Can We Re-Use Qualitative Data Via Secondary Analysis? Notes on Some Terminological and Substantive Issues
by Martyn Hammersley
The Open University
Sociological Research Online 15(1)5
<http://www.socresonline.org.uk/15/1/5.html>
doi:10.5153/sro.2076
Received: 14 Jul 2009 Accepted: 13 Dec 2009 Published: 28 Feb 2010
Abstract
The potential gains and practical problems associated with secondary analysis of qualitative data have received increasing attention in recent years. The discussions display conflicting attitudes, some commentators emphasising the difficulties while others emphasise the benefits. In a few recent contributions the distinctiveness of re-using data has come to be questioned, on the grounds that the problems identified with it - of data not fitting the research questions, and of relevant contextual knowledge being absent - are by no means limited to secondary analysis. There has also been a more fundamental claim: to the effect that these problems are much less severe once we recognise that, all data are constituted and re-constituted within the research process. In this article I examine these arguments, concluding that while they have much to commend them, they do not dissolve the problems of 'fit' and 'context'.
Keywords: Re-Use of Qualitative Data, Secondary Analysis, Qualitative Data Archiving, Constructionism
Introduction
1.1 There has been considerable recent discussion about the ‘re-use’ of qualitative data available in archives, such as those accessible through Qualidata, and of ‘secondary analysis of qualitative data’ more generally (see, for example, Mauthner, Parry and Backett-Milburn 1998; Fielding and Fielding 2000; Heaton 2004; Parry and Mauthner 2004; Gillies and Edwards 2005; van den Berg 2005; Bishop 2006; Moore 2006; Blaxter 2007; Mason 2007; Moore 2007; Savage 2007; Silva 2007).[1] Quite a lot of this discussion has been concerned with problems associated with re-use, for example that there will often be a lack of fit between the data available and the requirements of a second analysis, and that lack of access to the original contexts in which and about which the data were generated will make them difficult to interpret or will lead to error (Mauthner et al 1998; Parry and Mauthner 2004; Gillies and Edwards 2005). However, some recent contributions (notably Moore 2006 and 2007) have questioned the existence or severity of these problems by challenging the concepts of ‘re-use’ and ‘secondary analysis’, and this is the stimulus for my discussion here.1.2 Part of this argument is that the problems with re-use of data highlighted by the critics are not restricted to it: they also occur in some kinds of ‘primary’ research. In particular, what is involved in the ‘re-use’ of archived data has parallels to some forms of analysis that are normally seen as ‘primary’, for example that which occurs in research teams where one member analyses data collected by others, in historical work relying on documents that may have been used previously by other historians, or in conversation analysis, where it is common to analyse data collected by others. On top of this, Moore (2007) suggests that the notions of re-use and secondary analysis involve a false conception of data because they fail to recognise that data are reflexively constructed within research processes rather than existing independently of these. From this point of view, data cannot be first collected and analysed and then ‘re-used’ by other researchers for the purposes of ‘secondary’ analysis. Indeed, data cannot even be ‘collected’ in the first place because they are always constructed, as Bateson (1984) pointed out long ago in the context of survey research. The conclusion drawn is that it is possible, and desirable, to use material that other researchers have generated; and that the process of analysis here is no different in epistemic status from that in primary research, because the data are necessarily constituted, contextualised and recontextualised within any project. As a result, the problems of ‘fit’ and ‘context’ are no more likely to arise in research using data from an earlier study than they are in one where ‘new’ data are produced.
1.3 What is at issue here is partly a terminological point – about the appropriateness of the phrases ‘re-use of qualitative data’ and ‘secondary analysis’, and their analogues – but the arguments outlined above also present a challenge to any idea that what these phrases refer to is a relatively weak and/or problematic form of research by comparison with ‘first-time use’ of data in ‘primary analysis’. More specifically, criticisms of the ‘re-use’ of qualitative data in terms of ‘fit’ and the role of ‘context’ are questioned. In this article, I want to examine both the terminological and the more substantive issues. However, it seems to me that Moore’s argument about the way in which data are reflexively constructed in all research is a more fundamental one than the others, and for the moment I will leave it on one side.
