Andrea Doucet and Natasha Mauthner
 
 

Data analysis: A neglected area?

While recent years have seen a proliferation of texts within the area of qualitative research (e.g. Mason 1997; Ribbens and Edwards, 1997; Sapsford and Judd, 1996; Wolf, 1996), one area which continues to receive only limited attention concerns the detailed processes of data analysis. While there are several important general texts which discuss data analysis (e.g. Bryman and Burgess, 1994; Coffey and Atkinson, 1996; Hayes, 1997; Sapford and Judd, 1996; Ribbens and Edwards, 1997; Silverman, 1993; Wolcott, 1994), as well as texts outlining particular methods of data analysis (Glaser and Strauss, 1967; Hammersley and Atkinson, 1983; Miles and Huberman, 1984; Riessman, 1993; Strauss and Corbin, 1990), detailed discussions of the ënitty-grittyí of data analysis and the actual step-by-step processes of how interview transcripts are analyzed are rarely presented. A similar point can be made about specific research studies that rarely report the concrete details, difficulties and dilemmas of data analysis; indeed, restrictions on length of publications, particularly in journals, often preclude lengthy discussions of such methods. Moreover, there appears to be an increasing move to equate computer ëcodingí with qualitative data ëanalysisí, with several recent texts on ëqualitative data analysisí focusing on the use of computer programs and which software package to employ (Dey, 1993; Fielding and Lee, 1991; Kelle, 1995; Tesch, 1990; Weitzman and Miles, 1995).

The literature on qualitative research methods therefore has relatively little to say about the detailed and concrete processes of data analysis. For example, how do researchers use and adapt particular methods or combinations of methods within the contexts of specific research projects? Why and how do they choose particular methods of data analysis? How does an individual researcherís processes of data analysis reflect her/his theoretical and epistemological orientations? Qualitative research projects produce mountains of data; why and how do researchers decide to focus on some issues rather than others? And what personal, professional, political, practical, theoretical, institutional constraints or hidden agendas influence these various decisions and processes? How are particular ëcodesí or interpretations are arrived at? How are they ëgroundedí in respondentsí voices and perspectives? Central to qualitative, and particularly feminist research, is the notion of listening to respondents and understanding their lives ëin and on their own termsí (e.g. Denzin and Lincoln, 1994; Finch, 1984; Gilligan, 1982; Graham, 1983; Oakley, 1981). How is this general methodological principle operationalized within the actual research process and particularly during the analysis of the data? Similarly, the issue of ëreflexivityí is critical to qualitative and feminist research alike (eg Altheide and Johnson, 1994; Harding, 1992; Stanley and Wise, 1993); yet little attention has been given to issues of reflexivity and power, voice and authority specifically in the data analysis stage of the research. Thus, data analysis presents researchers with the challenge of keeping respondentsí voices and perspectives alive, while at the same time recognizing the researcherís role in shaping the research process and product. How have researchers grappled with this tension and dilemma?

These issues are central to social research given that the processes through which we transform respondentsí private lives into public theories, and our role as researchers within these processes, are critical to assessing the validity and status of these theories. As Miles and Huberman point out, ëThe strengths of qualitative data rest very centrally on the competence with which their analysis is carried outí (1994: 10; see also Bryman and Burgess, 1994; Glucksmann, 1994).

Although we raise a range of epistemological, theoretical and ontological questions concerning data analysis processes in this paper, our intention is to describe the ënitty-grittyí of how we analyzed verbatim transcripts of depth interviews gathered in the course of our respective doctoral research projects. Throughout the discussion we draw on examples from our research which employed individual, joint, and repeat, semi-structured, open-ended interviewing. Andrea Doucetís research investigated the experiences of 23 British dual earner couples with dependent children who identified themselves as ëconsciously attempting to share the work and responsibility for housework and child careí (see Doucet, 1995a, 1995b, 1996). Natasha Mauthnerís research explored womenís experiences of motherhood and postnatal depression through interviews with 40 mothers (and 25 of their male partners) of young children living in England, 18 of whom experienced postnatal depression (see Mauthner, 1993, 1994, 1995, 1998). We begin by offering reflections on why data analysis is such a difficult task. Following this discussion, we turn to a detailed treatment of data analysis in practice. In our conclusions, we highlight several methodological and epistemological issues that have arisen for us in the processes of reflecting upon data analysis. Throughout the paper we give particular emphasis to issues that have been raised by feminist researchers.

The difficulties of articulating what we do when we analyze qualitative data

Why is articulating what we do when we analyze qualitative data such a difficult and elusive task? The latter stages of data analysis, which tend to be structured, methodical, rigorous and systematic, are often easily described. For example, once a critical set of issues has been identified, the data are systematically scanned for examples of particular themes. However, the initial stages of actually getting to know the data and identifying what are the key issues feel more intuitive than anything else. As Bryman and Burgess have noted: ë... much of the work in which investigators engage in this phase of the research process is as much implicit as explicití (1994: 12). Thus, as qualitative researchers, we engage in a somewhat unsystematic process of following up certain leads and seeing where they take us. In deciding which ideas to follow up we are undoubtedly influenced, whether consciously or not, by our own personal, political and theoretical biographies. But the reasons why we choose some ideas rather than others are not always immediately obvious to us; nor are there necessarily logical reasons for our choices and decisions. The early phases of data analysis can therefore feel messy, confusing and uncertain because we are at a stage where we simply do not know what to think yet. Indeed, this is the whole point of data analysis - to learn from and about the data; to learn something new about a question by listening to other people. But while this sense of not knowing and of openness is exciting, it is also deeply uncomfortable. These kinds of processes are very difficult to articulate, especially in the logical, sequential, linear fashion that tends to be required in a research text. A further difficulty and constraint is the gap we experience between how research is executed and carried out on a concrete and daily basis, and our written (re)constructions of our data analysis steps and processes.

