Jtl INTERPRETING QUALITATIVE DATA Methods for Analysing Talk, Text and Interaction David Silverman SAGE Publications London • Thousand Oaks • New Delhi PART FOUR SUMMARY 9 Six Rules of Qualitative Research t f In any text on social research methodology, there is a danger of reducing analytical questions to technical issues to be resolved by cookbook means. I attempt here, with a very broad brush, to raise some of the concealed analytic issues that lurk behind some apparently technical questions like observing 'private' encounters or interpreting interview data. Following Wittgenstein, to whom 1 return later, a touch of'hygiene' may be useful in clearing our minds about the nature of the phenomena that qualitative researchers attempt to study. An interesting case in point is Moerman's (1974) study of the Lue tribe in Thailand. As you may recall from earlier chapters, Mocrman began with the anthropologist's conventional appetite to locate a people in a classifi-catory scheme. To satisfy this appetite, he started to ask tribespeople questions like, "How do you recognise a member of your tribe?' He reports that his respondents quickly became adept at providing a whole list of traits which constituted their tribe and distinguished them from their neighbours. At the same time, Moerman realised that such a list was, in purely logical terms, endless. Perhaps if you wanted to understand this people, it was not particularly useful to elicit an abstract account of their characteristics. So Mocrman stopped asking, 'Who are the Lue?' Clearly, such ethnic identification devices were not used all the lime by these people any more than we use them lo refer to ourselves in a Western culture. Instead, Moerman started to examine what went on in everyday situations. Looked at this way, the issue is no longer who the Lue essentially are but when, among people living in these Thai villages, ethnic identification labels are invoked and (he consequences of invoking them. Curiously enough, Moerman concluded that, when you looked at the matter this way, the apparent differences between the Luc and ourselves were This chapter is a revised version of a paper entitled "Six Rules of Qualitative Research; A Post-Rom an tic Argument' that originally appeared in Symbolic Interaction, 12, 2, Fall 1939, 215-230. Six Rules of Oualtidtive Research 197 *yi- considerably reduced. Only an ethnocentric Westerner might have assumed otherwise, behaving like a tourist craving for out-of-the-way sights. Moerman draws our attention to the nature of representation: its forms and, perhaps, its politics. This means that qualitative research can no longer concern itself with discovering truths which are unmediatcd by the situated use of forms of representation. Yet British and American fieldwork still tends to respond, almost instinctively, to two older impulses (Silverman: 1989b). The Enlightenment urge lo categorise and count is found in attempts to locate 'tribes' and cultures in classilicatory schemes. Conversely, the desire to understand raw 'experience' (usually via in-depth interviews) harks back to the romantic movement of the nineteenth century. In this chapter, I summarise the main argument of this book in the context of what remains of Romanticism in qualitative sociology. Admittedly, the crasser forms of this perspective are restricted to student essays and to some of the speeches of the British ex-Prime Minister Margaret Thatcher ('there is no such thing as society', she once commented). Nevertheless, professional social science often still responds lo the Romantic impulse, particularly in fieldworkers' commitment to the sanctity of what respondents say in open-ended interviews. As we saw in Chapter 5, we are thus sometimes left with the unappetising choice between treating accounts as privileged data or as 'perspective!' and subject to check via the method of 'triangulation' with other observations. To talk about 'rules' invites charges of simplification, over-generalisation and so on. While much has had to be crammed into a small space, I hope a common thread will emerge which will tie together the preceding chapters. Throughout, 1 return to the situated character of accounts and other practices and to the dangers of seeking to identify phenomena apart from these practices and the forms of representation which they embody. Rule I: Don't Mistake a Critique for a Reasoned Alternative One of the bad things that happens to some students who take courses in social theory is that they end up being convinced that a whole series of theorists are little more than congenital idiots. Durkheim is a good example of (lie kind of 'straw man' figure that emerged in some people's imagination. How could anybody seriously assume, for instance, that such an individual act as suicide is a consequence of social structure? Surely, such students feel, no account of suicjflc is adequate when it