1 Interview 'Facts' as Evidence to Support Inferences to Eventual Theorization/ Representation Models DATA-COLLECTION (UP TO AND INCLUDING THE INTERVIEW) AND DATA-INTERPRETATION (FROM INTERVIEW TO RESEARCHER PRODUCT) In social research methodology we typically distinguish the collection of data from the interpretation of data. How these are to be related can then be seen in two different ways: by a common-sense hypothetico-inductivist model and by a hypothetico-deductivist model. Common-sense Hypothetico-inductivist Model In a fairly common-sensical model, the researcher collects 'all the relevant facts' and then examines them to see what theory is suggested by this set of 'all the relevant facts'. The theory thus 'emerges' from the data. This is the original 'grounded theory' tradition (Glaser and Strauss, 1968) in which theory emerges by a process of 'induction'. The facts are believed to suggest - or even 'require' or 'dictate' - the theorization. Anti-common-sense Hypothetico-deductivist Model The counter-model is anti-inductivist. It declares that there is no thing as 'all the relevant facts', there are only 'hypothesis-relevant facts', and that research must always start with a body of prior theory, if only to decide which set of 'collectable facts' should be collected or generated. It is this prior body of theory from which the researcher generates a particular hypothesis whose truth or falsity could be 'tested' by a particular selection of 'hypothesis-relevant facts'. The hypothesis-relevant facts are then collected, and the hypothesis is either supported by the evidence of those facts or it is refuted by them. Inductivist and Deductivist 'Moments' in Doing Research Both the 'inductivist' and the 'deductivist' models correspond to styles of doing research; both have philosophical flaws which enable them to be criticized for EVIDENCE TO SUPPORT INFERENCES TO MODELS 3 ever. In order to have survived their innumerable critics for so long (see, for example, Layder, 1998) each model must correspond to some real experience of researchers. I would argue that both are appropriate as descriptions of what researchers experience as happening at different moments of the research cycle and that another relationship between the inductivist and deductivist models may be that of level. I would regard myself as strategically being largely a 'deductivist' or a the-oreticist while fully appreciating the need for particular moments of inductive working. For many of the purposes for which I do research, my general 'deductivist' strategic theory dictates 'giving up control' at a tactical level to the person being interviewed. For other purposes, my same theoretical model - which will be elaborated later - dictates 'control being taken back' by the interviewer. However, other researchers may be strategically inductivist while at certain moments using deductivist tactics. To understand why one can feel both an inductivist and a deductivist at different times or even in the same moment of research, I find the concepts of different levels, or of the difference between strategy and tactics, to be helpful. Gregory Bateson's argument for a combination of 'loose' and 'tight' thinking is similar. SOME FEATURES OF DEPTH INTERVIEWING AS DESIGNED PRACTICE If these are features of the approach being taken, what are the main points that I wish to stress about the style of interview that I am talking about? • The interview is a research interview, designed for the purpose of improving knowledge. • It is a special type of conversational interaction: in some ways it is like other conversations, but it has special features which need to be understood. • It has to be planned and prepared for like other forms of research activity but what is planned is a deliberate half-scripted or quarter-scripted interview: its questions are only partially prepared in advance (semi-structured) and will therefore be largely improvised by you as interviewer. But only largely: the interview as a whole is a joint production, a co-production, by you and your interviewee. • It is to go into matters 'in depth'. I shall deal with these briefly in turn. The Interview is a Research Interview The 'form' of an interview - questioning by one person, answering by another -can be used for a variety of purposes (Dillon, 1990: 2, and below, p. 154). Teachers use questions in order to help students learn; psychotherapists use questions to heal; secret police use brainwashing questioning to break down the interviewee's grip on reality. In future, when I refer to 'interviewing' in this book - unless it is clear from context that I am using it differently -1 shall mean 'research interviewing'. To me - and in this book - scientific research has to do with 'getting a better understanding of 4 CONCEPTS AND APPROACHES TO DEPTH INTERVIEWING reality'.1 The ethics of the research interview are that, at minimum, the informant should not be changed for the worse: against certain objections, I maintain that the research interview is not designed to 'help' or 'empower', or 'change' the informant at all. In my interviews, I collect information with the purpose of 1 