Ethnography and Discourse Analysis

ETHNOGRAPHY AND DISCOURSE ANALYSIS: INCOMPATIBLE OR COMPLEMENTARY?*

Martyn Hammersley** (OPEN UNIVERSITY)

RESUMO: Este artigo examina a relação entre etnografia e análise de discurso. Começa delineando a variação considerável nos tipos de trabalho que se abrigam sob um desses termos. Similitudes e tensões entre essas abordagens são assim investigadas. O que tem sido descrito como a ‘recente crítica radical de entrevistas’ é usado como um modo de explorar as orientações colidentes de alguns tipos de etnografia e análise de discurso. Esta avaliação crítica está fundamentada em argumentos que podem ser encontrados dentro da etnografia, mas também têm sido motivados por algumas formas de análise de discurso. Na parte final do texto, sugere-se, que há boas razões para qualificar os argumentos construídos na crítica radical e que o valor da prática analítica discursiva é mais bem compreendida dentro de uma visão etnográfica menos restritiva.

PALAVRAS-CHAVE: discurso, etnografia, entrevistas.

^[1]

ABSTRACT: This paper examines the relationship between ethnography and discourse analysis. It begins by outlining the considerable variation in kinds of work that come under each of those headings. Similarities and tensions between these approaches are then investigated. What has been described as the ‘recent radical critique of interviews’ is used as one way of exploring the conflicting orientations of some kinds of ethnography and discourse analysis. This critique is founded on arguments that can be found within ethnography, but that have also motivated some forms of discourse analysis. It is suggested in the final part of the paper that there are good reasons to qualify the arguments built into the radical critique, and that the value of discourse analytic practice is best understood from within a broadly ethnographic orientation.
KEYWORDS: Discourse analysis, ethnography, interviews.

My focus in this paper is on two closely related questions: can ethnography and discourse analysis be combined; and should they be combined - in other words, would this be beneficial? Some commentators give positive answers to these questions, and there have often been attempts to bring the two approaches together in some way. Many years ago, having been taught conversation analysis as well as traditional forms of ethnography, I used them both in my PhD research looking at social interaction in secondary schools (Hammersley 1980; see also 1974, 1976 and 1977). I employed the work of Harvey Sacks (see Sacks 1992), largely implicitly, to understand the roles of teacher and pupils in the classroom. However, my aim was to draw conclusions of a rather conventional sociological kind about schooling and its functioning in society. So, while in one sense I was combining ethnography with conversation analysis, most conversation analysts would deny that I had done this. Instead, they would probably see me as having abused conversation analysis for the purposes of ethnography. I had extracted conversation analysis from its ethnomethodological context and put some elements of it to work within a different framework. What this makes clear is that much hinges on what one means by combining ethnography and discourse analysis: combined in what relationship?

A second problem arises from the fact that ‘discourse analysis’ and ‘ethnography’ are both umbrella terms: in other words, each covers a

variety of kinds of work. Therefore, any answer to the questions I have posed requires some discrimination amongst versions of discourse analysis and of ethnography. I will begin, therefore, by sketching some of the heterogeneity on each side.

Diversity within discourse analysis

In the case of discourse analysis, it is possible to identify three broad types of approach, in rather crude terms. First, there are those which are inspired primarily by post-structuralism, and especially by the work of Foucault. ${ }^{1}$ For Foucault, and for many post-Marxists, discourses are sociohistorical forms of practice that are constitutive of the subjects and objects that make up the world defined within particular truth regimes. Foucault himself looked at the discourses surrounding madness, medicine and various forms of social and psychological science, treating these as constitutive of modernity. In a more conventional historiographical application of this approach, three discourses have been identified that, in a conflictual configuration, were the framework within which eighteenth century French revolutionaries construed the ancien regime and sought to construct the new society that they wanted to bring about (Baker 1990). Other writers have looked at discourses operating in more recent times, for example those surrounding social inclusion (Levitas 1998) and sexuality (Hollway 1989).

The other two broad kinds of discourse analysis I will mention involve closer study of specific texts, or of transcripts of verbal social interaction. One approach is based on linguistics, of one sort or another. Here the same kinds of distinction that have been applied to language below the level of the sentence are deployed above that level: to whole paragraphs and texts in the case of written materials, and to turn-taking and storytelling in the case of oral talk. Moreover, it is often argued that the whole structure of language needs to be taken into account in interpreting any piece of discourse. ${ }^{2}$ Some influential versions of linguistics-based discourse

^[1]

analysis are also committed to locating patterns of discourse within a wider social context, for example identifying how these function to reproduce or challenge existing social structures. This is true of the critical linguistics of Fowler, Kress and Hodge and of the critical discourse analysis of Fairclough (Fowler 1991; Kress and Hodge 1979; Fairclough 1995 and 2001).