depends on the 'distortions' of official statistics and fails to refer to the motives of the actor. Disconcertingly, however, there are curious kinds of similarities between Durkheim's account of suicide and research by Atkinson (1978) which draws on an apparently opposed theoretical perspective. Like 100 ImerptcUng Qualitative Dato Dnrkheim. Atkinson was not interested in psychological accounts Of suicide which involve reference lo (he meaning of (lie act lo flic aetWI. Modi Durkheim find Alkinson are concerned With the social organisation m whirl) suicide is embedded although Atkinson's cdmomclhodnlogical ptrspcciive locales (hal organis.ilion in Ihc practical reasoning of coronets radier (han in forms of social integration. Curiously, also, neither sociologist would question suicide statistics, although, once again, their reasons would differ. In Durkheim's case, rales of suicide provide him with (he social facts (ha( need explanation and official statistics are the only record of such rates. For Atkinson, lo question such statistics would imply that you had a belter way of measuring suicide. This would only make an irony of real social practices (defining the nature of unexpected deaths, collating local and national statistics, recommending policies to reduce rales «>| suicide, etc.). Such practices should be topics for so* i< »logical investigation. So, although Alkinson considerably redefines Durkheim's problematic, he is, in some ways, quite close (t) his position, certainly in a common opposition (o the sludcnl critique (sec Silverman: 1985. 32-33). Il is also useful lo recognise (he limited nature of many of the claims of (I suppose we have to say) the 'Founding Fathers' of the discipline. For instance, again diking Ihc despised Durkheim, it is important to note lhal Durkheim's polemical characterisations of 'society* have primarily a methodological status. We should read him as telling us how it is useful to view society, not what society essentially is, Thus, just as psychologists would generally resist turning 'psychological' phenomena into puicl\ neurological processes, so sociologists would usually not waul lo reduce social phenomena to Ihc psychological dispositions of the people concerned (Durklieim: 1974, 24-25). In both cases, Ihc problem is an uncritical reductionist form of thinking which fruitlessly searches for some more 'basic' level of analysis. Durkheim's soluiion, wiih which I fully concur, is to slay al one level of analysis and |o sec what you can sny about data al (hat level, withoul seeking to resolve philosophical. or occasionally participants', questions about the 'essential' character of 'reality'. This may seem an obvious point but, judging by what is written in undergraduate examination answers, il docs not usually lodge in students' consciousness. A further feature of such unswers, at least in Great Britain, is sludcms' general horror of what ihcy call 'positivism'. Now. of course. dm lei m may sometimes icier to certain practices, such as * rude quantified ■variable analysis', which still go on and, in certain terms, should be erilicised (Blumer: 1950, Cicourel: 1964). Mul usually sludcids (and some qualitative sociologists') use 'positivism' as a 'catch-all' term which seems lo encompass anything dies don't l^f in .»m ial science I lie problem lhal then arises is (hal il does nol seem to have a referent, since I caunoi think of any corilcmporai v sociologist »>i indeed philosopher of science who adopts (he label 'posilivisC (although, ironically, 1 recently did - Silverman: 1989a). 'Positivism', (hen, now serves as a lerm of abuse and perhaps conceals Six Rules of Qualitative Research 199 that i(s eiitics have no cohcieni alternative. The status of 'posihvism' as a rhetorical device is underlined when beginning Rraduate students hud that lhey lack the resources to translate their critique into a reasoned research proposal. Rule 2: Avoid Treating the Actor's Point of View as an Explanation How could anybody have thought this was die case in social science'* How could anybody (hink (hal wha( we ought to do is lo go out inlo (he field to rcporl people's exciting, gruesome or intimate experiences? Yet. judging by the prevalence of what I will call 'naive' interview studies m qualitative research, Ibis indeed seems to be (he case. Naive interviewers believe thai the supposed limits of structural sociology arc overcome by an open-ended interview schedule and a desire lo catch 'authentic' experience. They fail (o recognise what ihey have in common with media interviewers (whose perennial question is "How do you/does it feel?') and wilh tourists (who, in their search for the 'authentic' or 'different', invariably end up with more of ihc same). They also lofally fail to recognise the problematic analytic status of interview dala which are never simply raw but are both situated and textual (Misltler: 1986). Such