developing/constructing a 'model' of some aspect of reality that I hope will be found to be in accordance with 'the facts' about that reality, or 2 testing a constructed model to see whether it is confirmed or falsified by 'the facts', and, more usually, 3 doing both the above. I shall argue in Part II that the 'semi-structured depth interview' normally involves the interviewer in a process of both model-building and model-testing, both theory-construction and theory-verification, within the same session or series of sessions.21 shall explore further the considerable difficulties that arise with the concept of 'the facts', the concept of 'being in accordance with the interview facts' so easily (but only provisionally) assumed in statements 1 and 2 in the list above. The Interview is a Type of Conversational Face-to-face Interaction Psychologists, social psychologists, sociologists, anthropologists all study human interaction in general and face-to-face interaction in particular. People work together, live together, make love together: there are co-operative aspects to (interview) interactions, and conflictual ones. Interactions can be between people of similar statuses or those of different statuses. They can be ritualized, heavily formatted by custom, or they can be more exploratory, uncertain and potentially innovative. They can leave people feeling positive, negative, or nothing. They can involve expensive material resources, or none at all. They can be remembered, they can be forgotten. They can be 'typical of the time and place and of their type', they can be 'different'. When you come to study and plan interviewing, bear fully in mind all the knowledge derived from your discipline (sociology, psychology, cultural studies, history, etc.) about face-to-face interaction and about the specificity of the society and the setting and the types of people involved, especially yourself. Don't put on a set of blinkers marked 'research methodology' to exclude other considerations! The interviews that you do or that you study are not asocial, ahistorical, events. You do not leave behind your anxieties, your hopes, your blindspots, your prejudices, your class, race or gender, your location in global social structure, your age and historical positions, your emotions, your past and your sense of possible 1 This philosophical position has been called a 'realist' one, and there are other philosophies: for example, those that hold that there is no such thing as a 'reality' out there but only an inter-subjective agreement to pretend to the existence of such an independent-seeming reality. Strategically I work with the axiom that there is a historically occurring reality out there; tactically, methodologically, we should always suspect that our most recent account is a fiction requiring further rectification. 2 It also is liable to involve tlie informant in such a process as well as they struggle to make sense of the interviewee's responses and non-responses and how to respond and not-respond to the interviewer. The interviewer is as engaged in hypothesis-formation and theory-rectification as you are. EVIDENCE TO SUPPORT INFERENCES TO MODELS 5 futures when you set up an interview, and nor does your interviewee when he or she agrees to an interview and you both come nervously into the same room. Nor do you do so when you sit down to analyse the material you have produced. It has to be Particularly Well-prepared (Designed) to Allow it to be Semi-Structured Semi-structured interviews are designed to have a number of interviewer questions prepared in advance but such prepared questions are designed to be sufficiently open that the subsequent questions of the interviewer cannot be planned in advance but must be improvised in a careful and theorized way. As regards such semi-structured interviews, they are ones where research and planning produce a session in which most of the informant's responses can't be predicted in advance and where you as interviewer therefore have to improvise probably half - and maybe 80% or more - of your responses to what they say in response to your initial prepared question or questions. In particular, I am concerned with semi-structured interviews where the interviewee is asked to tell a story, produce a narrative of some sort regarding all or part of their own life-experience. These biographic-narrative interviews are of considerable interest in their own right, and they also illustrate rather well more general principles of semi-structured interviewing. Very often, semi-structured interviewing is seen as 'easier' in some not very clear way. Novice researchers often feel that, with interviews that are only semi-structured, they do not have to do as much preparation, they do not have to work each question out in advance. This is a terrible mistake. Semi-structured interviews are not 'easier' to prepare and implement than fully structured interviews; they might be seen as more difficult. They are semi-structured, but they must be fully planned and prepared. Improvisation requires more training and more mental preparation before each interview than simply delivering lines prepared and rote-learned in