The issue of terminology
2.1 A first issue is whether the use of data produced by a previous research project amounts to a sufficiently distinctive kind of work as to require its own label, such as the ‘re-use’, ‘reworking’, or ‘secondary analysis’ of data. It is certainly true that there is no absolutely clear dividing line between use and re-use, and that there is some inconsistency of usage. However, despite this, in my view there is some value in these labels, and they could be applied in relatively consistent ways.2.2 In order to draw a distinction between use and re-use we need to be able to differentiate between studies where the data being employed have been produced by previous researchers and those where they have not. It is worth examining the three sorts of problematic case mentioned above in order to see how these should be classified. Data analysis within a team, where one researcher analyses data collected by another, is probably best seen as use rather than re-use of research data, because members of the team are all working on the same project. Historians analysing documents should also be treated as engaged in primary research, since those documents will not usually have been produced by other historians or for the purposes of historiography. By contrast, conversation analysts using audio-recordings and/or transcripts that someone else has produced, and analysed, seems to be a clear case of the re-use of research data; though, where the data have not previously been analysed, we seem to have a genuine borderline case between use and re-use.[2]
2.3 Of course, we must also recognise that ‘re-using research data’ is not a homogeneous category. As has been noted in the literature, the term ‘re-use’ has been employed to cover a range of rather different enterprises: re-analysing the data from a project in order to assess the validity of its findings versus using the data from a previous project to address new research questions; using data from a previous project in which one was involved versus using data from someone else’s research; relying solely on the data from a previous project versus combining this with data from other sources. However, this heterogeneity does not undercut the value of having a phrase to cover the situation where previously produced research data are used again; it simply indicates that the term must be used with clarity about what is involved on any occasion when such variations are relevant.
2.4 It is also necessary to note that, in order to find the use/re-use distinction of value, we do not need to assume that the secondary analysis automatically involves problems that primary analysis always avoids. It would be sufficient that there is a tendency for certain sorts of problems to be more likely to arise in re-using existing research data than in projects where at least some of the data are specifically generated by the researchers to answer their own research questions. And I will argue that this is true both as regards the problem of lack of fit between data and research questions, and the issue of sufficiency of contextual knowledge.
2.5 So, terminologically speaking, the phrases ‘re-using’ or ‘reworking’ data do not seem to be so problematic as to be without value. They mark a roughly defined, albeit not entirely clear-cut, distinction that may be of relevance on many occasions. And while the division does not mark out some set of methodological problems that only occur on one side and never on the other, it does, I suspect, pick up a difference in the likelihood of particular sorts of problem on each side. This is sufficient warrant for use of these terms, I would have thought, given that many of the other conceptual distinctions we use only pick up tendencies rather than absolute differences.
More substantive matters
3.1 While these terminological issues are of some significance, they do not get to the core of what is involved in the debates that have taken place about the re-use of qualitative data. As already noted, one of the substantive points often put forward is that the two main criticisms made by those who are sceptical about the possibility and/or value of re-using qualitative data do not apply uniquely to this practice; they are found elsewhere too, in forms of research whose viability and value is generally accepted.3.2 Take, first, the problem of lack of fit. This can arise in any research project: it is by no means always possible to obtain all the data one needs. However, a gap is especially likely to arise in some kinds of work, for instance historical investigations that must rely upon documents from the remote past. What is fundamental here, of course, is a distinction between projects in which the researcher observes relevant situations and/or interviews relevant people in order to try to generate the data required to answer a set of research questions, and studies where the researcher can only use whatever data are already available in order to address those questions, whether this is historical work using extant documents or someone investigating a contemporary group or situation where there is little or no access to data via observation or interviewing (for example, current British Cabinet meetings). Clearly, the re-use of research data is closer in character to the second situation than to the first.