analytic issues, moreover, arc not even touched upon in the elegant methodological 'remedies' of survey research. If we reduce micro-sociology to the naive interview, we lose much of the thrust of ihc tradition from which it emerged. As I noled in Chapter 3. you only have lo look al interactional work Írom lbe Chicago School in the 1930s and 1940s to see (he presence of a much more vital approach-Using their eyes as well as listening lo whal people were saying, diese sociologists invariably located 'consciousness1 in specific patterns of social organisation. As we saw, Whyle (1949) showed how the behaviour ol barmen and waitresses was a response to Ihc imperatives of status and the organisation of work roudnes. The experiences of such staff needed to be contc.xtcd by knowledge of such features and hy precise observation of the territorial organisation of restaurants. This issue of (he situated nature of people's accounts directly arose in our study of a paediatric cardiology unil (Silverman: 1987). As just noted in Chapter 8, when we interviewed parents alter (heir child's tirst clinic visit, most said that they had a problem taking anylhing in. They reported that one ol Iheir major problems in concenlraling properly was i uiscd by the crowded room in which (he ci/isultation took place - as i( was a leaching hospital, several oilier doctors as well as nurses and researchers were present. Although we could enipalhise with the parents' response, we thought it worthwhile lo go back to our tapes of the encounters (hey were discussing. As I reported in Chapter 8, it turned out lhal the number of questions 24 7204 78 MO **&P*tmg Qualitative Data pii reu is asked was directly related (o the number of stuff present (not inversely related as (heir interview answers would have suggested^ As is often the case after such a counter-intuitive finding, we found quite a simple explanation. Perhaps when the senior doctor broke off (he consufl.itioii to :isk questions ol the |unioi doi h>i. pi- •..hi . quite uiiinien lioiially, ilus created a space for parents to think ahoiil what they had been (old mi fur iind lo formulaic their questions withoul being 'on stage' in direct eye contact with the doctor. Iliis explanation was supported in another unit where parents also asked many questions after they had had some time on their own while (he doctor studied clinical data (Silverman; 1987, 91-94). This took unpack to our interview material with the parents. We were not prepared to treat what they had told u$ ironically, i.e. as self-evidently mistaken in the light of (he objective data. Such simple-minded 'triangu-lation' of data fails to do justicí to the embedded, situated nature of accounts. Instead, we came to see parents' accounts as 'moral tales' (BAfUCh: 1982, Voysey: 1°75). Our respondents struggled to present their .Kiiniľ. m (lie conti-\i <>f nutí.il wiMoris of rôpomibír parenthood in -i situation where the dice were loaded against then) (because ol the risks to lil<- .nul (he high-technology means of diagnosis and treatment). Parents' reference lo lbe problems of the crowded consu/lalion room were now treated not as an explanation of (heir behaviour at the time bul as ■ situated appeal to the rationality and moral appropriateness of that bebivfoiir Similarly, in a study of lifty British general practice consul-lations, Webb and Slimson (lu7n) noted how the subsequent accounts of patients look on a dramatic quality in which the researcher was encouraged to empathise with the patient's difficulties in the consultation. A story was told in which a highly rational patient had behaved actively and sensibly. By contrast, doctors were routinely portrayed as acting insensitively or with poor judgment. By telling 'atrocity stories', Webb and Stimson suggest lhal patients were able to give vent to thoughts which had gone unvoiced at the time of the consultation, to redress a real or perceived inequality between doctor and patient and to highlight (he teller's own rationality. Equally, atrocity stories have a dramatic form which captures the hearer's attention - a point which field researchers become aware of when asked to give brief accounts of (heir findings. In a certain sense, once again we see how field rexenrchers have come back, in a full circle, to a Durkheimian position. Like Oiirkheim, Stimson and Webb arc rejecting the assumption (hat lay accounts can do (he work of sociological explanation!. Neither wants to take the actor's point of view as an explanation because this would be to equate common sense with sociology - a recipe for the lazy lickl researcher. Only when such a researcher moves beyond the ga/e of the tourist, bemused with a sense of bizarre cultural practices ('Goodness, you ilo things differently here'), do the interesting analytic questions begin. A p.uallel issue arose m j siu.h In Gilbert Ifld MuUCa) (1963) Of Six RuiOiof