Perhaps data analysis is also difficult to articulate because in doing so we are directly confronted with the subjective, interpretive nature of what we do - having to interpret respondentsí words in some way, while realizing that these words could be interpreted in a multitude of ways. It is now well recognized in many feminist critiques, as well as within work associated with postmodernism and with the longer-standing hermeneutic (Dilthey, 1900/1976; Gadamer, 1975; Ricoeur, 1979) or interpretive traditions (Blumer, 1969; Mead, 1934), that all research contains biases and values, and that knowledge and understanding are contextually and historically grounded, as well as linguistically constituted. These critiques, while distinct in many ways, share a common emphasis on deconstructing Enlightenment and modernist ideals of objectivity, scientific thought, dualisms and rationality (see Du Bois, 1983; Harding, 1992; Heckmann, 1990; Stanley and Wise, 1983, 1993). Feminists, for example, have argued for over two decades that understanding and knowledge come from being involved in a relationship with our subject matter and respondents, and not through adopting a detached and objective stance (e.g. Smith, 1974). Indeed the production of knowledge must contain a systematic examination and explication of our beliefs, biases and social location (Harding, 1992). This reflexivity ensures that the politics underlying the methods, topics, and governing assumptions of our scholarship are analyzed directly and self-consciously, rather than remaining unacknowledged (Crawford and Marecek, 1989). While these are laudable intentions, these biases and beliefs may be extremely difficult to uncover or even to notice: we may not have the practical means to do so in the busy process of analyzing interview transcripts; it may be quite uncomfortable to do so; a profound level of self-awareness is required to begin to capture the perspectives through which we view the world; and it is not easy to grasp the ëunconsciousí filters through which we experience the world.

In other words, in analyzing data we are confronted with ourselves, and with our own central role in shaping the outcome. Indeed, perhaps this is part of the reason why computer programs have been so popular: the use of technology confers an air of scientific ëobjectivityí onto what remains a fundamentally subjective, interpretative process. This is not to deny the obvious practical benefits to be gained from computer programs. The point we wish to make is that we need to think critically about how and why we use these programs.

A further reason why we might be tempted to gloss over the question of how we analyze our data stems from our anxiety about whether we have analyzed the data in ëthe right wayí. Even when researchers draw on specific methods of data analysis they use and interpret these methods in their own individual ways. Indeed, researchers who jointly develop a particular method can actually use the same method differently. The case of Glaser and Strauss (1967), and the difference of opinion or ëhead on clashí (Melia, 1996: 368) that developed between them as to what exactly ëgrounded theoryí is, provides a particularly good example of this (see Glaser, 1992; Melia, 1996). We might follow the general principles of a method but not go through all the steps that are specified. Or we might go through all the steps for a select number of cases, and analyze the remainder of the data set in a more speedy fashion particularly when resources of time, energy and money are running out. This can engender a sense of anxiety that we have not proceeded correctly; and rather than be open about exactly what we did and did not do, we might be tempted to simply gloss over the details of data analysis. But this issue has raised a number of unanswered questions for us. Are research texts on data analysis intended to be literally followed step by step? How many researchers who describe using particular methods actually follow all the steps as specified within the original texts? To what extent is it necessary to go through all the steps with each one of the transcripts? Do the researchers writing these texts actually go through all these steps themselves? Is there one right way to use a particular method? And to what extent do methods evolve as different researchers use and adapt them (see also Strauss and Corbin, 1990)?

Data analysis is our most vulnerable spot. It is the area of our research where we are most open to criticism. Writing about data analysis is exposing ourselves for scrutiny. Perhaps it is for these reasons that data analysis fails to receive the attention and detail it deserves.

Data Analysis in Practice: Analyzing domestic and maternal voices

Our discussion here focuses specifically on the analysis of interview transcripts. Clearly, the processes through which we analyze qualitative data also encompass other types of information such as fieldnotes, information gleaned during the setting up of interviews, incidental meetings or conversations with respondents, observational data during the interview, and non-linguistic ëdataí such as bodily and facial expressions, and non-verbal interactions between the couple in the case of joint interviews. However, in this paper we have not included any detailed discussions of how such information informs or interacts with the analysis of interview transcripts.