The third category of discourse analysis is even more of a mixed bag than the others. It includes social psychological and sociological approaches, ranging from the work of van Dijk through that of Potter and Wetherell all the way to conversation analysis of the kind championed by Schegloff and others (van Dijk 1991; Potter and Wetherell 1989; Schegloff 1997). And there has been considerable mutual influence among these. Indeed, the version championed by Potter and Wetherell, perhaps the most widely influential of all, was explicitly a combination of several extant approaches. And the subsequent work of these two authors has come to diverge in significant ways: Wetherell now labels her work critical discursive psychology, while Potter has moved closer to conversation analysis. Even within conversation analysis there have been major disputes, notably over whether institutional talk is different in character from more informal kinds of talk (Boden and Zimmerman 1991; Drew and Heritage 1992), and over the possibility of feminist and ‘critical’ versions (Speer 1999; Kitzinger 2000; Bogen and Lynch 1990; McHoul 1994). Indeed, what has happened over time is that further hybrid varieties have emerged, combining one or more of the existing strands as well as elements from beyond. Thus both Fairclough’s critical discourse analysis and Wetherell’s critical discursive psychology are influenced by post-structuralist ideas, yet there still remain significant differences between these two projects (Fairclough 1992; Chouliaraki and Fairclough 1999; Wetherell 1998).

Heterogeneity within ethnography

If we turn to ethnography, it is much more difficult to give a clear outline of the variety in approach to be found under this heading. The term has been in use for a much longer period, and has been shaped by

^[1]

disciplinary differences and rivalries as well as by many shifts in theoretical orientation and empirical focus. A range of philosophical, theoretical, and methodological ideas have shaped ethnographic work over the course of its history. These include positivism, functionalism, Marxism, pragmatism, phenomenology, hermeneutics, structuralism, post-structuralism, and constructionism. ${ }^{3}$ Only an extremely crude characterisation of the effects of these influences can be provided here. Positivism and functionalism encouraged ethnographers to document actual patterns of social behaviour, rather than just artefacts and beliefs, and to explain these in terms of the needs that any individual or society must meet. By contrast, Marxism, pragmatism, phenomenology, and hermeneutics led to greater emphasis on differences in orientation not only between societies but also within them, and to greater concern with documenting the cultural interpretations that inform patterns of action. Structuralism, poststructuralism, and constructionism have encouraged a focus on language and discourse that has led to some important departures from previous practice. In some cases, linguistics has been taken as a model for how to understand all aspects of human social life, treating these as generated by the deep structure of human nature. More commonly today, ethnographers come to pay closer attention to the language people use in the course of what they are doing, and in describing their lives in interviews, with a view to understanding how in some sense they ongoingly and discursively construct the social phenomena that characterise the society to which they belong. And from this it is a short step, though one with large implications, to a more radical kind of constructionism that views ethnographers themselves as effectively constituting social reality through their writing (see, for example, Clifford and Marcus 1986). These various influences have generated divisions within ethnography: between work that adopts a micro-focus in which the role of local context is emphasised and work which emphasises understanding the local in terms of the societal or global context; between approaches that insist on the need for an analytic perspective and those that demand a ‘critical’ orientation; and between those which seek to document hidden realities and those that are concerned with the discursive construction of ‘reality’.

In terms of research practice, we can also map significant patterns of change and diversification from early forms of anthropological fieldwork through to some more recent types of research that are labelled

ethnographic within the social sciences. For most anthropologists, from the early twentieth century onwards, ethnography involved actually living in the communities of the people being studied, round the clock, with the fieldwork taking place over a relatively long period of time, usually years. By contrast, today, the term ‘ethnography’ is standardly used to refer simply to studies that involve some participant observation and/or indepth, relatively unstructured interviews. For example, sociological ethnographers do not usually live with the people they study, they typically focus on what happens in a particular situation or institution when it is in operation, so that their ethnography is part-time in relation to the lives of the people studied. In fact, this is now increasingly true even of Western anthropologists where they study ‘at home’ in their own society or in other large complex societies. To a considerable degree, this reflects the nature of those societies: that people do not all live and work together in the same places in the way that they did previously in more traditional communities. Equally important, however, the fieldwork carried out by ethnographers is likely to last months rather than years. This reflects, perhaps, the increasing pace of academic life, notably the demand that publications be produced rapidly. However, it has also been stimulated by the turn to more micro-focused analysis, encouraged by the availability of lightweight audio- and video-recorders, which allow much more detailed recording of data than did the previous reliance solely on fieldnotes.