advance. Compared with fully structured interviews, semi-structured interviews to be successful require • as much preparation before the session, probably, and certainly • more discipline and more creativity in the session, and certainly • more time for analysis and interpretation after the session. Given an equivalent amount of time and money, you can 'do' (prepare, do and analyse) far fewer semi-structured interviews than you can do fully structured ones. They may yield much more than fully structured ones can, under the right conditions. Under the wrong conditions, they may yield nothing at all. They are high-preparation, high-risk, high-gain, and high-analysis operations. Thinking clearly about the right conditions for being successful in such skilled interviewing and analysis is what this book is about. It is an Interview 'in Depth' What is meant by 'depth' will be discussed in more detail in Part II (see, however, Rorty, 2000 for a recasting of the concept of 'depth' into one of 'width'). To attempt to 'formalize' the concept and give it a bogus rigidity and precision at this point would not be helpful. There are two meanings for 'depth' which are useful to distinguish. 6 CONCEPTS AND APPROACHES TO DEPTH INTERVIEWING 1 To go into something 'in depth' is to get a more detailed knowledge about it. 2 To go into something in depth is to get a sense of how the apparently straightforward is actually more complicated, of how the 'surface appearances' may be quite misleading about 'depth realities'. Perhaps this second is more in tune with the approach in this book. To go into something in depth means to get a deep understanding of how little you knew about it, and how provisional one's 'formulations of truth' have to be - even by, or about, depth-interviewing. Let us try to see how these generalities might work in a particular analysis. THE 'HARD AND ONLY FACTS' OF INTERVIEW INTERACTION, AND INFERENCES As said above, for the purposes of this section, I have assumed the relatively inductivist position that, if properly recorded by tape-recorder and properly transcribed, the words spoken in the interview are relatively non-controversial 'facts'.3 From the words of the interview, what sort of inferences to extra-interview realities can I make? More importantly: how do I make and question such inferences? Old Wu: Knowledge of Discourse, Referents, Subjectivity What follows is an extract from an interview conducted with an elderly Chinese lady. I have chosen this text partly because most readers of this book will, like me, be unfamiliar with the cultural context. At this point in the interview, she is talking about her experience of the Maoist period in China. 'The Anti-Rightist campaign started. My ex-husband was accused of vilifying the Party and was labelled a rightist. In China, the Communist Party exercises leadership in everything. My ex-husband was jailed for being a counter-revolutionary and a suspected spy for speaking out and having overseas relatives. For the future of the family, and especially for the children, he thought it would be better that I divorced him. The Party secretary from my workplace had discussed my case at a mass-meeting and suggested that I get a divorce. He told me that love was not just for love's sake; love must have a political basis. My husband was an enemy of the people and did not deserve my love; if I didn't divorce him, then I supported the enemy. The Party secretary told me not to be afraid, as the masses would support and help me to divorce my husband [a doctor at a hospital]. It seems inconceivable and ridiculous to young people nowadays, but this kind of propaganda beguiled our generation for several decades. I thought my children would suffer for ever if I did not divorce him. Furthermore, because my husband was in prison, he could not get any salary. I was a nurse and did not earn much. In the end, I decided to get a divorce and I married another man who was a worker. continued 3 Later on, when dealing with the task of transcription (p. 212-23 onwards) this model will be moderately questioned; here it is used as a basis for further argument. EVIDENCE TO SUPPORT INFERENCES TO MODELS 7 continued My son was only a teen-ager and was strongly against my divorce. After five years of prison life, my ex-husband died of some illness in prison. When my son heard the news of his death, he could not control himself and went mad. It was a great tragedy for my family. I am very fond of my son, and his mental illness has led me to study spirituality ever since. My ex-husband was not pardoned until 1981. I am not fully satisfied with my second marriage. I don't think I have enough courage to divorce again. Divorce belongs to the young. I am too old to divorce, I do not want to catch the vogue. I love my children more than myself. That is the great love of Chinese women.' (Farmer, 1993: 91-2) Let us assume that this transcript 'is' or contains the data of the interview - as we shall see later, it is not 'raw' but 'processed' data, not only because it is a translation but also because it is a transcription - and consider what inferences might be made. What can we learn from