Furthermore, in focusing specifically on the analysis of interview transcripts we are inevitably presenting a somewhat static and simplified picture of what is in fact a complex, dynamic process. ëData analysisí is not a discrete phase of the research process confined to the moments when we analyze interview transcripts. Rather, it is an ongoing process which takes place throughout, and often extends beyond, the life of a research project. For example, our interpretive work started when we first accessed the sample of people we wished to include in our studies. During the interviews, we were actively listening to participantsí stories, asking questions and leading respondents down certain paths and not others, making decisions about which issues to follow up, and which to ignore, and choosing where to probe. We were guided by our initial research agenda and questions, what each respondent said to us, and our interpretations and understandings of their words. The interview content was therefore a joint production (see Mishler, 1986) and part of what we were doing in shaping the interview was following our own analytical thinking. Moreover, with each interview, and with the analytical work we did during and after each interview, we formulated new ideas or approaches, and modified our interview questions so as to ëcheck these outí. The process of analysis continued in a more explicit way as we transcribed the interviews and began to immerse ourselves in the data through full transcript readings. We began to interpret the meaning of each respondentís stories, and noted areas of difference and overlap with other participantsí accounts. Finally, data analysis overlapped with and was ongoing during the writing up the research.

However, what we wish to focus on in this paper is the intensive period of 17 months when our research efforts were devoted primarily to analyzing our interview transcripts. Some of this time was spent analyzing transcripts within the context of a research group which we describe. We discuss the analytical procedures which we conducted during this concentrated period which we present in two distinct phases of data analysis: (1) our use of our adapted version of the ëvoice-centered relational methodí of data analysis involving four readings, case studies and group work; and (2) summaries and thematic ëbreaking downí of the data. The first stage took approximately 15 months and the second stage two months.

Stage One - A voice-centered relational method of data analysis: Four readings, case studies and group work

The voice centered relational method of data analysis: theoretical and ontological assumptions:

One of the main reasons that we both used the ëvoice centered relational methodí of data analysis was because, as graduate students, we were presented with the opportunity to join a small graduate research group set up by Carol Gilligan during her term as visiting professor at the University of Cambridge. The aim of the group was specifically to learn how to use this method as well to explore the theoretical and methodological ideas that underpin it. This method of data analysis was developed over several years by Lyn Brown, Carol Gilligan and their colleagues at the Harvard Project on Womenís Psychology and Girlsí Development at the Harvard Graduate School of Education (see Brown and Gilligan, 1992, 1993; Brown et al., 1988; Gilligan et al., 1990; Taylor et al., 1995). ³ The method has its roots in clinical and literary approaches (Brown and Gilligan, 1992), interpretive and hermeneutic traditions (Brown et al., 1989, 1991; Gilligan et al., 1990), and relational theory (Belenky et al., 1986; Brown and Gilligan, 1992; Gilligan, 1982; Gilligan, 1988; Gilligan et al., 1990; Miller, 1976/1986).

The method holds at its core the idea of a relational ontology. In contrast to the predominant ontological image in liberal political thought, neo-classical economics and neo-Kantian philosophy of a separate, self-sufficient, independent and rational ëselfí a ërelationalí ontology posits the notion of ëselves-in-relationí (Ruddick, 1989: 211), ërelational beingí (Jordan, 1993:141) and interdependent ëselfí (Taylor et al., 1995: 30), a view of human beings as embedded in a complex web of intimate and larger social relations (Gilligan, 1982); and a ëdifferent understanding of human nature and human interaction so that people are viewed as interdependent rather than independentí (Tronto, 1995: 142). The relational ontological assumptions which directly inform the development of this method of data analysis arose out of extensive developmental psychological and educational research, and clinical practice, with women and girls across class and race. While the method is rooted in this tradition, it is important to emphasize that this relational ontology has also been developed and theorized in other disciplines, particularly in political theory, feminist philosophy, feminist legal theory, and feminist economics (Baier, 1993; Benhabib, 1987, 1992; Folbre, 1994; Gilligan, 1988; Minow and Shanley, 1996; Ruddick, 1989; Tronto, 1989, 1993, 1995). Furthermore, there are conceptual parallels between a relational ontology and sociological emphases on understanding individuals within their social contexts and on exploring the ëdualityí of social structures and human agency (see Giddens, 1984). Also, a relational ontology is consistent with many of the assumptions espoused by the interpretive or symbolic interactionism tradition within sociology (Blumer, 1969; Mead, 1934) in that there is an emphasis on the self and identity as eminently social interactive processes.

The voice-centered relational method represents an attempt to translate this relational ontology into methodology and into concrete methods of data analysis by exploring individualsí narrative accounts in terms of their relationships to the people around them and their relationships to the broader social, structural and cultural contexts within which they live. While using the method under the guidance of Carol Gilligan, it is important to emphasize that we were also simultaneously developing our own version of it. Thus we drew on the excellent work which had begun at Harvard University but we adapted it so as to reflect our interdisciplinary backgrounds and our specific research interests. In particular, we were both interested in emphasizing and refining its application for projects that include a sociological focus. Also, our version of the method is also deeply rooted within the broader tradition of feminist research practice and the increasingly rich and wide field of qualitative research (see Mauthner and Doucet, 1998).