Some of the shifts in the practice of ethnography are thrown into relief by the growth, in recent years, of so-called internet or virtual ethnography (see Hine 2000, Mann and Stewart 2000, and Paccagnella 1997). Here, all the data are usually collected on-line without meeting the people concerned face-to-face. This raises the question: does ethnography depend upon the physical presence of the ethnographer in the midst of the people being studied? Or does the assumption that an ethnographer must be physically present involve an outdated idea of what is required for ethnographic work? Perhaps it even implies a false notion of personhood in a postmodern world? Equally important, the situations being studied in internet ethnography are virtual ones, rather than being particular physical locations. Given this, should we talk about on-line cultures? Or can we only understand what happens on-line in the context of the ordinary, that is off-line, lives of the people who produce blogs, put messages on message boards, participate in chat rooms, set up their own

web-sites, and so on? Alternatively, should we recognise that even the cultures that ordinary ethnography studies are also ‘virtual’, in a certain sense? After all, they are not objects that we can see or touch. And perhaps the very distinction between on- and off-line is misleading? Mark Poster, amongst others, has argued that postmodernity de-centres and disperses identities, and blurs the boundaries between humans and machines (Poster 1990). One of the effects of these, and other, developments is bring some forms of ethnography quite close, in practice, to forms of discourse analysis.

I hope I have been able to give here at least some sense of how ideas about ethnography have changed over time, generating substantial variation in how it is pursued and in what it produces.

Compatibilities and incompatibilities

For a variety of reasons, there are considerable overlaps in orientation and practice between the various kinds of work that come under the headings of ethnography and discourse analysis. In recent times, they have mostly been resolutely qualitative rather than quantitative, with a focus on meaning-construction and a reliance on unstructured observational and/or interview data. Moreover, some of the incompatibilities that exist are between particular forms of each type of inquiry, rather than between the two approaches overall. Nevertheless, there are some important differences and tensions that need to be examined. In what follows I will focus primarily on the third kind of discourse analysis outlined above, though I will begin by briefly examining the compatibility of the other two with ethnography.

If we take the first sort of discourse analysis, influenced primarily by post-structuralism, in some respects this involves much the same challenge to ethnography as structuralism did in the past. For some anthropologists, Levi-Strauss’s structuralism was compatible with ethnography, whereas others saw it as in conflict with what they held to be most important in ethnography. ${ }^{4}$ It is clearly possible to use ethnographic data to document

^[1]

the character and operation of various discourses in much the same way that Foucault and others have drawn on documentary evidence. There is, at the same time, conflict between this type of post-structuralist work and elements of the theoretical orientations that have previously shaped ethnography, some of which have emphasised the role of the perspectives and actions of individual actors in generating social phenomena. Those who have sought to combine ethnography with this sort of discourse analysis have recognised ‘conflict and discontinuity’ but emphasised ‘convergence’ (see Tambouku and Ball 2003:ix); whereas other ethnographers are more sceptical about the possibility or value of this. Whether we find an unacceptable conflict here will depend upon what aspects of previously influential theoretical and methodological ideas we build into our definition of ethnography, and also on how much emphasis we give to the need for an internally consistent orientation.

Turning to the second broad category of discourse analysis, that which draws on linguistics, there is often a significant difference in focus from ethnography; one that reflects commitment to different disciplines. Where linguistic discourse analysts are often concerned to identify the rules that generate particular discursive patterns, ethnographers are usually more concerned with what the discourse is being used to do, with the social factors that drive its use as part of patterns of situational and institutional action, and with the social consequences of that use. Another difference is methodological in character: discourse analysts usually rely entirely on textual or audio-recorded data, whereas ethnographers will often combine these sources with a much wider range of data. Of course, these differences in orientation have already been negotiated, to some degree, within anthropological linguistics and socio-linguistics (see, for example, Gumperz and Hymes 1972; Hymes 1977 and 1996; Gumperz 1982). Moreover, with critical linguistics and critical discourse analysis, there is much less of a difference in orientation, at least in terms of focus. ${ }^{5}$

^[1]

In the case of the third category of discourse analysis, there are also both compatibilities and tensions. I can illustrate this by discussing a dispute that has arisen over ethnographers’ use of interview data, which has drawn on arguments that are central to the rationale for discourse analysis.