this piece of data, this datum? At the moment, and only for the moment, I shall set aside the interview interaction as an 'object of research' in its own right (reserving this for Chapter 3). Just for the moment, I shall assume we are only interested in it as a means to study other things. What extra-interview objects of study can this material be used for studying? There is no end to the list of possible 'objects of study' for which I might wish to gather information by way of depth interviews. For convenience's sake, I shall create three categories, and then apply them to the Old Wu text. We can use our interviews to know more about 1 discourse 2 objective referents 3 subjectivity. Discourse The discourse is the mode of talk spontaneously chosen by the subject. A 'linguistic performance' occurs (carefully registered by the tape-recorder) and this can be analysed in a number of ways. One mode of analysis, that associated with the work of Foucault and of Chomsky, is where you attempt to identify a 'deep structure' which underlies or (as some would argue) generates the 'surface performance' of the things actually said. This is like a system of rules that creates 'patterned productions' of things likely to be said ('sayables') and things unlikely or impossible to be said ('unsayables') within that particular 'regime of discourse'. • For example, Hollway (1989:53-7) identified four discourses around sexuality: (i) the discourse of male sexual drive; (ii) the 'have-hold' discourse; (iii) the permissive discourse; and (iv) feminist discourse. She describes this identification of types of discourse as being 'tools to help in organizing ... accounts'. • For example, in economics, we can distinguish the 'neo-liberal discourse' promoting unregulated world free trade, from various 'social and ecological discourses' concerned to protect humans, animals and the environment from the ravages of the search for corporate profit. 8 CONCEPTS AND APPROACHES TO DEPTH INTERVIEWING • For example, Pierret (1993) has identified four different discourses about health and illness in France and compared them to the distribution of English ways of talking about health. In general, researchers attempt to identify the systemic nature of discourses which enable certain sorts of things to be said and make other sorts of things difficult to say. I might wish to explore the 'structure of discourse' underlying the flow of utterances of Old Wu in this segment of text. What are the organizing principles of her flow of talk? There are serious problems in exploring this, since all I do have as raw material is the mode of talk, the joint performance, in which two particular people engaged on the special occasion of this particular interview. In the Old Wu text, I do not have any idea as to what the interviewer said to prompt the flow of talk: the interviewer's contribution, their questioning, is absent. Was the original interview question about 'the great love of Chinese women' (the final sentence in the extract)? Was it 'What happened to your family in the Maoist period?' Was it 'How did Maoism work out in practice?' The significance of the interviewee's response cannot be gauged without understanding the implications of the questioning for the production of that response. I may wish to 'infer', from this co-produced joint mode of talk, what one of them (the interviewee; Old Wu, in our example above) would produce 'spontaneously' (on his or her own? to themselves in internal speech?) but this is an inference, not a fact. Some 'discourse analysts', as we shall see (for example, Potter and Weatherell, 1987), argue that all speaking (discoursing) is bound to the contexts and especially to the other people for whom it is produced, and so I can only say that somebody is capable of producing the 'discourse' that they did produce, that they have that particular discourse within their repertoire of discursive performances. They argue that I cannot be sure that one has a more privileged, or 'authentic', status - always more preferred - than any other. Nonetheless, whether I see people as having an 'authentic discourse' and a number of 'socially constrained' ones, or whether I think of all modes of talk as being equally constrained and equally (in)authentic, I can use interview material to learn more about discursive productions and performances (see e.g. Squire, 2000). However, if I wish to make inferences from the discourse in the interview to other realities, I can move in either or both of two directions. One direction is towards knowledge of that which is being talked about (the topic, the referent); the other is towards knowledge of the time-and-place located subjectivities who are doing the talking. We shall deal with these in turn. Objective Referents One such set of other realities may be that of the objective referents: those things that are referred to (hence 'referent', that to which reference is made) in the talk, or about which information can be gleaned 'through' the talk (excluding from the latter the 'subjectivity' of the individual informant, to be discussed later). See Figure 1.1. These are sometimes called the 'topics'. If a