Volume 6, No. 1, Art. 30 – January 2005
Reanalyzing Qualitative Interviews From Different Angles: The Risk of Decontextualization and Other Problems of Sharing Qualitative Data
Harry van den Berg
Abstract: In contrast to survey interviews, qualitative interviews are seldom reanalyzed. Besides obvious reasons such as ownership—and especially the culture of individualistic ownership—that impede reusing data, there is also methodological skepticism about secondary analysis. In this paper, I will argue in favor of sharing qualitative data on behalf of secondary analysis. The argument is partly based on—and much inspired by—the discussions during the preparation of a collaborative project of thirteen researchers who were invited to analyze the same set of interview data from their own theoretical/methodological viewpoint (VAN DEN BERG, WETHERELL & HOUTKOOP-STEENSTRA, 2003). During these discussions several methodological arguments against secondary analysis were put forward. In this paper I will deal with some of these arguments, especially the doubts about the usefulness of secondary analysis and the argument concerning the assumed risk of decontextualization: Is secondary analysis possible without in-depth knowledge of the context? Different theoretical and methodological positions concerning the contextualization of interview discourse will be scrutinized. On the one hand I argue against the tendency to include the ever-widening societal and historical context on behalf of the analysis of interview discourse. This tendency runs the risk of speculative social theorizing as a framework for interpreting interview discourse. On the other hand I do not think that the neglect of every social context outside interview talk—as advocated by some strands within conversation analysis—is fruitful or even possible. This neglect runs the risk of abstract empiricism. The main argument is that the kind and measure of contextualization of interview data needed on behalf of discourse analysis should depend on the research goal and the type of data.
Keywords: secondary analysis, qualitative interviews, discourse analysis, conversation analysis, contextualization, methodology
Table of Contents
2. Reanalyzing Qualitative Interviews From Different Angles: A Collaborative Project
3. Secondary Analysis of Qualitative Data: Is It Really Useful?
4. Secondary Analysis of Qualitative Data: Is It Really Feasible?
4.1 Context as extra-discursive template
4.2 Context as intra-discursive product
4.3 Context as conditions of discursive production
5. Contextual Information on Behalf of Secondary Analysis of Qualitative Interviews: Some Guidelines
Up till now, secondary analysis is still considered as something belonging to the world of quantitative research. In qualitative research, secondary analysis is rare. More than that, it is not widely recognized as a useful research practice notwithstanding the efforts of those who are involved in organizing qualitative data archiving and stimulating the development of a research culture favorable to secondary analysis (CORTI, 2002). 
In this paper I will discuss the experiences with a collaborative project of researchers who were invited to reanalyze the same set of interview data from their own theoretical/methodological viewpoints. In some respects, discussions during the preparation and the realization of the project were at least as informative as the ultimate outcome of the project itself. These discussions offered in-depth insight in existing obstacles, objections and aversions against secondary analysis of qualitative data in general. In this paper I will focus on some of the main methodological issues that were raised during these discussions. The two issues seem especially relevant because they are related to more general methodological problems of qualitative research. The first issue concerns the desirability of secondary analysis of qualitative data: Why should reanalyzing existing data be recommended as something useful? What is exactly the surplus value of secondary analysis above the analysis of data produced in the context of your own research? The second issue concerns the feasibility of secondary analysis of qualitative data: Is it really possible to analyze data without in-depth knowledge of the research context as well as the social context within which these data were produced? Before dealing with those issues, it is necessary to give some background information on the project. 