Traditionally, interview data have been used by ethnographers in at least two ways. First, they have been employed to provide information about events and settings that were not open to observation by the ethnographer him- or herself: because they happened before the fieldwork began or because for some reason the ethnographer was not able to gain access to them. Equally important, informants’ accounts have also been used as sources of additional information about events or settings that ethnographers have been able to observe, either to cover parts of them that were not accessible or to offer a different perspective on what was seen. In other words, interviewees have frequently been used as surrogate researchers to provide second-hand information about the world: to give researchers access to what informants have themselves witnessed or to information that they have acquired from others.

The second ethnographic use of informants’ accounts is as a way of gaining understanding of the perspectives, attitudes, beliefs, or values of the people being studied. These are often not accessible via observation, and ethnographers have insisted that without understanding them it is sometimes impossible even to describe accurately what people are doing, even less to explain why they do what they do. So, here, interview data are used to identify stable attitudes or perspectives on the part of informants; in other words, attitudes or perspectives that are taken to govern their behaviour in other contexts. An example would be interviewing employees within an organisation about their attitudes towards the management, these being assumed to guide much of their everyday behaviour at work.

A recent ‘radical critique’ of interviews has rejected both of these standard uses of interview data (Murphy et al 1998:120-3; for examples of this critique, see Dingwall 1997; Silverman 1997; Atkinson and Coffey 2002). Moreover, it points in two rather different directions as regards the practice of research. One possible conclusion is that interviews should be abandoned, on the grounds that the only reliable form of data comes from observation of naturally occurring social interaction in which the researcher is not a key participant. A quite different response, more widely

adopted, is to argue that while interviews cannot be used in the usual ways, they can be employed to generate data about the discursive resources or practices by which people construct social phenomena. These two rather different recommendations match the attitudes of conversation analysis, on the one hand, and the sort of discourse analysis originally advocated by Potter and Wetherell, on the other. Either way, the radical critique seems to require, in effect, the reconstruction of ethnography into a form that is much more like discourse analysis.

In order to understand what is at stake here, we need to give attention to the arguments underlying the radical critique of interviewing. As we shall see, elements of all these arguments can be found within the ethnographic tradition, but in the context of this critique they take on more extreme forms. So, in this respect, like many others, there is both continuity and difference between ethnography and discourse analysis.

The first argument is what has sometimes been referred to as naturalism (see Hammersley and Atkinson 1995:ch1). This amounts to a desire to understand social phenomena as they occur naturally in the world (rather than under artificial conditions), so that, as far as possible, the data employed must be uninfluenced by the research process and the researcher. This was the basis on which many ethnographers rejected the use of experimental approaches to understanding human behaviour, and it also led them to adopt roles in collecting data that are designed to minimise reactivity. While this naturalism rarely resulted in their rejecting interviews as a source of data, it did encourage both an emphasis on the centrality of participant observation and the adoption of relatively unstructured approaches to interviewing, designed to simulate ordinary conversation. By contrast with this, some versions of the radical critique do use naturalism as a basis for rejecting the standard uses of interview data - on the grounds that people’s responses in interviews are so heavily shaped by the context, and especially by the influence of the interviewer, that reliable inferences about their knowledge of, or behaviour in, other situations are impossible. Where, previously, reactivity in interviews was seen as an important problem to be addressed, it is treated by the radical critics as irremediable.

The second argument behind the radical critique contrasts the results of scientific observation by the researcher with the everyday accounts produced in interviews by informants. Within ethnography, there had

always been recognition that informants’ accounts are not normally based on rigorous data collection and recording, and therefore are likely to be selective, inferential, and subject to distortion by memory. As a result, they were seen as requiring careful checking. Once again, the radical critique pushes this argument further. Drawing on what we might call phenomenological empiricism, it insists that for analysis to be rigorous the data employed must be, in some sense, immediately given rather than fallible. In the context of phenomenological philosophy the data would be phenomena given to consciousness, once all background assumptions have been bracketed out. In the context of discourse analysis the given data are audio- or video-recordings, and transcriptions of these. In short, then, reliance on informants’ accounts must be abandoned because, since they are not the product of rigorous scientific analysis, there is no way of discounting their methodological weaknesses so as to extract valid knowledge. Equally important, an increase in the rigour of observation by the researcher is demanded, this is now to be limited to what is evidenced in the behaviour of the people being studied, as recorded or transcribed.