policeman interviews a witness about a scene in a supermarket that led to an act of violence, he is interested in the 'objective facts' to which the individual informant was a witness: who did what, said what, what happened next? ... The EVIDENCE TO SUPPORT INFERENCES TO MODELS 9 Utterances in Discourse Knowledge about the text -referents FIGURE 1.1 From Utterances, to Discourse, to Knowledge of Referents policeman is interested in the 'discourse' or the 'subjectivity' of the witnessing subject only to the extent that he must evaluate the reliability of the witness in order to make up his own mind about what inferences about 'what really happened' he is justified in drawing from the witness's statements. The individual subject is being asked to talk so that some information, not about him or her, but about a present or past 'context' which happens to be 'carried' by him or her, can be obtained. 'Interviewing for the facts', analysing interview material for 'factual content', for information about certain referents, is a very frequent, and obviously important, use of interview material. For an interesting discussion of knowledge-elicitation, see Firlej and Hellens (1993). In the case of the Old Wu extract, I can ask what knowledge about the relationships of Old Wu's family, her local Communist Party and its officials, the treatment of family members inside and outside the family, etc. can be gained from this account and what knowledge about such relations more generally can also be inferred. Subjectivity Or I may wish to interview and use interview material because of my interest in making inferences about subjectivity (Figure 1.2). Utterances in Discourse Knowledge about text -► subjectivity(ies) inter/subjectivity FIGURE 1.2 From Utterances, to Discourse, to Knowledge about Subjectivity(ies) Subjectivity is a term for a model I construct of what I see as some of the permanent or transient characteristics of the subject who is acting as informant in the interview. Our research may be concerned with them as a unique person, or as a 'representative' of a certain 'social type', or in some other way. Though our hypothetical policeman is only interested in subjectivity of the witness as a possible 'contaminating factor', I may be interested in the subjectivity of the interviewee as a clue to 'witness psychology', for example. In a more therapeutic situation, I might suggest that the interviewee talk about any number of subjects (e.g. what happened to them the day before) not because I am interested in that objective referent but because I am interested in how he or she talks about it, thus revealing their subjectivity. The topics themselves would be of subsidiary interest. What can I gather about Old Wu's (changing) subjectivity and the subjectivities of those inside and outside the family with whom she had a complex of voluntary or forced relationships? The 'subjectivity of Old Wu within the interview' may be different from that which she exhibits or possesses or is subject to in other practical and relational contexts of her life. The interplay between my and your subjectivity in an interview interaction is a matter for research. It can be argued (see e.g. Hollway, 1989; 10 CONCEPTS AND APPROACHES TO DEPTH INTERVIEWING Henriques et al., 1984) that subjectivity should be more fully understood as not just a predisposition within the individual but also as a complex interplay between people - for example, between the interviewer and the interviewee. I refer to this as being interested in the inter/subjectivity of the interaction. Inferring the Three Types of Knowledge from the Transcript Data Given a research focus on discourse, the referents, and subjectivity, what points can I infer about each from the raw data of the interview transcript? The 'Old Wu' 'case-narrative' is rich, and Figure 1.3 represents a 'first run through' of attempting to 'fill the columns, with appropriate material. Bear in mind, though, as the overlapping nature of the material suggests, the answers to these questions are inter-related, and that inter-relation will need to be explored. If I am to understand any one of the domains (discourse, the referent, subjectivity), I have to understand more about the other two, since all are imbricated in any one understanding. The assertions in Figure 1.3 should be regarded as hypotheses which are inferred from, and supported by, or at least not disconfirmed by, the transcript data. They are hypotheses about different types of knowledge that I as researcher might be interested in: knowledge about the way she talks about things, knowledge about the changing social and cultural contexts she has lived through, knowledge about Old Wu herself: discourse, referents, subjectivity. It may be useful to have a model of the relation of these three types of inferences from the interview interaction data to knowledge about discourse, about objective referents and about subjectivity see Figure 1.4.4 The diagram of how inferences can be made from utterances recorded in interview texts has certain implications: • It suggests that, as a researcher working towards a particular research project, I might be