One of my research interests concerns qualitative interviews and especially the methodology of qualitative interviews. In several respects, the methodology of qualitative interviewing is an underdeveloped field. First, it is widely recognized that the interviewer plays a decisive role in producing qualitative data with sufficient depth. Nevertheless, little is known about interviewer behavior in qualitative interviews. Methodological handbooks on qualitative research contain a variety of normative guidelines how to behave, but neither the functioning of these guidelines nor the actual behavior of interviewers is systematically studied. Second, the methodology of analyzing qualitative interviews (and qualitative data in general) shows important blind spots. Especially if one is interested in structural features such as composition, sequential order and different levels of meaning-structures, analytic instruments procedures and guidelines on behalf of a systematic analysis are missing.1) 
Fortunately, there are new opportunities to tackle both problems. Especially the developments in the field of conversation analysis and discourse analysis may contribute to our understanding of interviewer behavior, interaction in interview discourse and more generally the mechanisms by which social realities, opinions, and attitudes in interview discourse are produced. Discourse analysis covers a variety of different approaches that developed since the end of the eighties. The common denominator of these different approaches is the renewed attention for language as a form of social behavior. This linguistic turn is highly relevant for the methodology of analyzing textual data such as interview transcripts. Conversation analysis has a much older history going back to the sixties. Traditionally, conversation analysis was exclusively oriented to the analysis of informal talk but since the end of the eighties, there is a growing interest in the conversation analysis of institutionalized forms of talk such as research interviews. Therefore, in order to tackle the black spots in the methodology of qualitative interviewing one could profit from both developments in conversation analysis and discourse analysis. 
In order to make use of these opportunities, I developed a plan to invite representatives of different research strands of discourse analysis including conversation analysis on behalf of a workshop on the question how to analyze qualitative interviews. Such a workshop could possibly offer an overview of what those strands may contribute to our understanding of interview discourse. In order to facilitate the comparison of those contributions it seems worthwhile to ask participants to analyze the same set of interview data. Together with my colleague, Hanneke HOUTKOOP-STEENSTRA, I prepared the workshop. The first obstacle we met, concerned the selection of the interview data. As data archivists may know, qualitative researchers are not very eager to share their data with other researchers on behalf of reanalyzing those data. Besides, it was not feasible to use for example my own data because those interviews are in Dutch language. Due to the hegemonic position of English as the common language in the international scientific scene, we had to restrict our focus to qualitative interviews held in Anglo-Saxon countries. Fortunately, Margaret WETHERELL was generous and agreed to share interviews that were part of a large-scale research project she conducted in the mid-1980s on racism and race relations in New Zealand.2) 
We selected three interviews that were re-transcribed in fine detail to make them suitable for as diverse a range of styles of analysis as possible. The choice of interviews on race relations for the collaborative exercise was not accidental. Open-ended interviews on sensitive and controversial topics such as prejudice, ethnocentrism, ethnic categorizations, and stereotyping are difficult to interpret. These interviews very often produce many ambiguous statements. Traditional qualitative research on the fields of ethnicity, racism and nationalism, and gender and sexism has encountered severe difficulties in coping with the ambiguities and contradictions within interview discourse on these topics. Therefore, it was challenging to invite discourse analysts from different strands to reanalyze these interviews from their perspectives. 
It is difficult to summarize the results of the project in a few lines. Anyway, the experience of collaborative reanalyzing the same set of qualitative interviews and discussing each other's contribution proved to be quite stimulating. Of course, such a secondary analysis offers the opportunity to validate and refine the original analysis of the data as presented in the book of Margaret WETHERELL and Jonathan POTTER (1992). 
But, more important, this collaboration showed what discourse analysis may contribute to our insight in the research interview as a social activity. Besides it demonstrates different styles of discourse analysis. Therefore we decided to publish a selection of the contributions (VAN DEN BERG, WETHERELL & HOUTKOOP-STEENSTRA, 2003). 
An important feature of the book is that the full transcripts of the selected interviews are enclosed. As a consequence, the reader has the opportunity to check the interpretations developed by the authors and to facilitate the comparison of different approaches of discourse analysis. It is possible, for example, for the reader to reanalyze the fragments selected by the authors within the context of the transcribed interview as a whole and to compare the selected fragments with other parts of the interview. So in this respect, the book may contribute to the development of a spirit of secondary analysis. 
However, as mentioned in the introduction, during the preparation and the realization of the project, we were also confronted with serious criticisms concerning the aims of the workshop and especially the idea of reanalyzing the same set of interviews as a mean to realize these aims. It is worthwhile to scrutinize those critical arguments because they constitute an important barrier against secondary analysis of qualitative data in general. 