3.3 It is certainly true that the problem of lack of ‘fit’ does not disqualify re-use of research data, any more than it does other kinds of research that must rely exclusively on extant documents. This is partly because it is a matter of degree: the problem will be more or less serious in different cases. While re-using data does, in principle, restrict what questions can be addressed, compared with collecting new data for that purpose, this does not make all re-use of data especially problematic; even less does it make it impossible. However, note that this leaves open the likelihood that the problem may be intractable on some occasions. Indeed, there will be questions that are simply not open to investigation on the basis of already available data. In short, the problem of fit is more likely to arise with re-use of research data, by comparison with research where it is possible to generate new data, but this problem is not exclusive to that situation and it is not always a serious obstacle, but it will sometimes be.
3.4 The same sort of argument applies to the other criticism made of the re-use of qualitative data: the relative lack of contextual knowledge. Here, again, the problem is not unique to this sort of work, and is a matter of degree. There can certainly be significant differences between what is available as data to the initial researcher and what is accessible to a later researcher who re-works the data, whether for a similar or for a very different purpose. For one thing, in the process of data collection researchers generate not only what are written down as data but also implicit understandings and memories of what they have seen, heard, and felt, during the data collection process. For instance, not everything that is experienced in the course of participant observation or in-depth interviewing will be or can be written down. Nevertheless, it will often be drawn on tacitly, and perhaps sometimes consciously, in the course of analysis, perhaps playing an important role in making sense of the data that have been recorded (see Weaver and Atkinson 1994:3-4; Mauthner and Parry 1998; Gillies and Edwards 2005; but see also van den Berg 2005). What some have referred to as ‘head-notes’ (Ottenberg 1990) are generally unavailable to someone who did not carry out the data collection in the initial study, and to one degree or another to the same person seeking to re-use her or his own data after many years. Moreover, sometimes not even all of the recorded data will be included in archives, and the unavailability of what is missing to anyone seeking to re-use the data will not just be a matter of the absence of some information, once again it may affect how the data that are available are interpreted. Also unavailable, usually, will be the detailed knowledge about how the data were collected, and about how the analysis was carried out and written up, that the original researchers possessed; and this could also be of significance for what inferences should and should not be drawn from those data. It may be possible to provide some of this information in archives, but even aside from the problem that much of it is tacit or experiential, there are constraints arising from minimising the burden placed upon depositors (Bishop 2006).
3.5 Of course, researchers engaged in ‘primary’ analysis are sometimes faced with similar problems, for example when working as part of research teams or when using documentary data. Furthermore, it should not be assumed that those engaged in generating new data are in immediate contact with reality, even as regards the data collection process.[3] Mediation, of some kind, is always involved. Even where we are able actively to generate data, for instance through going to observe in relevant situations or to interview informants, we will often not know all that we later decide we needed to know about the contexts in which the data were produced in order to interpret them soundly. A good example here would be the complexities introduced into analysing interview data once we recognise the extent to which these have been ‘co-constructed’ or situationally generated. Having been a participant in an interview does not mean that one has an exhaustive knowledge of what occurred. So, the knowledge of context involved here is a matter of degree, and the problem may be no more serious in some cases of secondary analysis than it is in some forms of primary analysis. How serious it is depends upon purpose and circumstance. The example of conversation analysis is illuminating here: if we are studying turn-taking procedures such background knowledge may be of little significance. However, for other purposes it may be of much greater import. Furthermore, in some instances of secondary analysis the problem will be very severe: the data available may be ambiguous or obscure in ways that could only have been resolved if we had had more knowledge of the contexts investigated and of the research process.