finally more interested in subjectivity - in which case, knowledge about the referents is a means to that end - or I might be more interested in the referents - in which case, in its turn, knowledge about subjectivity is regarded as the means to that other end. I might of course be spontaneously interested in both. Whether I am or whether I'm not, I am arguing that I must be interested in all the aspects in order to be relatively objective about any one of them. • It also - at the bottom of Figure 1.4 - points out the importance of our existing contextual expertise and misinformation. Such prior expertise and misinformation strongly influences the way that I make inferences (from left to right, in the diagram) from the hard facts of interview interaction to the relatively low level of inferences - but still inferences - about discourse used in the interaction, and then to the more assumption-rich inferences about subjectivity and about referents and the objective context. I refer to 'misinformation' rather polemically so as to help me keep in mind that 7 should expect myself to be carrying misinformation as well as information, prejudice as well as knowledge, and that 4 In Figure 1.4,1 have extended the concept of 'hard facts of interview interaction' to include nonverbal as well as verbal interaction, as well as the experience of the interviewer of that interaction as recorded on paper or tape by him or her in debriefing 'session notes' immediately after the interview session. EVIDENCE TO SUPPORT INFERENCES TO MODELS 11 Interview Discourse Objective referent Subjectivity interaction (evolving context) (recounting subject) Fluent in recounting CCP exercises Remembers past narrative of her 'leadership' in everything mode of decision- divorce and re- (now and at time of making ('this kind marriage divorce) and there is no of propaganda') and - sense in resisting it can contrast it with Generalizations about - current mode of self and husbands Workplaces had Party 'being less beguiled and children secretaries who discussed by propaganda' - 'cases' at 'mass meetings' - Fluent in the and gave strong advice Now very cynical 'politicized' speech about courses of action about 'propaganda' of Communist China - and 'catching the of her generation She divorced her husband vogue' - who died in prison, but - Discourse about was 'pardoned' in 1981 Is concerned to study 'love' of husbands, - 'spirituality' and of children, of She thinks her son 'went does so ever since chairman Mao, of mad' because of this - her son's mental self she has other children illness started - but no information about - Discourse about them except that she Love is very important generational loves them for her and she differences - loves her children - The 1950s marriage law of more than herself Discourse about free choice in marriage - 'self-control' and divorce 'became Is not fully satisfied ('my son could not hollow and meaningless' with her second control himself and - marriage went mad') Young people now see - - what was 'inconceivable Explains her non- Apparent general and ridiculous' in earlier divorcing by fluency of discourse: periods but can't 'lacking the no recorded understand how it was courage' given that hesitations, non- 'believed in' at the time she is 'too old' grammatical sentences, etc. FIGURE 1.3 Old Wu - Hypotheses about Three Types of Knowledge Gained or to be Gained through the Raw Data this should increase my (and your) caution as I 'infer and interpret' the interview data I collect. While strategically my prior contextual knowledge and misinformation 'governs' the process of 'making inferences from the data', prior knowledge, expertise, ignorance and prejudice luckily need not be all-powerful. As I struggle to make sense of (make sensible inferences from) the data, I hope to find that my previous body of knowledge and misinformation gets disturbed and enriched: i.e. I 'learn' from the process. 12 CONCEPTS AND APPROACHES TO DEPTH INTERVIEWING INCREASING INTEREST OF RESEARCHER IN END OUTPUT 'Hard facts' of interview interaction 'A' Tape & transcript text plus non-verbal communication and experience evoked in 'session notes' Objective knowledge about discourse Discourse Soft inference from 'hard facts' of 'A' Objective knowledge about the referent(s) 'context' Objective knowledge re subjectivity/ies CONTEXTUAL KNOWLEDGE AND MISINFORMATION Objective knowledge about Subjectivity(ies) Objective knowledge about referent(s) in context FIGURE 1.4 Reading Interview Interaction 'Facts' for Discourse, Objectivity, Subjectivity in the Light of Contextual Knowledge and Misinformation In Figure 1.4, the first set of inferences are from the 'hard facts of that particular interview' to a model of the interviewee's discourse. There is then a second set of inferences which can either take a top-line movement from inferences, knowledge and misinformation about objectivity towards a model of subjectivity, or take a middle-line movement from inferences, knowledge and misinformation about subjectivity towards a model of objective context and the referents. In either case, the models developed then have to be rendered coherent with each