The first critical question is: Why is a collaborative reanalysis of the same data set useful? More specifically, why should using the same data facilitate the comparison of different approaches of discourse analysis? Different approaches are characterized by different methodological/theoretical assumptions and procedures. The selection of data is not something completely unconnected to these assumptions and procedures. On the contrary, the selection of data and the way data are constructed as analyzable data (for example in the form of transcripts of qualitative interviews) are deeply entwined with those assumptions and procedures. Even the way interviews are transcribed is based on theoretical assumptions. So, on behalf of a "fair" comparison of different strands of discourse analysis, each strand should present research in which the type of data is constructed that fits in with the assumptions and procedures of that strand. According to this view, secondary analysis of data is something that is only useful within the boundaries of a specific set of theoretical and methodological assumptions. But using the same data on behalf of the comparison of different approaches fuels the false empiricist notion of data as something "objective," something only to be assembled instead of something that has to be produced. 
At first sight, this critical argument concerning the desirability of secondary analysis of qualitative data seems rather plausible. Nevertheless, the argument is defective in several respects. 
First, the argument is based on the assumption that different approaches of discourse analysis (or qualitative research in general) can be viewed as different scientific paradigms that are incommensurable. This assumption is often stated but never founded. In fact it may be doubted that different approaches of discourse analysis can be treated as self-sufficient paradigms (see also: HAMMERSLEY, 2003). The overhasty and premature paradigmatization of theoretical and methodological differences has severe consequences for scientific discussion. The supposed incommensurability implies that the idea of comparison based on analyzing the same data is made senseless. More generally, a common ground for a rational debate between different approaches is denied. The choice between different approaches is reduced to a matter of subjective preferences. 
Second, the argument is based on a one-sided overestimation of the constructed nature of empirical data. Of course, the naïve concept of empirical data as something given, something "out there" only to be observed and assembled by the researcher belongs to the outdated positivist illusions of the former century. But the rejection of this naive concept does not imply that empirical data are just the derivatives of theoretical and methodological assumptions of the researcher. The "empirical" is not just a reflection or a dress up of the "theoretical." The "empirical" is undoubtedly connected to the theoretical, but it has a momentum of its own. For example, research interviews are not the instruments to "open up" respondents to get access to the already existing opinions and emotions (the illusion of old positivism and naturalism as well; see HAMMERSLEY & ATKINSON, 1995), nor are they completely determined by the interview strategies, tactics and techniques designed on behalf of some theoretical goal defined by the researcher. As several contributions in the publication (VAN DEN BERG, WETHERELL & HOUTKOOP-STEENSTRA, 2003) based on the workshop abundantly demonstrate, research interviews are essentially co-constructed in the interaction between the interviewer and the interviewee. Therefore, the outcome of an interview is always unpredictable because the results are partly dependent on the interviewee. That is why doing empirical research is exciting, challenging and sometimes surprising. 
Third, the argument implies an unnecessary and risky restriction of the kind of data to be analyzed by a specific approach of discourse analysis. It is assumed that each theoretical/methodological approach can only be judged according to how it analyses the data produced by research based on that approach. This assumption neglects the risk that in this manner data are selected/produced that fit in well with pre-existing theoretical expectations. To avoid this risk it seems far more fruitful to stick to the idea that the strength of a specific theoretical/methodological position should be related to the scope of discursive phenomena to which it can be applied. One should remember that this idea played an important role as one of the motives in the development of conversation analysis (see for example: SILVERMAN, 1998). Speech act theorists were heavily criticized for staying behind their desk and constructing invented sentences that fit in well within their research goals. The aim of conversation analysis was to study the variety of everyday conversations as they take place in society. This aim implies a broad scope of the discursive phenomena that should be covered by conversation analysis. 
To summarize, the objections against the desirability of secondary analysis are badly founded. They tend to restrict the possibilities of comparison and rational debate between different approaches by constructing imagined paradigmatic walls between those approaches. They overestimate the constructed nature of empirical data as derivates of specific theoretical aims and assumptions. They neglect the importance of showing the relevance of specific approaches by demonstrating what these approaches can produce in analyzing a broad variety of empirical data. 