Qualitative Reseatch 201 sciential*' accounts of scientific practice. As they point out, one way of hearing what scientists say is as haul data which bears on debates in the philosophy of science about (he character of scientific practice. It is then templing to treat such accounts as 'inside' evidence ('from the horse's mouth', as it were) about whether scientists are actually iiillueitccd by paradigms and community affiliations more than by sober attempts to refute possible explanations. Confusingly, Gilbert and Mulkay's scientists used both quasi-Kuhnian and quasi-ľoppcrian explanations of scientific practice. Understandably, however, they were much keener lo invoke the Popperian ('sober refutation') account of how they worked and the Kuhnian ('community context') account of how certain other scientists worked. Were these accounts to be treated as a direct insight into how scientists do their work or how they experience things in the laboratory? Not at all, at least in any direct sense. Instead, this interview data gave Gilbert and Mulkay access to the vocabularies lhal scientists use. These vocabularies were located in two very different discourses: - a 'contingent' discourse, in which people were very much influenced by political considerations, such as institutional affiliations, ability or inability to get big research coulrncts. etc. - an 'empiricist' discourse, where science was a response i<> dala 'out there' in tbc world. Neilhei discourse conveyed the 'hue' sense of science - there is no more an essential hum of scientific practice than there is a single reality standing behind 'atiocily stories'. Everything is situated in particular contexts. Scientists, loi instance. Gilben and Mulfcaj note, arc modi mora lllcely to use a 'contingent' discourse in a discussion at a bar than in a scientific paper. So the issue ceases to be 'What is science?' and becomes 'How is a particular scientific discourse invoked? When is it invoked? How does it stand in relution to other discourses?' Rule 3: Recognise that the Phenomenon Always Escapes Webb and Stimson, like Gilbert and Mulkay, remind us of the occasioned, situated nature of lay and sociological seeing, saying and doing. In (his sense, the link with Durkbein) is clearly broken. Given patients' and scientists' skilful invocation of discourses in appropriate social contexts, Durkheuu's faith in a stable reulity^sepatatc from somebody's seeing, saying and doing, is misplaced. Clearly, the botanist classifying a plant is engaged tri a less problematic activity than an anthropologist classifying a tribe. In both the studies I have been discussing, the researchers disabused us of our common-sense assumptions about the stable realities of particular collectivities. So patients, conceived as a stable phenomenon, escaped the totefpretmg Qualitative Data "£ää. k ass "s í r*'** h»,ng A paper by Woolgar 1985) ľ h( G"be" and Mulk.y'<.«tak. '"""■Sen«'. notes how pa ,S„« ,1"™'" """J"«» Wi,h 'a"ifid" ■her own activities a5 ins aľcfľf - ľ ? " '""V '* ,cluc,a"' "> »«■ Gilbert and Mulkay, CTul"?*,"™Z ?*" phcnome"a- "* Vcl he rep„r,s that/when T.Z .o „ľ '" *' *0dotoW of •*"■ •*«.....M work, each lab. n, v ,~ f ,?' '° ,abo">">'™ '"P""" «M. iV J.- For whatever «JÄ^ÄS »J- p,a« "> '"-riga.e *< »ha, Scien,ifie work really should bľo" he o I ?TT,ýÍ n°'"•"» d^ÄäÍ: r« 55? work bdn8 science. Cíl t0 somc °,her «it« as the home of 'hard' 'elsewhere'. As eaeh new tes, ?j!íľ* was "» P««ived ,o be ground, were ciled ,o find i t£Zla7"'C™S"""C '«'"' AI appeared, l-rgcly rejected because evenI lahZ.' a'"°U8 lWn« '«' » « '."man reasoning and AI, a machmc ™, ,"" '"C ÍatUmx "«*•« wi.hout being 'intelligent' ^vc„"Lhay "**„ *! Sl"""«'»'g intelligence' «elcvisionsdunngcornnterci^willnöthr *" '"«""""v «1*1. off »■nee, it is held, ch.s is a response to I,„ rcC°Smscd « ■" «ample of AI th« In programme centenľ , t° £*"? S .br0adCas' ">al '*»« argues, has generated a seemhľlv „1 8Cn"inc' Al' W°°'8»' 'he phenomenon always ctSJU?* "*■"* P">S'«mme in which p^cnľtľo^ľ;;::.:^»^ ;dealiscd »-*—- research, dissolving into sets „f 'ÍJ|ľľP °"h S "■"" ol V"»»* field Nowhere is this clearer than iľ.he fii , I ^ " '" parlicular «B***-Chap'or 3, pp 56-581 »7r k ofstud.es of the family (sce also given the privacy of the hoMrtS Cis ľ"' rep°r'8 °' hn% life If .here were some authentic site of fam |v f7'!" J" 'de",iscd «-»--PUI «"de, the researcher's microsc. „ľ ?„«'ď ? C°U'd * ÍS0,aled and are applied in varying ways ľ3. ľ"*11' *""■ of family life courts of law, clinics and radio «|SXf ľ"****' """> '" «*•*.■» available for research.invesHgatio! Pr0grammes' are P*"0 «Id readily m.»placed. Their assumptionhľX "nilľT ""'" 'ifC ,0Oks '" <* more like a common-sense way o ZľÄ .