We emphasize this point that what we are describing in this paper is very much our own version of this method of analysis. In reading the work of other researchers who have also used this method at Harvard University (e.g. Geismar, 1996; Rogers, 1994; Tolman, 1992; Way, 1994), it has become apparent to us that individual researchers use and adapt particular methods in their own individual ways. Researchersí individuality, their particular topics, their samples, the theoretical and academic environments and social and cultural contexts in which they work all influence the ways in which they use particular methods. Thus, although we both used this method, we picked up on and emphasized different elements of it. The method worked differently for us because of our different topics but also because of differences in our own biographies. In discussing how we used a voice-centered relational method we are therefore not discussing ëtheí method but rather our own individual interpretations, adaptations and versions of it.

Our version and adaptation of the voice centered relational method: The method revolves around a set of three or more readings of the interview text, and the original tapes can be listened to as these readings are carried out. We actually conducted four readings of selected interview transcripts. This paper will deal only with the first two readings (for greater detail, please consult Mauthner and Doucet, 1998). We also will discuss how the voice centered relational method formed part of a wider approach to data analysis which also included group work, case studies, summaries and themes.

Reading 1: Reading for the plot and for our responses to the narrative: The first reading we did comprised two elements. First, the text is read for the overall plot and story that is being told by the respondent - what are the main events, the protagonists, and the subplots. We listened for recurrent images, words, metaphors and contradictions in the narrative. This element is one which is common to many methods of qualitative data analysis (see Riessman, 1993; Strauss and Corbin, 1990).

In the second ëreader-responseí element of this first reading, the researcher reads for herself in the text in the sense that she places herself, with her own particular background, history and experiences, in relation to the person she has interviewed (Taylor et al., 1995). The researcher essentially reads the narrative on her own terms - how she is responding emotionally and intellectually to this person. Lyn Mikel Brown describes this process:

 The first reading of the interview text thus represents an attempt to come to know our response to the respondent and her or his story. The underlying assumption here is that by trying to name how we are socially, emotionally and intellectually located in relation to our respondents we can retain some grasp over the blurred boundary between their narratives and our interpretations of those narratives. If we fail to name these emotions and responses, they will express themselves in other ways such as in our tone of voice or the way in which we write about that person. The aim of this reading is also to lay down the evidence of our responses for others to see. A further assumption underlying this ëreader responseí reading is that our intellectual and emotional reactions to other people constitute sources of knowledge; it is through these processes that we come to know other people (see also Hammersley and Atkinson, 1983; Miles and Huberman, 1994).

Reading 2: Reading for the voice of the ëIí : The second reading we conducted was similar to that described by Brown and Gilligan (1992) and focused on how the respondent experiences, feels and speaks about herself. In an attempt to get some sense of this through the empirical data, we followed the method of using a coloured pencil to physically trace and underline certain of the respondentís statements in the interview transcript ? namely where the respondent uses personal pronouns such as ëIí, ëweí or ëyouí in talking about themselves. This process centers our attention on the active ëIí which is telling the story; amplifies the terms in which the respondent sees and presents herself; highlights where the respondent might be emotionally or intellectually struggling to say something; and identifies those places where the respondent shifts between ëIí, ëweí and ëyouí signaling changes in how the respondent perceives and experiences herself. Spending this time carefully listening to the respondent creates a space between her way of speaking and seeing and our own, so we can discover ë... how she speaks of herself before we speak of herí (Brown and Gilligan, 1992: 27-28).
 

 This second reading represents, in a sense, the first step of a phased process of listening to this person as they speak about themselves and the life that they live and the world that they inhabit. From the point of view of sociology, this second reading represents an attempt to hear the person, agent or actor voice their sense of agency, while also recognizing the social location of this person who is speaking. This stage of the data analysis represents an attempt to stay, as far as it is possible, with the respondentsí multi-layered voices, views and perspectives rather than simply and quickly slotting their words into either our own ways of understanding the world or into the categories of the literature in our area. In our view, this detailed and focused attention on the voice(s) of the ëIí increases the possibility of creating more or less space within which to hear our respondentsí voices; and to take more or less time doing so.

This reading for the personal pronoun statements strikes us as being one of the key features which distinguishes a voice-centered relational method of data analysis from grounded theory, a method which is used widely by sociologists conducting qualitative data analysis. According to Strauss and Corbin, grounded theory is less interested in ëpersons per seí and more interested in action/interaction:

The aim of theoretical sampling is to sample events, incidents, and so forth, that are indicative of categories, their properties and dimensions, ... we sample incidents and not persons per se! Our interest is in gathering data about what persons do or donít do in terms of action/interaction; the range of conditions that give rise to that action/interaction and its variations; how conditions change or stay the same over time and with what impact; also the consequences of either actual or failed action/interaction or of strategies never acted on. (1990: 177)

Thus grounded theory seems more concerned with action and interaction and less so with the processes of reflection and decision making which are key concerns of a voice-centered relational method of data analysis. Thus, it may be that in researching areas of ëprivateí life where process-oriented values and ways of being are emphasized rather than the more ëpublicí goal-oriented values and ways of being (Ribbens and Edwards, 1997), the voice-centered relational method was quite instrumental in helping to shed light on the meanings, processes, relationships and contradictions which are central to domestic life.

The first two readings are the ëstaplesí of the method in that researchers using this method of data analysis would always undertake these. Generally speaking, researchers have conducted two further readings of their own choice depending on their research topic (see Brown and Gilligan, 1993; Mauthner and Doucet, 1998).