The third argument behind the radical critique might usefully be labelled discursive psychology (see Edwards 1997). This is a general philosophical argument to the effect that mind is displayed in public behaviour rather than being a matter of internal, private experience. Interestingly, this is analogous in some ways to George Herbert Mead’s social behaviourism, on which symbolic interactionist ethnography was based (Blumer 1969). Mead argued that we must understand human psychology in terms of a dynamic interplay between an active force that is rather fluid in character, the ’ I ‘, and various definitions of self, actual and imagined, the ‘me’, that are introjected by the actor in seeking to make sense of others’ behaviour and anticipate their reactions to his or her future actions. Despite the similarities with discursive psychology, Mead’s social psychology was not interpreted by ethnographers as denying either the existence or importance of insider or relatively private knowledge, and they have generally been concerned with gaining access to such knowledge, often by means of interviews. By contrast, from the point of view of discursive psychology, accounts of ‘subjective’ phenomena - of beliefs, attitudes, past experience, etc - are not to be treated as more or less accurate representations of cognitive activity or of private knowledge or experience

but analysed instead as public, discursive displays through which actions, contexts, and even subjectivities are ongoingly constituted. In these terms, the use of interviews to gain insider accounts is ruled out. Instead, we must analyse the narrative forms or rhetorical strategies people employ when they talk about their experiences.

The final line of argument that can be identified within the radical critique is scepticism about the idea that accounts can ever represent reality at all, whether this is ‘external’ or ‘internal’ reality. There has long been an element of scepticism built into some influential kinds of ethnography, in the sense that the perspectives of the people studied have often been viewed as constructed on the basis of distinctive cultural assumptions, and/or as designed to serve various functions, rather than as simply reflecting the nature of the world (Hammersley 1992:ch3). And there has sometimes been a tendency for recognition of cultural pluralism to turn into epistemological relativism: in other words, the idea that ‘the world’ is constructed in different ways by different cultures has resulted in the conclusion that there are multiple realities rather than a single one. The radical critics sometimes draw on such relativism, and effectively push it into scepticism, suggesting that informants’ accounts must be treated not as true or false but rather as constitutive - as themselves producing one of many possible versions of events. From this point of view, reality is constructed in and through the telling, rather than having characteristics that are independent of this. Thus, the radical critics argue that we should not judge informants’ accounts in terms of whether or not they are accurate. Rather, we should study how people construct accounts, what discursive resources they employ, and perhaps why they portray things in the ways that they do. Equally important, this argument involves an agnosticism, at best, about whether there are such things as feelings, attitudes, values, etc that motivate people’s behaviour. There is a distrust of any appeal to relatively stable, underlying factors that generate appearances, in favour of an emphasis on the complexity and flux of what is on the surface of social life.

These arguments are not entirely compatible with one another, at least in their fully-extended forms. For example, scepticism is in conflict with both phenomenological empiricism and discursive psychology, since the latter involve claims to knowledge. Perhaps as a result of this, in the context of discourse analysis we find varying emphases on each of these

arguments, leading to rather different forms of research. As an illustration, we can contrast two forms. The first accepts naturalism, insists on a severe, empiricist threshold for the acceptance of evidence, adopts discursive psychology, but subordinates scepticism to these other arguments. This is the position taken by many conversation analysts. By contrast, some other discourse analysts reject naturalism in favour the idea that the researcher cannot but be part of the social world being investigated, and therefore must both influence it and be influenced by it. As a result, they also do not place much emphasis on empiricism. However, they do adopt discursive psychology, and draw on scepticism. In fact, neither of these positions is entirely stable, and most discourse analysts probably draw on all four arguments at one time or another.

It is worth emphasising that these arguments at the core of discourse analysis have radical implications not only for interviews but also for the analysis of observational data, indeed for the question of what does and does not count as such data. In relation to ethnography, they largely rule out the use of fieldnotes, since these cannot simply be a record of what happened: human observation and note-taking are necessarily selective and inferential. In writing fieldnotes we formulate in particular ways who was involved and what they did, and there are always alternative possible and valid formulations. So, for discourse analysts, instead of the observer being the research instrument, reliance is to be placed on mechanical means of audio- or video-recording. ${ }^{6}$ These are taken to produce records rather than interpretations or inferential accounts. And, for conversation analysts, inference is to be limited in the process of analysis too: to what is ‘evidenced’, so that the only context to be employed is that which is displayed as relevant within the data.

Now, it seems to me that none of these arguments is entirely convincing, either as a rationale for rejecting the normal uses of interview data or as a basis for discourse analysis. While the first, naturalism, points to an important threat to validity, seeking to minimise reactivity is not the only way to deal with it; equally important is to try to understand its effects, to adopt a reflexive stance (Hammersley and Atkinson 1995).