other in the light of the mutual inspection of models of discourse, objectivity and subjectivity, and of the diagram's bottom line: the resources and handicaps with which I start: my existing prior knowledge and assumptions, prejudice and ignorance. The process is iterative. Inference = Assumptions + Evidence-handling Argument Against Counter-arguments Assuming (for the moment) that our record of the 'hard and only verbal facts of an interview interaction' is correct, I then put forward hypotheses (or knowledge-claims) on the basis of those facts about discourse and/or objective referents and/or subjectivity. These are put forward on the basis of inference-drawing about the significance of the interview interaction data. Inferences require assumptions and an argument. The arguments take the form of bringing evidence for and against various alternative accounts that might be considered, each of which takes 'the facts' into account in a different way to different inferred conclusions (knowledge-claims) about discourse, objective referents, or subjectivity. Such quasi-judicial EVIDENCE TO SUPPORT INFERENCES TO MODELS 13 arguing/or certain inferences and against rival inferences depends to a considerable extent on making and questioning assumptions. These assumptions may be based on general cultural common-sense or on specialized knowledge and expertise, on information prejudice and misinformation, or usually on a mixture of all these. One assumption that I make is that of intended truthfulness. One assumption that I make in coming to the provisional inferences collected on p. 6 about Old Wu is that she is a real person recounting her own experience. If I were to have further information, identifying the speaker of the transcript words as an accomplished actress with a quite different personal history, then many of the inferences made in the diagram above would be undermined. One interviewer I know has never been able to decide whether a particular interviewee of his was involved with the IRA in the 1970s in the frightening way he described or whether he wasn't. As the transcript was re-examined, the researcher became increasingly unhappy about making the assumption of intended truthfulness and, in the end, decided that the interview could not be used. Even if I make the assumption that the interviewee is trying to be as truthful as they can, I may have to consider what errors they may be making. For example, people may make mistakes about dates: perhaps Old Wu's husband was only pardoned in 1982? Myths We Live By (Samuel and Thompson, 1990) has useful discussions and examples of how our 'accounting for the past' may not be as true to the facts as we like to think, since such accounting is subject to a myth-making propensity both at the individual and at the collective level (see also papers in Perks and Thomson, 1998). Similarly, I may make assumptions which question the significance of the assertions in the interview. For example, in the text on page 7, Old Wu explains the madness of her son by his hearing of the death of his father in prison; it may be that this madness was also (or instead) the result of other factors of which she was unaware. It might also be the case that his 'madness' might be better described in other words: 'madness' is a cultural category, a historical 'label', and other terms might be more appropriate for the Chinese concept or for her son's reality, or both. I might assume that she is knowledgeable about the circumstances of her husband's life but less knowledgeable about the nature of her son's disturbance. I might make a further general assumption. I might make the assumption that people attempt to minimize information about their history that puts them in a worse light than they are comfortable with. Making that assumption, I might entertain the speculative hypothesis that Old Wu might be aware that her son became disturbed partly as a result of her remarrying (in itself, or how it turned out), but might not wish to acknowledge this to herself or to the interviewer. This 'further general assumption'changes the inferences that I might make from her data and provides me with further research questions: Was the son even more against the remarriage than he was against the divorce? If she loves her children more than she loves herself, as she now presents herself as doing, why did she divorce, and remarry? What other children does she have and why are their responses to the situation not presented? These questions arise from changing the assumptions that I make to generate inferences from the datum of the transcript. A further assumption may be in order, based on knowledge about Chinese society. Classically, sons were valued more than daughters. I might speculate about 14 CONCEPTS AND APPROACHES TO DEPTH INTERVIEWING inferring that, since she refers to one son, the rest of the category 'children' are made up of one or more daughters. I have argued, using the case of Old Wu as an example, that it is convenient to think of the interview interaction as a single 'event' that occurs at a particular moment in social history, that I attempt to use 'interview material' as evidence