More important than the objections against the desirability is the critical stance concerning the feasibility of secondary analysis: Is it possible to analyze data without in-depth knowledge of the research context as well as the social context within which these data were produced? 
This question is far more difficult to deal with than the first one, because analyzing social events within their social context is generally considered as one the hallmarks of qualitative research. And neglecting the risk of decontextualization is precisely one of the main objections of qualitative research against quantitative research based on survey-designs or experimental designs. Therefore one could ask if secondary analysis of qualitative data such as qualitative interviews is compatible with the requirement to bring in the context on behalf of the analysis of interview discourse. 
In the field of qualitative research, the prevailing answer tends to be negative. There is a strong skepticism about secondary analysis of interview discourse and qualitative data in general. 
For example, Jan BLOMMAERT (1997) discusses in a paper about workshops and data sharing why interview transcripts from his research on communication problems between immigrant women and Belgian welfare workers are not sharable (BULCAEN & BLOMMAERT, 1997). A colleague from another university asked permission to use those interview transcripts. But besides legal and ethical barriers for sharing these data, there was in his view "a more fundamental obstacle":
"We felt that our colleague could not understand the full depth of what was going on in the transcripts. The complex forms of interaction involving peculiar forms of troubles talk, administrative talk, psychosocial counseling, but also intimidation, distancing and so on: the interaction of personal-emotive, professional and bureaucratic voices; the clash of jargons and cultural schemata observable in many professional-client interactions; the wider context of cultural stereotyping, the structure of welfare work to immigrants, the intertextuality between various cases and professional discourses; and the transformation of a narrative into written notes, case reports, team meeting talk, summaries given to other professionals (doctors, police officers) and so on: all these elements and their intricate interplay had only gradually become clear to us, and their was no way in which simple things could be said about any of the data samples we had collected" (BLOMMAERT, 1997, p.32). 
The argument specifies several types of context ranging from cultural, discursive and linguistic resources used in the interaction between interviewer and interviewee and including aspects of the wider social context such as structure of the well-fare work and the wider cultural context such as cultural stereotyping. Besides specific historical background, several features are mentioned concerning intertexuality, transformations of text in different genres and last but not least the "intricate interplay between all those (contextual) elements." 
Such a variety of assumed relevant features evokes several questions: Which aspects of the conditions of discursive production should be taken into account as the relevant context? Are there criteria to determine what should count as relevant or is it up to the researcher to determine the boundaries of the context to be included in the interpretation of the text? And last but not least: How to avoid the endless regress of the argument about context? After all, to interpret the context of a text, you need to study the wider context, etcetera. 
The quoted argument against data-sharing is an example of the more general viewpoint that an intense personal involvement in the fieldwork constitutes a necessary prerequisite in order to grasp the relevant context and to interpret interview transcripts. From this viewpoint, secondary analysis of qualitative interviews amounts to doing the impossible. The researcher defines him or her self as the privileged insider and as a consequence any outsider is declared as unable to reanalyze the textual data. A further implication of this approach is that it is impossible to check the interpretation and analysis of these data. In other words, this argument on context may function as immunization against possible criticism. In this respect, the argumentation resembles the way context is used in everyday discourse as a rhetorical device to undermine undesired interpretations and to impose desired interpretations of someone's utterances. 
In my view this is not a fruitful approach of the notion of contextualization in qualitative research. The problem with the whole notion is that it all depends on what is precisely meant by contextualization. The general phrase "bringing in the context" is in fact very ambiguous. It means different things in different strands of qualitative research and there are different conceptions of which context is relevant and how this context should be used on behalf of the interpretation of interview discourse. Therefore it is necessary to unravel those different conceptions on the relation between textual data and the context to be accounted for in interpreting those data. 
These differences lie behind an ongoing debate about the relation between text and context (BILLIG, 1999a, 1999b, BLOMMAERT, 2001, SCHEGLOFF, 1992, 1997, 1998, 1999a, 1999b, VERSCHUEREN, 2001, WETHERELL, 1998, 2001, 2003). 
