‘knowledge-based economy’, emphasizing the production and use of information as a means of generating national wealth. For primary care, this has meant that knowledge which has commercial value (such as genetic discovery) has been privileged over knowledge that has other value (such as understanding patients’ perspectives).

What do these approaches to discourse analysis have in common?

Some discourse studies tend to draw on more than one approach (for instance, Study 3 draws on discursive psychology as well as a Foucauldian approach). Despite the diversity of origin and definition, discursive approaches share several conceptions about social life.

Firstly, ‘language and interaction are best understood in context’. Insightful interpretation of data involves understanding contexts such as local circumstances (e.g. setting, participants) and/or wider discourses that shape language and interaction. For instance, Study 4 explored documents from 1972 to 2005, helping to illustrate that research discoveries do not simply ‘happen’, but are products of history and the shift towards a knowledge-based economy.

Secondly, ‘social reality is socially constructed’. This is a concept which is difficult to grasp because it challenges a traditional, rationalist view of an objectively discoverable social world, instead acknowledging that social worlds are subjectively understood and experienced. Constructivists argue that all knowledge—including taken-for-granted, common sense knowledge—is derived from and maintained by social interactions.15 For instance, a category like severely mentally ill which may appear to be natural and obvious is actually an artefact of a particular culture or society (see above).

Thirdly, discursive research ‘looks beyond the literal meanings of language’. For example, in Study 2, cancer sufferers may or may not ‘really’ ‘think positive’: discourse analysis is not interested in whether beliefs and attitudes are ‘true’, but is interested instead in the social functions of talk (for example, the way that talking about thinking positive bonds members of a group or moves discussions on from difficult topics). Meaning therefore depends upon the context of an interaction, and in Study 3 thinking positive has many different meanings.

One approach to doing discourse analysis: a worked example from a doctor–patient consultation

There is no set formula for how to do discourse analysis. However, as with other qualitative approaches, there are a number of practical steps that can guide researchers (see Box 1). Background reading is an essential part of the research process, helping to refine research questions and understand how theoretical ideas and approaches might be relevant to the research.19 The focus of the research then guides the kind of data that will be gathered. Data may be drawn from a number of sources in order to preserve a sense of the contexts in which things occurred. Data may be ‘researcher generated’ (such as interviews or field notes) or ‘naturally occurring’ (such as published documents or recordings of conversations).

Practical steps in discourse analysis

Start with a general problem area.

Undertake background reading about discourse analysis and about the topic you want to study (both within and outside of the health/medical field).

Seek advice and/or support from a social scientist with experience of discourse analysis (if this is not your area of expertise).

Begin to focus your research questions, continuing to review and refine it/them throughout.

Decide on the type of data you wish to study and collect data.

Familiarize yourself with the data through repeated reading, watching and/or listening, asking questions of the data (such as ‘What is the context for this interaction?’, ‘What is happening and why?’) and begin to note interesting features.

Transcribe any spoken data you might have collected, paying close attention to detail.

Index for analytic themes and discursive features. Look for patterns.

Make analytic notes as you go along, using paper and pen and/or a computer-assisted qualitative data analysis software package (such as NVIVO or Atlas.ti).

Test intuitive hunches against the data, being critical and looking for counter examples.

Discuss your emerging analysis with colleagues, especially those from other disciplines.

Start writing preliminary analyses, moving between writing, reading and analysing.

Continue redrafting analyses, being prepared to return to earlier steps as needed.

Adapted from16,17 and 18

Discourse analytic studies often start with a general problem area, developing more focused research questions as the research progresses so that researchers can remain genuinely open to new insights. To demonstrate this process, we draw on a study (undertaken by JB) exploring consultations for coughs and colds in family practice: we describe the formulation of the research question and present a worked example of analysis which exemplifies a discursive psychology approach (similar to Study 2). As we describe above, this is only one of many approaches to thinking about and analysing discourse.

Existing literature suggests that consulting with minor illness may be seen as inappropriate by health professionals.20 To explore potential misunderstandings and conflicts in consultations for coughs and colds, 33 consultations between doctors and patients with upper respiratory tract infections were videorecorded, with doctors and patients interviewed afterwards. In the process of becoming familiar with the data, it became apparent that patients gave surprisingly long and involved accounts of apparently ‘minor’ symptoms, which raised the question: ‘what purpose do these accounts serve?’ The literature suggests that patients need to persuade doctors that their visit is appropriate,7 particularly if the problem might be labelled by the doctor as minor. Analysis therefore paid particular attention to the use of persuasive language (i.e. rhetorical structure).21 To demonstrate how this might be done, we now analyse part of one patient's account of their symptoms (see Box 2).

We could approach the data in Box 2 from a number of different perspectives. Reading the passage as a ‘medical history’, the data can be summarized in just one line (six days of sore throat, itchy respiratory passages, three days of fever), with most of the rest of the account deemed largely irrelevant. However, there are dimensions of social interaction going on apart from ‘giving a medical history’.

Data extract

The following excerpt is taken from near the beginning of a consultation between a GP and a 30-year-old man (Mr K) with respiratory symptoms:

13 right em (..) I started last Wednesday (…) with a little

14 (.) it wasn't sore throat [repeated stroking hand motions over throat and face]

15 (.) it wasn't- (..) very dry em throat and all my em nasal

16 (…) channels or whatever you call it in my ears

17 (.) very itchy all around (..) [wavy finger motion near ears]

18 em that was on Wednesday evening

19 .hhh on Thursday evening (.) seemed like em you know everything was fine

20 (..) nothing at all (.) but then on Friday- (.)