3.6 We should not, of course, assume that what a primary researcher ‘knows’ and ‘understands’, in terms of context, is always right. Interpretation is always required, always depends upon assumptions, and can therefore go astray. Furthermore, this researcher will never be in receipt of all that could have been relevant as evidence for interpreting what is going on and why. In other words, what is produced and experienced in the course of data collection is always, necessarily, made sense of; it is not a matter of the reproduction of reality. And it is made sense of from a particular, situated location, reflecting who the researcher is, her or his relations with the people being studied, from what physical location the observation took place, how the interview unfolded as a piece of social interaction, etc.[4] Similarly, when we read data produced by others we ‘read in’ some sense of what the situation must have been like, what must have been happening, who the people were, and so on, drawing on our cultural knowledge. The readings generated by a secondary analyst may well be true in many respects, but there are more constraints on our interpretations of data when we have been involved ourselves in the situation in which the data were collected than is usually the case when we are analysing data produced by others. And this may, but need not always, affect the validity of our research findings. So, once again, the problem is not unique to re-using data, nor is it necessarily severe, but it is more likely to occur in that kind of work than in what are currently dominant forms of qualitative research, and it can be very difficult and sometimes impossible to handle.
The construction of data?
4.1 Finally, I want to consider a more fundamental challenge to the distinction between use and re-use of data. This is to the effect that data do not exist independently of the research process, so that it is never a matter of re-using existing data: one is always constituting the data for the first time in any research project (Moore 2007). There are elements of truth in this argument, but I think it is false overall.4.2 Part of the trouble here derives from ambiguity in the meaning of the word ‘data’, a term that is routinely used but rarely defined. If we think of data as what is used to serve as grounds for inference, there is certainly an important sense in which data are constructed through the research process. What serves as grounds for inference, and inference to which conclusions, depends upon what research question(s) are being addressed and how the analysis develops. In other words, there are important respects in which what count as data, in this sense of the term, and what the data mean, are determined functionally within the research process. One way of formulating this is to recognise that data are signs, and that by their very nature the meanings of signs are given by what they are taken to be signs of, by what conclusions they are used to reach. From this point of view, it may seem that data cannot exist independently of the research process, whether this is ‘primary’ or ‘secondary’: they are formed by the processes in which they are given meaning through being used as a basis for drawing inferences.
4.3 However, there is a counter-intuitive quality to this argument that ought to trouble us. After all, surely we do not and should not make up our data? I take this to be true whether or not one is a realist or even a ‘latent positivist’ (Mauthner et al 1998:736 and 743; Moore 2006:11; Moore 2007:3.3), though there are no doubt some who would challenge the point.[5] What this means is that the data must in some ways constrain what inferences we make and the conclusions we reach, rather than being freely constructed in and through our inferences. And this implies that they must, in some sense, exist prior to and independently of the research process.
4.4 This relates to a different meaning that can be, and almost inevitably is, accorded to the term ‘data’, which implies that what is used as data is somehow ‘given’ rather than created by the researcher. Moreover, this applies even to those forms of data production, like writing fieldnotes and transcribing audio-recordings, where the metaphor of ‘collecting’ or ‘gathering’ data seems especially problematic, by contrast with the case of using written documents or artifacts. In producing observational or interview data, in effect, we try to write down what is/was happening in a scene, in certain respects. We do this, in part, by recording the words used, in the actual sequence in which they occurred, but more fundamentally by generating descriptions of the actions and events in which they were involved (Hammersley 2010). There can, of course, be ambiguities here. To draw on a famous example, we may transcribe the words ‘Let him have it’, but then have to decide what action these represent: an incitement to murder or a call to give up the booty. Nevertheless we must try to represent what actually happened. My point here is not that we can engage in this process passively, without making selections and interpretations, but rather that the selections and interpretations we make are always from/of what G. H. Mead referred to as ‘the world that is there’ (see Miller 1973:42-3 and passim), and that in making them we try to represent relevant parts of this world. The sense in which the data are ‘given’ is even more obvious where we are using documents that have been produced by others and whose existence pre-dated the research. Here, the metaphor of ‘collecting’ or ‘gathering’ data approaches being literally true. We treat these documents as independent of us even though we necessarily interpret them in trying to understand what was being said and done, so as to infer answers to our research questions. Contrary to the excesses of some reader response theory, reading and writing are very different activities, we do not produce a text when reading in the same way as when we write it (see Eco 1994 and 2008). Moreover, as I have already emphasised, when we are writing fieldnotes, transcripts, or research reports, by contrast with producing short stories or novels, we are nevertheless seeking to document what happened and why in some situation.