for making inferences from 'the facts of the interview' to possible truths about extra-interview realities (whether they be about the person being interviewed, the society and culture they inhabit, the sequence of events they have lived through, or any other subjective and non-subjective realities). I have argued that, in such arguments about extra-interview realities, I use not just evidence to question or argue cases but also assumptions, contextual expertise and misinformation with which I question or buttress those arguments. I shall return to these matters in some detail, particularly in Part IV dealing with the analysis of interview materials. FINAL PROPOSED RE/PRESENTATION PRODUCT/ PURPOSE DETERMINES APPROPRIATE DESIGN OF DATA-GENERATION/PROCESSING 'Fitness for purpose' is the criterion by which I design and choose instruments and procedures for my social research. How can I use this idea in respect of the instruments of semi-structured depth-interviewing and of the procedures for collecting and analysing such interview data? I need to be clear about the intended use of such analyses, about what 'product of interview analysis' I wish to provide for myself and others. Once I know what the shape of the desired product should be, then I can be more specific about what needs to be done in order to produce such a product. This involves thinking very precisely about the type of text that I wish to produce, or that our sponsors or market will wish, or allow, us to produce. However, to attempt to define all the possible purposes/products that might be developed on the basis of semi-structured depth-interviewing practices would be impossible. In this book, I shall take up a particular position about social research products using depth interview material: other positions are perfectly possible and legitimate. This position will be argued for below. Very broadly, I shall argue, following the work of Kuhn (1970), that all social research work and production is largely determined by the norms and exemplars of professional practice characteristic of the research community to which the researcher belongs or wishes to belong. Consequently, to know what your research community wishes to see, you need to study examples of 'very good, very recognized practice' within your discipline. Your training has or will provide you with an experience of the intensive study of 'the classics' (different in each discipline, changing with the society in question), each of which serves to convey to you a model of what is recognized as good practice within the discipline in the society or university at the time. This is known as the 'paradigm' or as an/the 'exemplar'. A collection of such exemplary texts is sometimes called the 'canon' - as in the 'canonical works of English Literature'. Obviously, learning a paradigm is more than just reading texts: it is 'working the tools' and 'interpreting the product' (making inferences from the data) produced by that proficient working of the tools. EVIDENCE TO SUPPORT INFERENCES TO MODELS 15 One of the directions for any system of social knowledge hoping to be accepted as (relatively) 'scientific' is that existing 'theory' must always be 'critiqued' and 'better theory' must be argued for. In this theory-focused model, material from interview research is used to develop better theory than that which existed before; what counts as 'better' is of course always a matter of intense argument. In the next subsection, such an orientation to theory-development and improvement is taken on board as a way of thinking about the uses of interview material in social research. There is another approach to that focused upon critique. It is concerned with 'adequacy of representation', and it is not solely focused upon 'representation as theory'. A painting of a person may be criticized as a bad likeness or representation, but such a critique will not criticize the 'theory' of the painting or of the painter. Instead, the painting will be asserted to be a bad description or model of that which it is attempting to describe or model. In the next subsection, such a concern to use interview materials to provide improved representations, models, or descriptions will also be located. This book, therefore, attempts to be useful both to social researchers who believe that the object of research is that of advancing theory and to those who believe it is that of advancing (verbal/ discursive) description. Both groups place great emphasis, consciously or unconsciously, upon the conceptual frameworks we have to use in any assertion of theory and any modelling of reality. I therefore now go on to consider conceptual frameworks used in the study of the practice of interview interaction, and to clarify the notion of 'conceptual framework' itself as it relates to different uses of the term 'theory'. The need for such frameworks needs to be first established, though: I shall attempt to do this by first indicating the 'problematic' nature of interview interaction, in a rather more 'applied' way than I have done so far, and then by suggesting that certain conceptual frameworks can help us to handle this problematic nature.