Case studies and group work

The work in our research group involved us writing up our ongoing thoughts and analyses about a particular respondent in the form of case studies. This detailed and time-consuming work was valuable for understanding the depth and complexity of individualsí experiences, as well as the very significant differences between our respondentsí narratives. Working within the context of a group was useful because, having read extracts from our transcripts, others were able to point out where we might have missed or glossed over what they regarded as key aspects of the interview narrative. This made us acutely aware of our own role and power in choosing the particular issues we emphasize and pick up on, and which we ignore or minimize.

One of the drawbacks of how we used a voice-centered method is that ideally it requires a great deal of time. As a result we found it impossible to systematically conduct all four readings with each and every one of our respondents; we were only able to focus such detailed attention on a select number of cases. Nevertheless, the energy and time we put into these few cases served the function of ëtuning our earí. We read the remaining narratives listening for the issues or voices we had by then identified as both critical in terms of understanding the experiences of our respondents, and also ënewí or challenging within our particular disciplinary areas. What is clear to us, however, is that in moving from the slow and careful work with individual cases to the more speedy process of reading through the other transcripts, we begin to selectively focus in on certain issues while shutting out others. This seems inevitable to us. In part, we reach a ësaturation pointí where we have enough and even too many ënewí issues we wish to write about and contribute to our areas of work. But shortages of time and resources are obviously a further constraint on the extent to which we analyze the data. It is important to recognize and acknowledge that these processes are taking place.

Overall the four readings of the interview transcripts emphasize the multi-layered nature of narratives and trace voices across and within a particular transcript. This approach is fundamentally different to the thematic organization characteristic of most methods of data analysis, including those assisted by computer programs. It delays the reductionistic stage of data analysis when transcripts are cut up into themes and aggregated. This process shifts data analysis away from traditional ëcodingí, which implies fitting a person into a pre-existing set of categories, whether those of the researcher or those of established theoretical frameworks. Furthermore, tracing voices through individual interview transcripts, as opposed to linking themes across interviews, helps maintain differences between the respondents.
 

The detailed and lengthy focus on individual interviews embodies respect for individual respondents within the research context. If we do not take the time and trouble to listen to our respondents, data analysis risks simply confirming what we already know. If this is the case, in no way has the respondent changed our view or understanding, thus defeating the point of doing the study in the first place. At the same time, this approach respects the role of the researcher and indeed the necessity of the researcher having her own voice and perspective in this process. By providing a way of reading and listening to an interview text ë... that takes into account both our stance as researchers and the stance of the person speaking within the textí (Gilligan et al., 1990: 96), this approach respects and to some extent exposes the relationship between researcher and researched. As Gilligan notes, ëthe relationship has to be maintained throughout the writing, and you donít write over, or voice over, other peopleís voices.... Itís an attempt to try to work as a writer would work, by giving people their voice, by giving ourselves a voice in our work, and then thinking very consciously about the orchestration of the pieces we writeí (Kitzinger with Gilligan, 1994: 411). It is in bringing the listener into responsive relationship with the person speaking that this approach or method is characterized as a relational one.

Stage Two - Summaries and Thematic ëBreaking Downí of the Data:

In addition to the detailed case studies, which we did for 10 individuals (or five couples), we wrote up summaries for the remaining individuals which represented short portraits of each respondent of one to two pages. We also both felt it was necessary and important to try to move from the holistic understandings of individual respondents described above to tackling the data set as a whole. This decision to ëcut upí the transcripts was a difficult moment in our research. Having spent so many months on a relatively small number of respondents, we felt anxious to make the huge volume of data more manageable. At the same time, we were frustrated not to be able to devote the same amount of time and energy to each one of our respondents. We felt we were short-changing many of them; we missed the process of getting to know and understand another story; and above all we feared that in ëcutting upí the data we would lose much of its complexity. Despite this apprehension, we proceeded to break up each transcript into a number of overlapping themes and sub-themes. Natasha did this manually on the computer (through cut and paste) while Andrea conducted a two staged process of working manually and then using a computer based programme (text-base alpha). Many of these themes, and sub-themes in particular, emerged as a direct result of the intensive case study work and provided a way of linking the details of individual respondents with the stories told by the data sets as wholes.

The analysis of the data therefore involved organizing the data in different ways (tapes; verbatim transcripts; 4 readings; case studies; summaries; themes) in order to tap into different dimensions of the data sets. It also involved a dialectical process of moving between different ways of organizing or representing the data, and between the details and particularity of each one of the individual respondentís experiences, and the overall picture of the samples as wholes.

Conclusions

We believe that data analysis is a critical stage in the research process for it carries the potential to decrease or amplify the volume of our respondentsí voices. As the site where their stories and ëvoicesí become ëtransformedí into theory, what goes on during data analysis strikes us as being central to the fundamental concern of feminists: ëthe intertwined problem of realising as fully as possible womenís voices in data gathering and preparing an account that transmits those voicesí (Olesen, 1994: 167). We cannot emphasise enough how these processes between ëdata gatheringí and transmitting ëthose voicesí has received only sparse attention in feminist research and in the more general field of qualitative research. Indeed in place of a move towards greater links between empirical research practice and epistemological discussions, we would concur with Maynard (1994: 22) who has noted that ëarguments about what constitutes knowledge and discussions about methods of research are moving in opposite directionsí. Our purpose in writing this chapter is to join those authors who are concerned about this increasing gap between abstract philosophical discussions about epistemology and research revealing the daily lives of women and men in domestic and private settings.