4.5 A related point is that in much qualitative inquiry there is a considerable difference between what is initially searched for, selected, and interpreted as data that are potentially relevant to the research and the later use of this material as a basis for inference to conclusions in the research report. In other words, the research questions, ideas about what would be relevant in answering them, judgements about what can and cannot be legitimately inferred from what, and so on, are all likely to change, to one degree or another, over the course of the research process. This certainly means that what constitute ‘the data’ will change, but it does not mean that they are entirely malleable, that we can (or should) make of them whatever serves our purposes. They constrain our analysis in at least two respects. First, there will often be gaps in what we come to judge to be the body of data necessary to confirm or challenge our conclusions. Secondly, at various points, there will be data already produced, and data that could be produced, that will count against some of our inferences from other data, generating interpretative problems of various kinds, and perhaps requiring us to abandon our earlier hypotheses.
4.6 I suggest that these two meanings of ‘data’, as constructed and as given, are both essential: they relate to different but equally important aspects of the material we use as grounds for inference in research. One way of marking the distinction would be to restrict the term ‘data’ to what is collected or generated as a resource, in the sense indicated above, while using the word ‘evidence’ to refer to what is eventually used as grounds for inference to research conclusions in publications. In these terms, we collect data as a resource and then use some of it, in particular ways, as evidence, in order to draw inferences relevant to our research focus; and we discover how to do this in the course of our work. So, in using data, we necessarily reconstitute it as evidence.
4.7 It is important to stress, however, that this is not an absolute distinction between what is completely independent of us in its character (data) and what is entirely constructed ab initio (evidence). Rather, these terms must be seen as labelling phases – or, perhaps it would be better to say, identifying two temporally overlapping faces – of a single process. In the early stages of research we 'hail' some noumena as phenomena that might be relevant to our research, or we may even work to produce them, but we are nevertheless constrained in what we can identify as data by the noumena that are available (as well as by our current conceptions of relevance). While this initial process may occur against the background of our research questions, to a large extent it will be more a matter of finding materials that are intelligible or 'readable' as relevant to our work. Later, we will use some of these materials much more selectively and deliberately as data to generate evidence, re-forming it again in the process because we are putting it into the context of arguments designed to answer our research questions. But here, as before, we are re-forming something that already exists, not making it up.[6]
4.8 So, when, for example, we are transcribing an audio-recording, whether from scratch or re-working an initial transcript produced by someone-else, we try to write down the words that were used, albeit inevitably doing so as part of some understanding of what those words are being used to mean and do. We are not listening to the audio-recording for evidence for or against some descriptive or explanatory hypothesis, or even primarily so as to generate theoretical ideas, although if any come along we will no doubt want to note them down. By contrast, when beginning the analysis of data produced in this way, we are primarily concerned with using those data as evidence relevant to the research focus. This may involve both inferring plausible descriptions or explanations from the data as well as using them to check existing descriptions and explanations, and perhaps also to interrogate some of our own starting assumptions. By the time of writing the research report, the concern will be entirely with evidence, in the sense of that term introduced above; and specifically with what sorts and amount of evidence need to be provided to satisfy and convince the target audience.