^[1]

The second argument, empiricism, highlights the need for research to draw on evidence about what people are doing that is beyond reasonable doubt in its validity. However, there is little justification for extreme forms of empiricism that claim to narrow down what can count as evidence to empirical givens that are beyond all doubt (see Hammersley 1992 and 2003). The attempt to do this is futile: mechanical recording devices are always selective, and in order to use what they produce in analysis further selection and also interpretation and inference are required. Thus, a transcription is not simply an audio-recording written down, it is a cultural product designed to represent certain aspects of what was caught on tape, for particular purposes. The point is even more obvious with transcriptions of video-recordings, where some means has to be found of representing non-verbal behaviour. The third argument, discursive psychology, also contains an important element of truth: that people’s accounts of their experiences are formulated in language that is by its nature a collective resource, and will be shaped by notions of who can say what, how, when, and to whom. However, this does not require us to deny that people have unique personal experiences, or that they have distinctive sources of information that may not be immediately accessible to others or publicly reported. Finally, while scepticism can be useful for methodological purposes, as an existential commitment it undercuts any kind of research, and indeed any kind of knowledge or belief. In methodological terms, by suspending interest in whether informants’ accounts are true we can open up a new area of investigation: the rhetorical strategies or narrative structures to be found in such accounts. In these terms, discourse analysis would not be in competition with ethnography, but simply a different enterprise with a different focus; and, indeed, one that might generate resources that could be of great use to ethnographers. By contrast, adopting scepticism as an epistemological commitment involves rejecting not just ethnography but discourse analysis too, in favour of producing self-subversive texts that continually undermine any knowledge claims they make (See, for example, Ashmore 1989).

So, while all four of the arguments underpinning discourse analysis can be found within ethnography, as rudimentary elements, any attempt to apply them in unqualified form rules out most of the sorts of data that ethnographers have traditionally employed, many of the uses they have made of those data, and almost all of the topics they have normally

addressed. Doing this demands a fundamental re-specification of the research enterprise, one which at best effectively reduces it entirely to a form of discourse analysis. The costs of this re-specification are judged by many ethnographers to be excessive and unnecessary. However, others are drawn to discourse analysis, believing that it avoids some of the methodological problems with which ethnography has always had difficulty in dealing, and provides an account of human social life that is more in keeping with its fundamental character. For the reasons I have indicated, in my view this is an illusion.

Conclusion

Returning to the questions I posed at the beginning of this paper, it is clear that there are ways in which ethnography and discourse analysis, of various kinds, can be combined. However, this usually involves abandoning some of the key ideas and practices on one or both sides in order to bring about compatibility. Forms of discourse analysis can be extracted from their usual methodological and theoretical infrastructures so as to make them compatible with ethnography. And ethnography can be re-engineered to make it compatible with, or to turn it into, discourse analysis. I have more sympathy with the first of these strategies than with the second: I have tried to show that the methodological arguments that underpin the radical critique of interviews, which motivate advocacy of discourse analysis as a competitor to ethnography, have long been part of the ethnographic tradition in moderate forms. And there are good reasons not to adopt more extreme interpretations of them. Indeed, doing so in a consistent way would make it difficult even for discourse analysts to sustain the claim that they can produce knowledge of the social world.

So, while discourse analysis is legitimate as an enterprise adopting a different focus from ethnography, being concerned with discursive patterns, rhetorical strategies, or narrative structures, it cannot supplant ethnography and other kinds of social research. At the same time, the ideas and methods that have been developed by discourse analysts can be valuable resources for ethnographers, enabling them to engage in more detailed analysis of talk and texts than was common in the past; and, of course, many ethnographers have already availed themselves of this.

It is perhaps worth emphasising, however, that discourse analysis also serves the function of highlighting some problems within ethnography, arising from the clash between the various methodological arguments on which it has relied. In fact, facing up to discourse analysis is, for an ethnographer, rather like looking into a distorting mirror that tells us more about ourselves than we may wish to know. The difficult, unresolved problems that I have identified - about the nature of agency, the relationship between representational and functional roles of concepts, how far we can rely on insider accounts, the problem of reactivity, and so on - have long been present within ethnography. Furthermore, discourse analysis also underlines problems ethnographers frequently have these days in talking about facts or about truth, or even about knowledge; we tend to put scare quotes around these terms, to indicate that we do not want to be committed to anything so naïve as realism. The appeal of discourse analysis for some ethnographers, I suspect, is that it seems to promise a way of carrying on doing research without that commitment. Yet, it cannot deliver on that promise. Instead, we must recognise that we cannot be researchers without, in effect, being realists, though not necessarily naïve ones.