As we have gradually come to appreciate our omnipresence throughout all the stages of the research, we now feel that the feminist aim of listening to women ëin and on their own termsí is to some extent impossible. We are thus critical of the tendency by some feminist researchers to simplify the complex processes of representing the ëvoicesí of research respondents as though these voices speak on their own (see, for example, Reinharz, 1992: 267), rather than through the researcher who has already made choices about how to interpret and which quotes and interpretations to present as evidence. There is therefore a contradiction, as we see it, between two of the principles which are fundamental to feminist research: the commitment to listen to women on their own terms and the recognition that it is the researcher who ultimately shapes the entire research process and product. Instead, we have found it helpful to think of the research process as involving a balancing act between three different and sometimes conflicting standpoints: (i) the multiple and varying voices and stories of each of the individuals we interview; (ii) the voice(s) of the researcher(s); (iii) the voices and perspectives represented within existing theories or frameworks in our research areas and which researchers bring to their studies. We view research both in terms of process (how we do research) and product (the production/social construction of knowledge) as a journey in which these three ëvoicesí or perspectives must be listened to, maintained and respected, and the processes whereby we make critical shifts between these three ëvoicesí be charted for other researchers to build on or to critique (see also Edwards and Ribbens, 1997). It should also be mentioned that in the end, further sets of voices enter the conversation and these include the voices and perspectives of those that ultimately read and interpret the texts we produce.

We also find it useful to think of the research process and product in terms of degrees rather than absolutes. We can never claim to have captured the ëpureí, ërealí, ërawí or ëauthenticí experiences or voices of our respondents because of the complex set of relationships between the respondentsí experiences, voices and narratives, and the researcherís interpretation and representation of these experiences/voices/narratives. However, there are ways in which we can attempt to hear moreof their voices, and understand more of their perspective through the ways in which we conduct our data analysis. This paper has highlighted one of the key dilemmas we face as researchers: on the one hand, we play a critical role in transforming private lives and concerns into public theories and debates and in voicing what might otherwise remain invisible and/or devalued issues pertaining to domestic life. On the other hand, in the process of transformation, the private account is changed by and infused with our identity - and thereby becomes a different story to that originally told by the respondent(s). We cannot be sure we have faithfully reported our respondentsí concerns. At the same time, as academic researchers, our role involves more than this for we are also required to theorise our respondentsí accounts and lives, and locate them within wider academic and theoretical debates. We have to accept the losses and gains involved in this process, and hope that a version of our respondentsí concerns is made public, even if it is neither their exact version nor necessarily all of the issues they regard as paramount. Moreover, whatever the losses and gains involved in moving from talk, to text, to theory, we must document the paths, detours, and shortcuts we have chosen at each stage of the research journey.

References:

Anderson, K. and Jack, D. (1991). Learning to listen: interview techniques and analysis, in S. Gluck and D. Patai (eds), Women's words:The feminist practice of oral history. London: Routledge. pp. 11-26.

Baier, A. C. (1993).  What do women want in moral theory?, in M. J. Larrabee (ed.), An ethic of care: Feminist and interdisciplinary perspectives. London: Routledge. pp. 19-31.

Belenky, M. F., Clinchy, B. M., Goldberger, N. R. and Tarule, J. M. (1986).  Women's ways of knowing: The development of self, voice, and mind. New York: Basic Books.

Benhabib, S. (1987). The generalized and the concrete other: the Kohlberg-Gilligan controversy and feminist theory, in E. Kittay and D. Meyers (eds), Women and moral theory. Totowa, NJ: Rowman and Littlefield. pp. 154-177.

Benhabib, S. (1992).  Situating the self. Cambridge: Polity Press.

Blumer, H. (1969).  Symbolic interactionism: Perspective and method. Berkeley: University of California Press.

Brown, L. M., Debold, E., Tappan, M. and Gilligan, C. (1991).  Reading narratives of conflict and choice for self and moral voices: A relational method, in William M. and J. L. Gewirtz (eds), Moral behaviour and development, Volume 2: Research. Hilldale, New Jersey: Lawrence Erlbaum. pp. 25-61.

Brown, L. M. (1994).  Standing in the crossfire: a response to Tavris, Gremmen, Lykes, Davis and Contratto. Feminism and psychology 4, 382-98.

Brown, L. M., Argyris, D., Attanucci, J., Bardige, B., Gilligan, C., Johnston, K., Miller, B., Osborne, R., Tappan, M., Ward Janie, Wiggins G. and Wilcox D. (1988).  A guide to reading narratives of conflict and choice for self and relational voice. Cambridge, MA: Harvard Graduate School of Education.

Brown, L. M. and Gilligan, C. (1992). Meeting at the crossroads: Women's psychology and girls' development. Cambridge, MA: Havard University Press.

Brown, L. M. and Gilligan, C. (1993) Meeting at the crossroads: women's psychology and girls' development. Feminism and Psychology 3, 11-35.