4.9 Data are, then, in an important sense given as well as constructed: they are not created out of nothing in the research process, nor should we construct whatever inferences we wish to on the basis of them. At the same time, it is important to recognise that they are also constructed or produced in the course of research, and to be aware of aspects of this process that could be relevant to what would and would not be legitimate inferences from them. Generally speaking, we will know more about the construction process where we actively produce our own data than when we rely upon the data collected in another research project in which we were not involved, or even one that we were involved in but that took place a long time previously. While this is a matter of degree, and it is not a difference that is restricted to the contrast between use and re-use of data, nevertheless it is an important one that can create problems. It is not dissolved by the fact that all data are inevitably re-contextualised within research processes so as to turn them into evidence.
Conclusion
5.1 In this article I have examined some of the issues that have arisen in recent discussions of the ‘re-use’ or ‘secondary analysis’ of qualitative data. As regards terminology, I suggested that there is nothing illegitimate about these phrases, and that they can be of value. Furthermore, I argued that it is important to recognise that re-using existing research data can involve problems of ‘fit’ and ‘context’; even though it is equally important to be aware that these problems may also arise in ‘primary’ research and analysis. Indeed, it is rare for these problems to be entirely absent in any kind of research. What are involved are matters of degree, and their significance depends upon purpose and circumstance. At the same time, if we compare re-using research data with those forms of qualitative research that are currently dominant in many fields, involving a researcher setting out to collect data her or himself, both these types of problem are likely to be more severe in the former than the latter.5.2 It has been suggested that, once we recognise the constructed character of data, these apparently obvious points no longer hold: if the data for any project are constituted in and through the research process, there is no significant difference between ‘primary’ and ‘secondary’ analysis. However, this is not the case. It is true that data are constructed by researchers. What is and is not relevant evidence, what it means, what should be taken as strong and what as weak evidence, and so on, depends upon the research focus, the data available, and the line of argument that develops in the course of the research. However, if we think of data as resources that we subsequently come to use as evidence, then there is also an important sense in which data exist independently of the research project. All construction uses materials and resources that are in some sense given rather than created by those doing the construction work. The same point emerges if we think of data as consisting of signs: the sign-vehicle, and (if the inferences we draw are correct) also what they point to, exist independently of the research process.
5.3 This conclusion does not involve commitment to the idea that data are simply given, and can be interpreted by a ‘primary’ researcher against the immediacy of the data collection process in such a way that they automatically stand for reality. Selection and interpretation are always involved, and interpretations are underpinned by assumptions that could be wrong. However, this does not undercut the important role that knowledge about how the data were collected, the situations and people involved, etc, can play. For some purposes, ‘missing data’ can be crucial, preventing certain questions being addressed; and attempts by ‘secondary’ analysts to fill in the gaps when interpreting the data that are available could lead to error. This problem is not unique to the re-use of research data, but it does frequently arise in that task.
5.4 I have tried to outline how the two aspects of data, as constructed and as given, change over the course of the research process, labelling this as a shift from a concern with data to a focus on evidence. What is involved here is that the framework in terms of which we treat data as given and through which we construct it alters as a result of the developing research process. Initially, we look for and take what seems relevant to our work and use it to create an initial sense of the phenomena we are interested in investigating, as yet perhaps not well-defined. Here, we treat data as given but also work them up into some intelligible picture. And this is an interactive, not a one-way process. Later, we are more concerned with turning these data into evidence that can enable us to answer our research questions. And, at the point of writing the research report, we are primarily interested in whether we have the evidence to support our conclusions, from the point of view of the target audience, and how reliable that evidence is.
5.5 The point of all this is that recognising that the data construction process involves working on what is taken as given makes clear that there can always be variation in the degree to which the data ‘fit’ the requirements, and in whether we have sufficient knowledge of ‘context’ in order to make sound interpretations. If there were no sense in which data existed independently of the research process in which they are used, then the problems of ‘fit’ and ‘context’ would not occur. Once we recognise that there must be an important sense in which data are given, we can see that those problems are likely to be more and less severe, depending upon the circumstances and purposes of the research. Moreover, we can also see that those problems will, very often, be more severe in the re-use of research data, by comparison with the original research project in which and for which they were generated.