Í "" F""*1 """'* ,ook' analytic basis. Finding the family ^""t"8 P1""10"*'»» «I* little -eryday life, we can always locate ľnd , „, "f "" lW' '" °"r ys locate and understand -real' families by using Six Ruleaol Qu3Maove Research 203 the documentary method of interpreta I ion (Garfinkel: 1967) to search beneath a p pen ranees to locale the 'true' reality. In this regard, think of how social workers or lawyers in juvenile or divorce courts 'discover' the essential features of a particular family. Yet, for sociologists, how wc invoke the family, when we invoke the family and where wc invoke the family become central analytic concerns. Because we cannot assume, as lay people must, that families are 'available* for analysis in some kind of unexplicatcd way, 'the family', conceived as a self-evident phenomenon, always escapes. The phenomenon that always escapes is the 'essential' reality pursued in such work. The phenomenon thai can be made to reappear is Ihc practical activity of participants in establishing a phcnonomenon-in-contexl. Rule 4: Avoid Choosing between All Polar Oppositions The philosopher of science Thomas Kuhn (1970) has described sociology as lacking a single, agreed set of concepts. In Kuhn's terms, this makes sociology 'prc-puradigmatic' and in a state of competing paradigms. As I have already implied, the problem is that this has generated a whole series of sociology courses which pose different sociological approaches in terms of either/or questions. Such courses are much appreciated by students. They learn about the paradigmatic oppositions in question, choose A rather than B und report back, parrot fashion, all the advuntuges of A and the drawbacks of B. It is hardly surprising that such courses produce very little evidence that studeníš have ever thought about anything - even their choice of A is likely to be based on their teacher's implicit or explicit preferences. This may, in part, explain why so many undergraduate sociology courses actually provide a learned incapacity to go out and do research. Learning about rival 'armed camps' in no way allows you to confront field data. In the field, material is much more messy than the different camps would suggest. Perhaps there is something to be learned from both sides, or, more constructively, perhaps we start to ask interesting questions when wc reject the polarities that such a course markets. In my discussion of Rule 3, and in Chapters 3 and 4, we saw how the opposition between 'structure' and 'meaning' is not very instructive in a range of settings including families, tribes, laboratories and coroners' courts As I argued in Chapter 5, the same might be said about the analysis of interview data. Does this tell us simply about people's experiences (and thus about 'meaning')? Or arc interview responses instances of collective phenomena, such as moral forms and structures of narration? In the latter case, as Durkhcimian 'collective representations', interview data tell us about structures. So the field researcher necessarily is concerned with both structure and meaning. Here, as elsewhere, attempts to place lleldwork on one side or another of competing paradigms are misplaced. 82 204 Interpreting Qualitative Data Another area in which the 'purity' of particular models may be invoked arises in the decision lo use or to avoid quantitative methods, [n the'uritish sociology of ilio 1970s, the word got about that no good qualitative researcher would want to diny her or his hands with any techniques of quantification. Yet, although many of the criticisms of survey methods in the 1960s were well placed (Cicourel: 1964), so were some of the survey researchers' suspicions about field research. As 1 argued in Chapter 7, we are all familiar with the case-study report that advances its argument on the basis of 'a good example of this is ...' or 'X's comment was typical'. Of course, these are 'good' or 'typical' examples because the researcher has selected them to underline (he argument. Just choosing examples of phenomena stands in the way of both rigorous and lateral thinking. Yet, if you are trying to get some feci about your data as a whole or are actively pursuing deviant eases, it may sometimes be very useful to use certain quantitative measures, however crude they may be. For instance, in the study of a pacdialric cardiology clinic mentioned in Chapter 8,1 observed thai consultations with parents of Down's Syndrome children seemed very different in character to other consultations with parents of children who also had suspected congenital heart disease. To pursue my hunch, I examined closely (he form of the doctor's initial question to (he parents about whether they saw any symptoms in t heir-child. Simple counting then revealed very nicely the way in which the usual doctor's question ('A well child?' or 'is s/he well?') was transformed ('How is s/he?') with parents of Down's Syndrome children (Silverman: 1981). This apparently trivial finding proved to be crucial in an analysis of how primarily 'social' rather than 'clinical* categories came to be central to