Roughly, three different conceptions of context can be distinguished:
The "broad" concept: Context as extra-discursive template,
the "narrow" concept: Context as intra-discursive product,
the "intermediate" concept: Context as conditions of discursive production. 
The relation between discourse and social structure constitutes the main object of an approach that is known as Critical Discourse Analysis (FAIRCLOUGH, 1995, 2003, VAN DIJK, 1987, 1991, 1997, WODAK, 2001). The general aim of CDA is to develop empirically based social criticism. Therefore, CDA focuses especially on structural relations of inequality such as racial and gender discrimination and the role of discourse in reproducing and transforming these relations. Theoretically, discourse is conceptualized as socially constitutive as well as constituted by structural features of society. However, the research practice of CDA is mainly dominated by a tendency to detect how structural relations "are manifested in language." As a consequence, CDA is often characterized by a specific reading practice: Text is interpreted as the imprint of structural inequalities. In other words, the context is viewed as an extra discursive template that should be taken into account in analyzing discourse because the meaning of discourse is molded by this template. 
This conception of context and the related reading practice is rather problematic. 
First, in CDA the context is often assumed instead of scrutinized (BLOMMAERT, 2001, SCHEGLOFF, 1997, VAN DEN BERG, 1992, VERSCHUEREN, 2001). As a consequence, there is a remarkable discrepancy between the detailed study of small amounts of text and the a-priori claims made by CDA researchers about the general social context of those textual data. 
Second, the way the assumed context and the text are related to each other is also predefined. These relations are often theoretically modeled in terms of a functionalist paradigm (VAN DEN BERG, 1992, 2003). Discourse is analyzed as functional for the reproduction of the social system. This speculative presupposition concerning the social functioning of discourse may easily blind the researcher for the meanings actually produced in discourse. This can be illustrated by the problems in analyzing contradictions in interview discourse on racial and ethnic issues. For example, in CDA it is often assumed a-priori that western societies are fundamentally racist in structural as well as cultural respect. From this starting point it is tempting to interpret contradictions in interview discourse on racial topics preferably in terms of disclaimers that interviewees use to prevent inferences that could be detrimental for their self-presentation. As a consequence, other possible interpretations of contradictions in interview discourse are excluded prematurely. For example, contradictions may be due to the fact that interviewees switch between incompatible interpretative repertoire3). 
To summarize, the research practice of CDA exemplifies the problem of contextualizing textual data on behalf of the interpretation of these data. The researcher invokes a theoretically predefined context. But in stead of a-priori assuming the relevant context—characteristic for theoretical essentialism—it seems methodologically more adequate if the researcher should try to find out which context is relevant for interpreting textual data. 
To avoid the risk of arbitrary choices by the researcher of what should be considered as relevant context, there is a well-known methodological alternative approach developed within conversation analysis. According to Emanuel SCHEGLOFF (1992, 1997) the only relevant context is the context that is demonstrably made relevant by the participants themselves in their talk. As a consequence, if one is interested in the role of context, one should turn to the textual data to find out which contextual elements are invoked by the discourse as constructed by the participants. In other words, context is a discursive product and should be studied as such. From this viewpoint there is no real methodological problem of context. The only problem is that some researchers tend to bring in their own prejudices and political preferences to predefine arbitrary contextual elements and features. According to SCHEGLOFF (1997, p.165) this amounts to "academic and theoretical imperialism, which imposes intellectual's preoccupations on a world without respect to their indigenous resonance." 
Notwithstanding the elegance of SCHEGLOFF's methodological rule to define empirically which elements should be considered as relevant context of textual data, it falls short in terms of its own ambitions. His concept of relevant context is too narrow. 
First, what is relevant for the participants does not need to be demonstrated in observable features of their talk. During a social interaction participants may assume a common knowledge base or a common ground that is not completely and explicitly articulated in talk. That common ground may include knowledge about relevant contextual features. 
Second, any stretch of talk that is defined by the analyst is "data" is not something isolated but is a moment in a history of discursive processes. For example, participants may selectively use specific cultural sources in their talk. But the observation that some sources are actually used and other sources are neglected does not entail cues about the selection process: selection may be due to the differential availability of sources or to a deliberate choice of specific available sources. 