5.6 So should we still talk of ‘re-using’ data and ‘secondary analysis’? On the basis of the arguments presented here, the notion of re-use is unobjectionable: it indicates that the data were originally produced as research evidence by someone and that we are re-using this evidence as data, albeit constituted within the context of a second (or third, or fourth) research project.[7] While the term ‘secondary analysis’ is problematic because the same data can be used more than twice, in my view this does not undermine its usefulness. As a number of writers have pointed out, the fact that data are being re-used does not signal the presence of problems that never arise in the context of primary research, nor does it automatically imply that ‘fit’ and ‘context’ are going to be impossibly troublesome. But it does suggest that these problems are especially likely to arise. How important or difficult to resolve they turn out to be will depend upon the nature of the data available, and on the purposes for which we are wanting to re-use it.
Notes
1There are also now many discussions of the experience of carrying out re-analyses of data, of various kinds. See, for example, Thompson 2000; Bornat 2005; Gillies and Edwards 2005; Savage 2005; Bishop 2007.2 Another example would be where a whole corpus of data are published so that they are available for others to analyse. For example, Znaniecki reported that the original intention behind The Polish Peasant in Europe and America was ‘to present a collection of human documents, with such comments and interpretations as would make their utilization by other scientists possible’ (Znaniecki 1939:91).
3 This occasionally seems to be assumed, see for instance Kynaston 2005:2.4.
4 I do not take this to mean that its truth is relative, only that it is fallible and that its relevance is perspectival.
5 For a discussion of some forms of qualitative research that might deny any distinction between ‘making’ and ‘making up’, see Hammersley 2008.
6 And, even here, there may be a further distinction to be drawn between the use of the data in abductive and in inductive inference; in other words, so as to generate plausible descriptive and explanatory ideas, or in order to test those ideas.
7 The parallel that Heaton (2004), following Glaser (1962 and 1963), draws between the re-use of quantitative and qualitative data is quite legitimate, it seems to me. There are likely to be some differences in what is involved, over and above the fact that the data are in numerical or verbal form; but there are, in any case, substantial differences among projects re-using quantitative data just as there will be among those re-using qualitative data.
References
BATESON, N. (1984) Data Construction in Social Surveys, London, Allen and Unwin.
BISHOP, L. (2006) ‘A proposal for archiving context for secondary analysis’, Methodological Innovations Online, 1, 2. Available at (accessed 23.06.09):
BISHOP, L. (2007) ‘ A reflexive account of re-using qualitative data: beyond primary/secondary dualism’, Sociological Research Online, 12, 3. Available at (accessed 30.05.09): <http://www.socresonline.org.uk/12/3/2.html>
BLAXTER, M. (2007) ‘Commentary on “A reflexive account of re-using qualitative data: beyond primary/secondary dualism” (Libby Bishop)’, Sociological Research Online 12, 3. Available at (accessed 30.05.09): <http://www.socresonline.org.uk/12/3/7.html>
BORNAT, J. (2005) ‘Recycling the evidence: Different approaches to the reanalysis of gerontological data’, , Forum: Qualitative Social Research, 6, 1, Art.42. Available at (accessed 7.12.09): <http://nbn-resolving.de/urn:nbn:de:0114-fqs0501424>
ECO, U. (1994) The Limits of Interpretation, Indiana University Press.
ECO, U. (2008) Interpretation and Overinterpretation, Cambridge, Cambridge University Press.
FIELDING, N. G. and Fielding, J. L. (2000) ‘Resistance and adaptation to criminal identity: using secondary analysis to evaluate classic studies of crime and deviance’, Sociology, 34, 4, pp671-89.