the formulation of Down's Syndrome children with heart disease. This also tied into die doctor's policy of surgical non-intervention. Moreover, not only was I happier because I could account for all my data, instead of using selected examples, but I was able lo do this by counting in terms of the language used by the participants rather (han imposing my own categories on to the data prior to counting. Categories abstracted from the business of daily life usually impose a set of polarities (or continuums) with an unknown relationship to that business. One obvious example of such & priori polarised theorising is in the abstract models of decision-making found in the polarity of rational/ non-rational action. As Anderson, Hughes and Sharrock point out, such models, whether Weberian or social-psychological (e.g. Cyeri and March: 1963) fail to address: '(he essentially, socially organized character of the discovery, recognition, determination and solution of problems' (Anderson el al: 1987, (44). Using materials from audio-tapes of business negotiations, Anderson et al show that (he parties focus on problems and their provision of candidate solutions is embedded in how they play with the sequencing rules of natural language. For instance, a transition point to a next speaker or a next topic Six Rules #tf Qualitative Research 205 may not be accepted and so a parly can avoid a commitment until more is known of Ihe oilier party's game. Equally, requests for Clarification both buy time and give the ball back to the first speaker in a three-part sequence (clarification request/clarification response). In turn, these sequencing rules are enacted in the context of a set of 'business' relevances which, as Anderson et al show, depend on the display of 'competitiveness' coupled with a form of 'urbane affability* which takes for granted the reciprocity of personal and commercial relevances. Anderson el al's analysis reveals 'what adopting a businesslike attitude to the solution of routine problems means as an observable, interactional feature of daily life' (ibid, 155). in doing so, it emphasises our Rules 3 and 4: not only does it reject prior polar oppositions (say, between rational and nonralional elements in negotiation) but it also shows how 'business' disappears as a unitary phenomenon. As Anderson et al note, 'business life' is interwoven with social life: the purely 'rational' cannot be filtered out from the social. Rule 5: Never Appeal to a Single Element as an Explanation A further parallel beteen qualitative and quantitative work is (hat multifactorial explanation is likely to be more satisfactory than explanations which appeal to what I have called a 'single element'. Just because one is doing a case-study, limited to a particular set of interactions, does not mean that one cannot examine how particular sayings and doings are embedded in particular patterns of social organisation. Despite their very different theoretical frameworks, this is the distinctive quality shared by, say, Whytc (1949) and Moerman (1974). A classic case of this is found in Mary Douglas' (1975) work on a central African tribe, the Lelc. Douglas noticed that an anteater, which Western zoologists call a 'pangolin', was very important lo the Lclc's ritual life. For the Lele, the pangolin was both a cult animal and an anomaly. It was perceived to have both animal and human characteristics - for instance, it tended only to have one offspring at a lime, unlike most other animals. It also did not readily fit into the Lele's classification of land and water creatures, spending some of its time on land and some time in the water. Curiously, among animals that were hunted, the pangolin seemed to the I.ele to be unique in not trying to escape but almost offering itself up to its hunter. Fortunately, Douglas resisted what I called earlier Ihe 'tourist' response, moving beyond curiosity to systematic analysis. She noted that many groups who perceive anomalous entities in their environment reject them out of hand. To lake an anomalous entity seriously might cast doubt on the 'natural' status of your group's system of Classification. The classic example of the rejection of anomaly is found in Ihe Old Testament. Douglas points out that the reason why the pig is unclean, according to the Old Testament, is that it is anomalous. It has a cloven 206 interpreting Qualitative Data hoof which, following the classification system, makes it clean but it does not chew the cud - which makes it dirty. So it turns out that the pt£'* particularly unclean precisely because it is anomalous. Similarly, the Old Testament teachings on intermarriage work in relation to anomaly. Although you are not expected to marry somebody of another tribe, to m.'irry tlu- nlhpring of a marriage between a member of your tribe and an outsider is even more frowned upon. In both examples, anomaly is shunned. However, the