21 er (.) I made an appointment on Thursday morning just in case and they gave me

22 (.) today (.) of course

23 .hh em on Friday (..) it came back (..) unexpectedly em (..) massively (.) like em

24 (.) my sore (..) my throat was really really sore and my ears and etcetera (.)

25 .hh I left work em (.) in the evening (.) em I went home (.) I had a fever (.)

26 and that night was horrendous and Saturday was exactly the same (.)

27 I (.) could barely swallow

This transcript includes detail such as pauses, and some body conduct, but not the minute detail of how things were said and body conduct etc.17

(.) represents pauses of a tenth of a second.

.hhh represents an in-breath.

[italicized] represents notes about body conduct.

Discourse analysis involves looking beyond the literal meaning of language, understanding the context in which social interaction takes place and exploring what was said, when and why. Mr K gives specific details of what happened and when: the illness seemed to start innocuously on Wednesday (six days before his appointment) with a dry throat and itchy ‘nasal channels’ (lines 13–18). Then, on Thursday Mr K's symptoms seemed better (lines 19–20). In lines 23–25, he explains that the symptoms recurred ‘massively’ and ‘unexpectedly’ while he was at work. On the face of it, the detail about initial mild symptoms seem fairly irrelevant. However, discourse analysis allows exploration of the function that it serves: it suggests that Mr K had not overreacted to his symptoms, had not prematurely adopted a sick role and was behaving in a socially responsible way.

Mr K creates a persuasive account of his illness in a number of ways.21 For instance, he uses a three part list (a rhetorical device that captures attention), referring to (i) ‘really really sore throat’, (ii) ‘my ears’ and (iii) ‘etcetera’ (line 24). He also provides specific detail about events on different days of the week that makes his account more believable. He uses several extreme case formulations which are designed to be dramatic and persuasive—for example, ‘it came back massively’, ‘really really sore throat’, ‘horrendous night’ and ‘barely swallow’—which construct his illness as more severe than a mundane sore throat. The dramatic contrast between the initial mild symptoms and later severe symptoms is also persuasive in design. Mr K accompanies his account with hand gestures: repeated stroking motions over his throat and face while describing sore throat symptoms (line 14) and finger motion to convey itchiness (line 17) which help to demonstrate to the doctor an illness which is not visible.

In summary, Mr K does considerable rhetorical work which constructs his illness as worthy of attention and himself as not to blame, pre-empting any suggestion that consulting with minor respiratory symptoms is not appropriate. Awareness of this dimension of the social interaction (i.e. patients’ ‘hidden agendas’ about legitimacy) could help doctors and patients to avoid unintended loss of face and/or conflict.

Some debates about discursive research

Is discourse analysis just subjective opinion?

A common concern about discourse analysis is that study findings represent nothing more than researchers’ opinions. For instance, in the worked example above, how can we know for sure that Mr K has an unstated agenda about the legitimacy of his visit? This interpretation is not arbitrary but is justified by reference to the data and supported by ‘evidence’ from other sources (for example, literature about ‘inappropriate’ use of primary care, the rhetorical structure of other consultations and doctors’ and patients’ concerns in interviews). Discourse analysts see research findings as socially constructed, for example, products of historical, geographical, economic and other contexts, and influenced by the researchers themselves22 (e.g. disciplinary background, age, gender, ethnicity and so on). Discursive ‘findings’ are therefore seen as rigorously produced interpretations rather than ‘discoveries’.23 Providing detail about study settings, participants and methodologies allows readers to judge credibility and plausibility of findings.19 As is the case for other qualitative approaches, discursive findings are judged for the insights they can offer and are theoretically rather than statistically applicable to other situations.19

Can discursive approaches complement other methods?

Whether discursive methods can truly complement other approaches depends upon the methodological assumptions which underpin the research.24 Different approaches may have competing assumptions about the nature of data, what position the researcher holds in relation to the research participants, how data can be analysed, what conclusions can be drawn, how certain knowledge can be and how findings can be applied. Researchers from different traditions may be able to compromise to accommodate different ways of viewing the world. For example, in a study investigating ethnic minority students’ performance in exams, initial quantitative work describes the proportion of ethnic minority students failing final medical exams and subsequent discursive work then explores how and why this happens.25

In contrast, there are some research approaches which are incompatible with the conceptions underpinning discursive research. For example, a popular tool in health-related research is the attitude survey: the underlying assumptions of such surveys are problematic to the discourse analyst because concepts such as ‘satisfaction’ do not have a fixed, universal meaning and experiences are complex. Surveys fail to capture the context in which things are said: although the same questions are asked of all respondents, they will be interpreted in unique ways by different people.1

How is discourse analysis relevant to family practice?

Discourse analysis focuses on interaction, looking beyond the literal meaning of language. It lends itself to studying the complexities of day-to-day family practice, helping to unpick taken-for-granted (and often revered) ideas and practices. Discourse analysis adds a new methodological dimension to family practice research by drawing on theories and approaches from a range of disciplines, typically from outside medicine. Like other qualitative approaches, discourse analysis therefore brings a different lens through which we can potentially add to and deepen our understanding. Findings often have practical implications for family practice: for example, Study 1 identified how and why misunderstandings occur in patient–practitioner communication and resulted in a training video for health professionals and Study 3 informed the development of empowerment strategies to help patients to be more involved in decisions about their mental health care.

Our paper has explained the what, how and why of discourse analysis: we advocate that those allied to family practice take up the challenge of understanding, utilizing and extending the field of discourse studies within family practice.

Declaration

Funding: This work was supported by an interdisciplinary postdoctoral award from the Economic and Social Research and Medical Research Councils to SS and a Primary Care Researcher Development Award from the Department of Health to JB.

Ethical approval: None.

Conflicts of interest: None.