Bryman, A., & Burgess, R. (1994). Analysing qualitative data. London: Routledge.

Code, L. (1995).  How do we know? Questions of method in feminist practice. In  S.D. Burt and L. Code (Eds.), Changing methods: Feminists transforming practice.  Peterborough: Broadview Press.

Coffey, A. and Atkinson, P. (1996). Making sense of qualitative data: Complementary research strategies. London: Sage.

Connell, R.W. (1987). Gender and Power. Cambridge: Polity Press.

Denzin, N. K. and Lincoln, Y. S. (1994). Handbook of qualitative research. London: Sage.

DeVault, M. (1996). Talking back to sociology: Distinctive contributions of feminist methodology. Annual Review of Sociology 22, 29-50.

Dey, I. (1993). Qualitative data analysis: A user-friendly guide for social scientists. London: Routledge.

Dilthey, W. (1900/1976). The development of hermeneutics.  In W. Dilthey, Selected writings. Cambridge: Cambridge University Press.

Doucet A. (1996). Encouraging voices: Towards more creative methods for collecting data on gender and household labour.  In L. Morris and S. Lyon, Gender relations in the public and the private (pp. 156-173).  London: Macmillan.

Doucet A. (1995b). Gender equality and gender differences in household work and parenting. Women's Studies International Forum 18, 271-284.

Doucet A. (1995a). Gender equality, gender differences and care:  Towards understanding gendered labour in British dual earner households. (Doctoral dissertation, University of Cambridge, 1995).

Du Bois, B. (1983). Passionate scholarship: notes on values, knowing and method in feminist social science.  In G. Bowles and R. D. Klein (Eds), Theories of women's studies  (pp. 105-16).London:  Routledge and Kegan Paul.

Edwards, R. (1990). Connecting method and epistemology:  A white woman interviewing Black women. Womenís Studies International Forum 13(5), 477-90.

Fielding, N., & Lee, R. (1991). Using computers in qualitative research. London: Sage.

Finch, J. (1984). "It's great to have someone to talk to":  The ethics and politics of interviewing women.  In C. Bell and H. Roberts (Eds), Social researching:  Politics, problems, practice (pp. 70-87). London: Routledge and Kegan Paul.

Folbre, N. (1994).Who pays for the kids: Gender and the structures of constraint. London and New York: Routledge.

Gadamer, H. (1975). Truth and method. New York: Seabury Press.

Geismar, K. (1996). Bodies of knowledge: Transfiguring the pain of childhood trauma. (Doctoral dissertation, Harvard University, 1996).

Giddens, A. (1984). The constitution of society. Cambridge: Polity Press.

Gilligan, C. (1982). In a different voice. Cambridge, MA: Havard University Press.

Gilligan, C. (1988). Remapping the moral domain: New images of the self in relationship. In C. Gilligan, J. V. Ward, J. M. Taylor (with Betty Bardige) (Eds), Mapping the moral domain: A contribution of women's thinking to psychological thoery and education (pp. 3-20). Cambridge, MA: Harvard University Press.

Gilligan, C., Brown, L. M. & Rogers, A. (1990). Psyche embedded:  A place for body, relationships and culture in personality theory. In A. I. Rabin, R. Zucker, R. Emmons & S. Frank (Eds), Studying persons and lives (pp. 86-147).  New York: Springer.

Glaser, B. G. (1992). Emergence vs. forcing basics of grounded theory analysis. Mill Valley, CA: The Sociological Press.

Glaser, B. G. and Strauss, A. L. (1967). The discovery of grounded theory. Chicago: Aldine.

Glucksmann, M. (1994). The work of knowledge and the knowledge of women's work. In M. Maynard and J. Purvis (Eds), Researching women's lives from a feminist perspective (pp. 149-65). London: Taylor and Francis.

Graham, H. (1993). Caring: A Labour of love.  In J. Finch and D. Groves (Eds), A labour of love: Women, work and caring. London: Routledge and Kegan Paul.

Hammersley, M. and Atkinson, P. (1983). Ethnography: Principles in practice. London: Tavistock.

Haraway, D. (1988). Situated knowledges: The science question in feminism and the privilege of partial perspective. Feminist Studies 14, 575-599.

Harding, S. (1992). Whose science? Whose knowledge?  Milton Keynes: Open University Press.

Harding, S. (Ed.). (1987). Feminism and methodology: Social science issues. Bloomington: Indiana Univeristy Press.

Hayes, N. (1997). Doing qualitative analysis in psychology. Mahwah, NJ:  Lawrence Erlbaum.

Hekman, S. (1990). Gender and knowledge: Elements of a postmodern feminism. Boston: Northeastern University Press.

Holland, J. & Ramazanoglu, C. (1994).  Coming to conclusions: power and interpretation in researching young women's sexuality.  In M. Maynard & J. Purvis (Eds.), Researching women's lives from a feminist perspective (pp. 125-48). London: Taylor and Francis.

Jack, D. (1991). Silencing the self: Depression and women. Cambridge, MA: Havard University Press.

Jordan, J. (1993).  The relational self: A model of womens development.  In J. van Mens-Verhulst, K. Schreurs & L. Woertman (Eds.), Daughtering and mothering: Female subjectivity reanalysed. London: Routledge.