There may be a great deal to be gained from combining ethnography with discourse analysis. However, it should be clear from the arguments presented in this paper that this is not a straightforward matter. The relationship between the two highlights some quite difficult issues about the nature of the social world and how we can understand it. Putting the two together and glossing over the conflicts and tensions will not improve matters. However, engaging carefully with the range of methodological and philosophical arguments to be found under these two headings may lead to a more secure basis for improving social research in general.

References

ASHMORE, M. The Reflexive Thesis. Chicago: University of Chicago Press, 1989.

ATKINSON, P. & COFFEY, A. Revisiting the relationship between participant observation and interviewing. In J. F. Gubrium and J. A. Holstein (Eds.) Handbook of Interview Research. Thousand Oaks, CA: Sage, 2002.
ATKINSON, P. & HAMMERSLEY, M. Ethnography and Participant Observation. In N. K. Denzin and Y. S. Lincoln (Eds.) Handbook of Qualitative Research. Thousand Oaks, CA: Sage. 1995.

ATKINSON, P. A. & SILVERMAN, D. Kundera’s Immortality: the interview society and the invention of the self. Qualitative Inquiry, 3, 1997: 304-25.
BAKER, C. The search for adultness: membership work in adolescent-adult talk. Human Studies, 7, 1984: 301-23.
BAKER, K. Inventing the French Revolution. Cambridge: Cambridge University Press, 1990.
BLUMER, H. Symbolic Interactionism: perspective and method. Englewood Cliffs, NJ: Prentice-Hall, 1969.
BODEN, D. & ZIMMERMAN, D. H. (Eds.) Talk and Social Structure. Cambridge: Polity Press, 1991.
BOGEN, D. & LYNCH, M. Social critique and the logic of description: a response to McHoul. Journal of Pragmatics, 14, 3, 1990: 505-21.
BURAWOY, M. BLUM, J. A., GEORGE, S., GILLE, Z., GOWAN, T., HANEY, L., KLAWITER, M., LOPEZ, S. H., RIAIN, S., THAYER, M. Global Ethnography: Forces, Connections, and Imaginations in a Postmodern World. Berkeley: University of California Press, 2000.
CHOULIARAKI, L. & FAIRCLOUGH, N. Discourse in Late Modernity: rethinking critical discourse analysis. Edinburgh: Edinburgh University Press, 1999.

CLIFFORD, J. & MARCUS, G. (Eds.) Writing Culture: The poetics and politics of ethnography. Berkeley, CA: University of California Press, 1986.
DINGWALL, R. Accounts, interviews and observations. In Miller, G. and Dingwall, R. (Eds.) Context and Method in Qualitative Research, London: Sage, 1997.

DREW, P. & HERITAGE., J. (Eds) Talk at Work. Cambridge: Cambridge University Press, 1992.
EDWARDS, D. Discourse and Cognition. London: Sage, 1997.
FAIRCLOUGH, N. Discourse and Social Change. Cambridge: Polity Press, 1992.
FAIRCLOUGH, N. Critical Discourse Analysis. London: Longman, 1995.
FAIRCLOUGH, N. Language and Power. Second edition. Harlow: Pearson Education, 2000.
FOWLER, R. Language in the News: discourse and ideology in the press. London: Routledge, 1991.
GUMPERZ, J. Discourse Strategies. Cambridge: Cambridge University Press, 1982.
GUMPERZ, J. & HYMES, D. (Eds.) Directions in Sociolinguistics. Oxford: Blackwell, 1972.

GUTTING, G. Michel Foucault’s Archaeology of Scientific Reason. Cambridge: Cambridge University Press, 1989.
GUTTING, G. Foucault and the history of madness. In G. Gutting (Ed.) The Cambridge Companion to Foucault. Cambridge: Cambridge University Press, 1994.

HAMMERSLEY, M. The Organisation of Pupil Participation. Sociological Review, 22, 3, 1974: 355-368.
HAMMERSLEY, M. The Mobilisation of Pupil Attention. In M. Hammersley and P. Woods (Eds.) The Process of Schooling. London: Routledge and Kegan Paul, 1976.

HAMMERSLEY, M. School Learning: the cultural resources required to answer a teacher’s question. In P. Woods and M. Hammersley (Eds.) School Experience. London: Croom Helm, 1977.
HAMMERSLEY, M. A Peculiar World? Teaching and Learning in an Inner City School. Unpublished Ph.D. thesis, University of Manchester, 1980.
HAMMERSLEY, M. What’s Wrong with Ethnography? London: Routledge, 1992.
HAMMERSLEY, M. Sociology, what’s it for? A critique of the grand conception. Sociological Research Online, 4, 3, 1999.
http://www.socresonline.org.uk/socresonline/4/3/hammersley.html
HAMMERSLEY, M. Conversation analysis and discourse analysis: methods or paradigms? Discourse and Society, 14, 6, 2003: 751-81.
HAMMERSLEY, M. & ATKINSON, P. Ethnography: principles in practice. Second edition, London: Routledge, 1995.
HAMMERSLEY, M. & GOMM, R. Recent radical criticism of the interview in qualitative inquiry. In M. Holborn and M. Haralambos (eds.) Developments in Sociology, Volume 20, Ormskirk, Causeway Press/Edinburgh: Pearson Education, 2005.