GILLIES, V. and Edwards, R. (2005) ‘Secondary analysis in exploring family and social change: Addressing the issue of context’, Forum: Qualitative Social Research, 6, 1, Art. 44. Available at (accessed 30.05.09): <http://www.qualitative-research.net/index.php/fqs/article/view/500/1077>
GLASER, B. G. (1962) 'Secondary analysis: a strategy for the use of knowledge from research elsewhere', Social Problems, 10, 1, pp70-4. [doi:10.1525/sp.1962.10.1.03a00080]
GLASER, B. G. (1963) 'Retreading research materials: the use of secondary analysis by the independent researcher', American Behavioral Scientist, 6, 10, pp11-14.
HAMMERSLEY, M. (2008) Questioning Qualitative Inquiry, London, Sage.
HAMMERSLEY, M. (2010) ‘Reproducing or constructing? Some questions about transcription in social research’, forthcoming in Qualitative Research.
HEATON, J. (2004) Reworking Qualitative Data, London, Sage.
KYNASTON, D. (2005) ‘The uses of sociology for real-time history’, , Forum: Qualitative Social Research, 1, 3, Art.27. Available at (accessed 7.12.09): <http://nbn-resolving.de/urn:nbn:de:0114-fqs0501457>.
MASON, J. (2007) ‘“Re-using” qualitative data: on the merits of an investigative epistemology’, Sociological Research Online, 12, 3. Available at (accessed 30.05.09): <http://www.socresonline.org.uk/12/3/3.html>
MAUTHNER, N. S., Parry, O., and Backett-Milburn, K. (1998) ‘The data are out there, or are they? Implications for archiving and revisiting qualitative data’, Sociology, 32, 4, pp733-45. [doi:10.1177/0038038598032004006]
MILLER, D. L. (1973) George Herbert Mead: Self, language and the world, Chicago, University of Chicago Press.
MOORE, N. (2006) ‘The contexts of data: broadening perspectives in the (re)use of qualitative data’, Methodological Innovations Online, 1, 2. Available at (accessed 23.06.09):
MOORE, N. (2007) ‘(Re)using qualitative data’, Sociological Research Online, 12, 3. Available at (accessed 30.05.09): <http://www.socresonline.org.uk/12/3/1.html>
OTTENBERG, S. (1990) ‘Thirty years of fieldnotes: changing relationships to the text’, in Sanjek (ed.).
PARRY, O. and Mauthner N. S. (2005) ‘Back to basics: who re-uses qualitative data and why?’, Sociology, 39, 2, pp337-42. [doi:10.1177/0038038505050543]
SAVAGE, M. (2007) ‘Revisiting classic qualitative studies’, Forum: Qualitative Social Research, 6, 1, Art31. Available at (accessed 30.05.09): < http://www.qualitative-research.net/index.php/fqs/article/view/502/1081>
SILVA, E. B. (2007) ‘What’s [yet] to be seen? Re-using qualitative data’, Sociological Research Online, 12, 3. Available at (accessed 30.05.09): <http://www.socresonline.org.uk/12/3/4.html>
THOMPSON, P. (2000) ‘Re-using qualitative research data: a personal account’, Forum: Qualitative Social Research, 1, 3, Art.27. Available at (accessed 7.12.09): <http://nbn-resolving.de/urn:nbn:de:0114-fqs0003277>
VAN DEN BERG, H. (2005) ‘Reanalyzing qualitative interviews from different angles: the risk of decontextualization and other problems of sharing qualitative data’, Forum: Qualitative Social Research, 6, 1, Art 30. Available at (accessed 30.05.09): <http://www.qualitative-research.net/index.php/fqs/article/view/499/1075>
WEAVER, A. and Atkinson, P. (1994) Microcomputing and Qualitative Data Analysis, Aldershot, Avebury.
ZNANIECKI, F. (1939) ‘Comment by Florian Znaniecki’, in Blumer, H. An Appraisal of Thomas and Znaniecki’s The Polish Peasant in Europe and America, New York, Social Science Research Council.