Third, SCHEGLOFF's restriction of context to discursive cues produced in the discursive interaction (in his terms: proxy context) and his dismissal of any notion of socio-structural contextual phenomena as aimed by CDA (in his terms: distal context) assumes that talk and other forms of social behavior should be viewed as two separate worlds. But it is only an analytic distinction. In social life, the discursive and the extra-discursive cannot be separated so easily. 
The common problem of both concepts of context is that the concrete conditions under which the stretch of talk to be analyzed is produced, are neglected. These concrete conditions may contain relevant contextual features that the analyst should know of on behalf of the interpretation of the text. Participants may not always be aware of those features and as far as they consider those features as relevant, they need not to articulate those features as observable relevant in their talk. In these respects, this concept of context is broader than SCHEGLOFF's concept. 
Compared to the concept of context often used in CDA, this concept is less broad. It can be viewed as the interface between general characteristics of social structure on the one hand (such as different forms of structural inequality) and talk on the other hand (such as interview discourse and conversations). So context in this sense is much closer to the textual data to be analyzed then context as conceptualized in CDA. 
Conditions of discursive production entail both non-discursive circumstances (such as physical setting) as well as discursive conditions (such as available discursive sources and strategies). 
This concept of context fits in with the tradition of socio-linguistics and the ethnography of communications (HYMES, 1974). It is a general concept and as such it is rather indeterminate. But this indeterminacy is unavoidable (SILVERSTEIN, 1992) and it would be fruitless to try to elaborate this concept on behalf of a general set of concrete criteria to determine what should be taken into account as the relevant context. As a consequence, the kind and measure of contextualization needed on behalf of the analysis of textual data should be derived from the research design. 
First, the required contextual information should depend of the research goal. For example, if the research is focused on the reconstruction of rules used by participants during communicative interaction then a relatively formal analysis will suffice without extended knowledge about contextual features. 
Second, the required contextual information should also depend on the type of data. For example, the analysis of everyday conversations or institutionalized conversations such as counseling require much more in-depth of information about the history of the social relations between the conversational partners than in the case of analyzing interview discourse. 
These two guidelines with respect to the required contextualization are still very general. To avoid the risk of endless regress, the researcher should adhere to the principle of parsimony. This principle implies the recognition that complete contextualization is unattainable and that contextualization is always limited. Therefore one should restrict oneself to the contextual features that are assumed to be the most relevant for interpreting textual data. This principle implies also that in determining contextual relevance, the introduction of untested assumptions is unavoidable. 
What are the implications of this concept of context and the proposed general guidelines and principles for the feasibility of secondary analysis of qualitative interviews? 
First, the feasibility of secondary analysis depends on the research goal. For some research goals a vast amount of contextual information is required. If this is the case, research based on primary data collection is of course the preferred option. 
Second, the feasibility of secondary analysis depends also on the type of textual data. Generally, qualitative interviews are more suitable for secondary analysis then conversational data. 
Last but not least, the feasibility of secondary analysis depends on the amount of available contextual information. Therefore, it would be helpful to develop a set of minimum guidelines concerning the aspects of context that should be accounted for in the analysis of textual data and that should be made available on behalf of secondary analysis. Inspired by the discussions during the workshop about collaborative reanalyzing qualitative interviews from different angles, I would like to propose some minimum guidelines on contextual information required on behalf of secondary analysis of qualitative interviews.
Information about the discursive context of interviewee's responses: Although researchers are often mainly interested in tales and responses of interviewees, it is a prerequisite that information about the interaction between interviewer and interviewee is made available. Interview discourse is fundamentally co-constructed. Therefore, audiotapes or at least detailed transcriptions of interviews should be available. On behalf of these transcriptions, the standard developed in conversation analysis should be adopted.
Information about the discursive history of interviewee's responses: Interviewee's responses are not isolated pieces of information but elements of a trajectory of interview discourse. Therefore the whole interview should be made available instead of parts.