HAYES, E. N. & HAYES, T. (Eds.) Claude Lévi-Strauss: the anthropologist as hero. Cambridge, MS: MIT Press, 1970.
HINE, C. Virtual Ethnography. London: Sage, 2000.
Hollway, W. Subjectivity and Method in Psychology. London: Sage, 1989.
HOWARTH, D. Discourse. Buckingham: Open University Press, 2000.
HYMES, D. Foundations in Sociolinguistics. London: Tavistock, 1977.
HYMES, D. Ethnography, Linguistics, Narrative Inequality. London: Taylor and Francis, 1996.
KITZINGER, C. Doing feminist conversation analysis. Feminism and Psychology,

10, 2, 2000: 163-93.
KRESS, G. & HODGE, R. Language as Ideology. London: Routledge and Kegan Paul, 1979.
LEVINSON, S. Pragmatics. Cambridge: Cambridge University Press, 1983.
LEVITAS, R. The Inclusive Society? Social exclusion and New Labour. London: Macmillan, 1998.
MCHOUL, A. Towards a critical ethnomethodology. Theory, Culture and Society, 11, 1994: 105-26.
MACLURE, M. Discourse in Educational and Social Research. Buckingham: Open University Press, 2003.
MANN, C. & STEWART, F. Internet Communication and Qualitative Research: A Handbook for Researching Online. London: Sage, 2000.
MERQUIOR, J. G. Foucault. London: Fontana, 1985.
MURPHY, E., DINGWALL, R., GREATBATCH, D., PARKER, S., and WATSON, P. Qualitative research methods in health technology assessment: a review of the literature. Health Technology Assessment, 2, 16, 1998: 1-260.
<http://www.hta.nhsweb.nhs.uk/execsumm/summ216.htm>(Accessed on 14.08.02)

PACCAGNELLA, L. Getting the seats of your pants dirty: strategies for ethnographic research on virtual communities. Journal of Computer Mediated Communication, 3, 1, 1997.
http://www.ascusc.org/jcmc/vol3/issue1/paccagnella.html (Accessed on 12.01.05)

POSTER, M. The Mode of Information. Cambridge: Polity Press, 1990.
POTTER, J. and WETHERELL, M. Discourse and Social Psychology: Beyond attitudes and behaviour. London: Sage, 1987.
SACKS, H. Lectures on Conversation. Ed. G. Jefferson. 2 vols. Oxford: Blackwell 1992.

SCHEGLOFF, E. A. Notes on a conversational practice: formulating place. In D. Sudnow (ed.) Studies in Social Interaction. New York: Free Press, 1971.
SCHEGLOFF, E.A. (1997) Whose text? Whose context?. Discourse and Society 8: 165-87.
SCHIFFRIN, D., TANNEN, D. and HAMILTON, H. H. (eds.) The Handbook of Discourse Analysis. Malden, MS: Blackwell, 2001.
SEALE, C. Qualitative interviewing, in C. Seale (ed.) Researching Society and Culture. London: Sage, 1998.

SILVERMAN, D. Towards an aesthetics of research. In D. Silverman (ed.) Qualitative Research: theory, method and practice. London: Sage, 1997.
SPEER, S. (1999) Feminism and conversation analysis: an oxymoron?. Feminism and Psychology. 9, 4: 471-8.
STRENSKI, I. Malinowski: Second Positivism, Second Romanticism. Man, 17, 1982: 766-77.
STUBBS, M. Discourse analysis: the sociolinguistic analysis of natural language. Oxford: Blackwell, 1983.
STUBBS, M. Scratching the surface: linguistic data in educational research. In M. Stubbs Educational Linguistics, Oxford, Blackwell, 1986.

TAMBOUKOU, M. & BALL, S. J. (eds.) Dangerous encounters: genealogy and ethnography. New York: Peter Lang, 2003.
WETHERELL, M. (1998) Positioning and interpretative repertoires: conversation analysis and post-structuralism in dialogue. Discourse and Society, 9: 387-412.
VAN DIJK, T. Racism and the Press. London: Routledge, 1991.