Kelle, U. (1995). Computer-aided qualitative data analysis. London: Sage.

Kitzinger, C. with Gilligan, C. (1994).  The spoken word: listening to a different voice. Feminism and Psychology 4, 399-403.

Mason, J. (1996). Qualitative researching. London: Sage.

Maynard, M. (1994).  Methods, practice and epistemology.  In M. Maynard & J. Purvis (Eds.), Researching women's lives from a feminist perspective. London: Taylor and Francis.

Mauthner, N.S. (1994). Postnatal depression: A relational perspective. (Doctoral dissertation, University of Cambridge, 1994).

Mauthner, N.S. (1995).  Postnatal depression: The significance of social contacts between mothers. Women's Studies International Forum 18, 311-23.

Mauthner, N.S. (1998). Re-assessing the importance and role of the marital relationship in postnatal depression: Methodological and theoretical implications. Journal of Reproductive and Infant Psychology.

Mauthner, N.S. (1993). Towards a feminist understanding of "postnatal depression". Feminism and Psychology 3, 350-355.

Mauthner, N.S.  and Doucet A. (1998). Reflections on a voice centred relational method of data analysis: Analysing maternal and domestic voices.  In J. Ribbens and R. Edwards (Eds.), Feminist dilemmas in qualitative research: Private lives and public texts. London: Sage.

Mead, G. (1934). Mind, self and society. Chicago: University of Chicago Press.

Melia, K. (1996). Rediscovering Glaser. Qualitative Health Research 6, 368-378.

Miles, M. and Huberman, M. (1994). Qualitative data analysis: An expanded sourcebook. London: Sage.

Miller, J. (1976/1986). Towards a new psychology of women. London: Penguin Books.

Minow, M. & Shanley, M. L. (1996). Relational rights and responsibilities: Revisioning the family in liberal political theory and law. Hypatia 11, 4-29.

Mishler, E. G. (1986). Research interviewing: Context and narrative. Cambridge, MA: Harvard University Press.

Oakley, A. (1981). Interviewing women: a contradiction in terms.  In H. Roberts (Ed.), Doing feminist research(pp. 30-61). London: Routledge and Kegan Paul.

Olesen, V. (1994). Feminisms and models of qualitative research.  In N. K. Denzin & Y. S. Lincoln (Eds.), Handbook of qualitative research (pp. 158-74). London: Sage.

Ribbens, J., & Edwards, R. (1997). Feminist dilemmas in qualitative research: Public knowledge and private lives. London: Sage.

Reinharz, S. (1992). Feminist methods in social research. Oxford: Oxford University Press.

Ricoeur, P. (1979). The model of a text: Meaningful action considered as a text.  In P. Rabinow & W. Sullivan (Eds.), Interpretive social science: A reader. Berkeley, CA: University of California Press.

Riessman, C. K. (1993). Narrative analysis. Newbury Park: Sage.

Rogers, A. (1994). Exiled voices: Dissociation and repression in women's narratives of trauma. Wellesley, MA: Stone Center Working Paper Series.

Ruddick, S. (1989). Maternal thinking: Towards a politics of peace. Boston, MA: Beacon.

Silverman, D. (1993). Interpreting qualitative data: Methods for analysing talk, text and interaction. London: Sage.

Smith, D. (1987). The everyday world as problematic: A feminist sociology. Milton Keynes: Open Uiversity Press.

Sapsford, R., & Jupp, V. (1996). Data collection and analysis. London: Sage.

Stanley, L. and Wise, S. (1993). Breaking out again: Feminist consciousness and feminist research. London: Routledge and Kegan Paul.

Strauss, A. L. and Corbin, J. (1990). Basics of qualitative research: Grounded theory procedures and techniques. London: Sage.

Tesch, R. (1990). Qualitative research: Analysis types and software tools. New York: The Falmer Press.

Tolman, D. (1992). Voicing the body: A psychological study of adolescent girls' sexual desire. (Doctoral dissertation, Harvard University, 1992).

Tronto, J. (1995). Care as a basis for radical political judgements. (Symposium on care and justice) Hypatia 10, 141-49.

Tronto, J. (1993). Moral boundaries: A political argument for an ethic of care. New York and London: Routledge.

Tronto, J. (1989). Women and caring: What can feminists learn about morality from caring?  In A. M. Jaggar & S. R. Bordo (Eds.), Gender/body/knowledge: Feminist reconstructions of being and knowing (pp. 172-187). New Brunswick and London: Rutgers University Press.

Way, N. (1994). In their own words: Listening to inner-city adolescents speak their worlds. (Doctoral dissertation, Havard University, 1994).

Weitzman, E. A. & Miles, M. B. (1995). Computer programs for qualitative data analysis: A software sourcebook. London: Sage.

Wolcott, H. F. (1994). Transforming qualitative data: Description, analysis and interpretation. Newbury Park, CA: Sage

Wolf, D.L. (1996). Situating feminist dilemmas in fieldwork. In D.L. Wolf (Ed.), Feminist Dilemmas in Fieldwork (pp.1-55). Boulder, Colorado: Westview Press.