Information about background characteristics of interviewer and interviewee that are knowable or visible for the participants and could influence the course of the interaction. One of the issues in the controversies about contextualization concerns the information about background characteristics of interviewer and interviewee. At least those social characteristics that are knowable or visible for the participants themselves (such as age, gender, race, social class) should be made available.
Information about the place, time and setting of the interview, such as presence of third persons. As before, those characteristics of place, time and setting that are visible for the participants and are probably relevant for the interaction, should be included in the contextual information.
Information about how the interviewee is selected and approached to cooperate. Interview discourse is partly determined by the way the interview situation and the interview goal or research goal are framed. (VAN DEN BERG, 1996). The selection and the approaching activities are crucial in this framing process. Therefore, information about these activities should be made available.
Information about relevant others that are known to the interviewee as well as the interviewer such as gatekeepers, other interviewees, etcetera. The common knowledge base of interviewee and interviewer may include information about aspects of the wider social context. The interviewee may know something about other research activities of the researcher such as the involvement of other interviewees and gatekeepers. This information could be very relevant as contextual information needed on behalf of the interpretation of interview discourse (VAN DEN BERG, 1996). 
The contextual information outlined above and distinguished in six different types of information, constitute a minimum. As a consequence, it may be insufficient for some research goals. It is unavoidable that secondary analysis permits only a limited range of possibilities for research. For some research goals primary data collection and analysis will remain the only road to follow. For other research goals secondary analysis may constitute a very fruitful alternative provided that sufficient contextual information is made available. Therefore it is necessary to develop standards about the minimum of contextual information made available for secondary analysis. That should contribute to the usability of secondary analysis as a viable research practice. 
1) For example, the existing software on behalf of computer-assisted analysis of qualitative data is mainly oriented to a Grounded Theory approach of thematic analysis of interview transcripts and text in general. <back>
2) The outline of the research project and the main research findings are presented in detail in WETHERELL and POTTER (1992). <back>
3) See for a more elaborate discussion on how to analyze contradictions in interview discourse: VAN DEN BERG (2003). <back>
Billig, Michael (1999a). Whose Terms? Whose Ordinariness? Rhetoric and the Ordinariness in Conversation Analysis. Discourse & Society, 10, 543-558.
Billig, Michael (1999b). Conversation Analysis and the Claims of Naivety. Discourse & Society, 10, 572-576.
Blommaert, Jan (1997). Workshopping: Notes on Professional Vision in Discourse Analysis. Wilrijk: Antwerp Papers in Linguistics 91.
Blommaert, Jan (2001). Context is/as Critique. Critique of Anthropology, 21(1), 13-32.
Bulcaen, Chris & Blommaert, Jan (1997). Eindrapport VFIK Project 307: begeleiding van migrantenvrouwen en – meisjes in centra voor residentieel welzijnswerk. Antwerp: IprA Research Center.
Corti, Louise (2000). Progress and Problems of Preserving and Providing Access to Qualitative Data for Social Research—The international Picture of an Emerging Culture [58 paragraphs]. Forum Qualitative Sozialforschung / Forum: Qualitative Social Research [Online Journal], 1(3), Art. 2. Available at: http://www.qualitative-research.net/fqs-texte/3-00/3-00corti-e.htm.
Fairclough, Norman (1995). Media Discourse. London: Edward Arnold.
Fairclough, Norman (2003). Analysing Discourse; Textual Analysis for Social Research. London: Routledge
Hammersley, Martyn (2003). Conversation Analysis and Discourse Analysis: Methods or Paradigms? Discourse & Society, 14(6), 751-783.
Hammersley, Martyn & Atkinson, Paul (1995). Ethnography; Principles in Practice. London: Routledge.
Hymes, Dell H. (1974). Foundations in sociolinguistics; An ethnographic approach.
Philadelphia: University of Pennsylvania Press.
Schegloff, Emanuel A. (1992). In Another Context. In Alessandro Duranti, & Charles Goodwin (Eds.), Rethinking Context; Language as an Interactive Phenomenon (pp.191-229). Cambridge: Cambridge University Press.
Schegloff, Emanuel A. (1997). Whose Text? Whose Context? Discourse & Society, 8, 165-187.
Schegloff, Emanuel A. (1998). Reply to Wetherell. Discourse & Society, 9, 413-416.
Schegloff, Emanuel A. (1999a). Schegloff's Texts' as Billig's Data. Discourse & Society, 10, 558-572.
Schegloff, Emanuel A. (1999b). Naivety Versus Sophistication or Discipline Versus Self-indulgence. Discourse & Society, 10, 577-583.
Silverman, David (1998). Harvey Sacks; Social Science and Conversation Analysis. New York: Oxford University Press.
Silverstein, Michael (1992). The Indeterminacy of Contextualization: When Is Enough Enough? In Peter Auer & Aldo di Luzio (Eds.), The Contextualization of Language (pp.55-77). Amsterdam: John Benjamins Publishing Group.
Van den Berg, Harry (1992). Racismeonderzoek en discoursanalyse; Een analyse van "Communicating Racism" (1987) en "Racism in the Press" van Teun van Dijk. Migrantenstudies, 8(1), 38-61.
Van den Berg, Harry (1996). Frame Analysis of Open Interviews on Interethnic Relations. Bulletin de Methodologie Sociologique, 53, 5-32.
Van den Berg, Harry (2003). Contradictions in Interview Discourse. In Harry van den Berg, Margaret Wetherell & Hanneke Houtkoop-Steenstra (Eds.), Analyzing Race Talk: Multidisciplinary Approaches to the Interview (pp.119-138). Cambridge: Cambridge University Press.
Van den Berg, Harry, Wetherell, Margaret & Houtkoop-Steenstra, Hanneke (Eds.) (2003) Analyzing Race Talk: Multidisciplinary Approaches to the Interview. Cambridge: Cambridge University Press
Van Dijk, Teun (1987). Communicating Racism: Ethnic Prejudice in Thought and Talk. London: Sage.
Van Dijk, Teun (1991). Racism and the Press. London: Routledge.
Van Dijk, Teun (1997). The Study of Discourse. In Teun Van Dijk (Ed.), Discourse as Structure and Discourse as Process (pp.1-34). London: Sage.
Verschueren, Jef (2001). Predicaments of Criticism. Critique of Anthropology, 21(1), 59-81.
Wetherell, Margaret (1998). Positioning and Interpretative Repertoires: Conversation Analysis and Post-Structuralism in Dialogue. Discourse & Society, 9, 387-412.
Wetherell, Margaret (2001). Debates in Discourse Research. In Margaret Wetherell, Stephanie Taylor & Simeon J. Yates (Eds.), Discourse Theory and Practice; A Reader (pp.380-400). London: Sage.
Wetherell, Margaret (2003). Racism and the Analysis of Cultural Resources in Interviews. In Harry van den Berg, Margaret Wetherell & Hanneke Houtkoop-Steenstra (Eds.), Analyzing Race Talk: Multidisciplinary Approaches to the Interview (pp.11-31). Cambridge: Cambridge University Press.
Wetherell, Margaret & Potter, Jonathan (1992). Mapping the Language of Racism; Discourse and the Legitimation of Exploitation. New York: Harvester Wheatsheaf.
Wodak, Ruth (2001). What CDA Is About—A Summary of its History, Important Concepts and its Developments. In Ruth Wodak & Michael Meyer (Eds.), Methods of Critical Discourse Analysis (pp.1-14). London: Sage.
Harry VAN DEN BERG is an Associate Professor in the Department of Social Research Methodology at the Vrije Universiteit, Amsterdam. He teaches methods of qualitative interviewing and discourse analysis. His research interests include methods of discourse analysis, mixed methods research and methodological research of qualitative and survey interviews.
Harry van den Berg
Department of Research Methodology
De Boelelaan 1081-c
1081 HV Amsterdam
Van den Berg, Harry (2005). Reanalyzing Qualitative Interviews From Different Angles: The Risk of Decontextualization and Other Problems of Sharing Qualitative Data [48 paragraphs]. Forum Qualitative Sozialforschung / Forum: Qualitative Social Research, 6(1), Art. 30, http://nbn-resolving.de/urn:nbn:de:0114-fqs0501305.