Lele are an exception: they celebrate the anomalous pangolin. What ihis suggests to Douglas is that there may be no universal propensity to frown upon anomaly. If there is variability from community to community, ihen-this must say something about their social organisation. Sure enough, there is something special about the Lele's social life. Their experience of relations with other tribes has been very successful. They exchange goods with them and have little experience of war. Whal is involved in relating well with other tribes? It means fWCTtlfllll) crossing a frontier or boundary. But what do anomalous entities do? They cut across boundaries. Here is the answer lo the puzzle about why the Lele are different. Douglas is suggesting lhal ihe Lele's response to anomaly derives from experiences grounded in their social organisation. They perceive the pangolin favourably because ii culs across boundaries just as they themselves do. Conversely, the Ancient Israelites regard anomalies unfavourably because their own experience of crossings boundaries was profoundly unfavourable. Indeed. Ihe Old Testament reads as a scries of disastrous exchanges between the Israelites and other tribes. Douglas' uccount of the relation between responses to anomaly and experiences of boundary-crossing answers Ihe 'why' questions thai I discussed in Chapter 8. It can also be applied elsewhere. Perhaps bad experiences of exchanges with other groups (particularly the stale and Ihe media) explains why British sociologists for many years divided themselves between warring 'armed camps' (so shunning anomaly)? And again, the less apparent doctrinal battles in North American sociology suggest a more peaceful relation with the outside world Douglas' study of the Lele exemplifies ihe need to locate how individual elements are embedded in forms of social organisation. In her case, this is done in an explicitly Durkhcimian manner which sees behaviour as the expression of a 'society' which works as a 'hidden hand' constraining and forming human action. Alternatively. Atkinson's and Anderson et aľs work indicates how one>can follow Rule 5 and avoid single-element explanations by pursuing answers to 'how' questions, without treating social organisation as a purely external force. In the latter case, people cease to be 'cultural dopes' (Garlinkel: 1967) and skilfully reproduce [he moral order. Durkheim's contemporary, Saussure, provides a message appropriate to both these traditions when he reminds u% lhat no meaning ever resides in .i Six Rules of^ialitative Research 207 single term (sec the discussion of Saussure in Chapter 4. pp. 71-73). This is an instruction equally relevant to Douglas' structural anthropology as to Atkinson's (1982) interest in the sequencing of conversation in 'formal' settings. So we can take Saussure's message out of context from the kind of linguistics that Saussure himself was doing and use it as a very general methodological principle in qualitative research. What we are concerned with, as Saussure (1974) showed us. is not individual elements but their relations. As Saussure points oui, these relations may be organised in terms of paradigmatic oppositions (Ancient Israelites, British sociologists, etc.) or in terms of systems of relations which are organised in terms of what precedes and what follows euch item. An example that Saussure himself gives shows the importance of organisation and sequence in social phenomena. The 8.15 train from Zurich to Geneva remains Ihe 8.15 train even if it docs not depart till 8.45-The meaning of the train - its identity - only arises within the oppositions and relationships set out in the railway timetable. Let me illustrate the significance of this with an example drawn from a further case-study. Dingwall and Murray (1983) were concerned with how medical staff responded to patients presenting themselves at a British 'casualty' or emergency hospital unit. They note that Jelfery (1979) suggests that patients are typified by staff as either 'good' and 'interesting' or 'bad' and 'rubbish'. The former might be patients who lested the specialised competences of staff; the latter might be patients with trivial complaints and/or responsible for their own illnesses. Dingwall and Murray argue that Jeffery's polarity inadequately spells OUt the system of relations in which these labels arc embedded. They note, for instance, that children often have trivial complaints for which they themselves are responsible and yet are not usually defined by staff as 'bad' or 'rubbish' patients. Drawing upon McHugh's (1970) treatment of deviance, Dingwall and Murray suggest that casualty staff assign such labels only after assessing whether the patient is 'theoretic' (i.e. perceived to be able to make choices) and the situation is 'conventional' (i.e. that it offers a choice for the patient to make). On this basis, Dingwall and Murray offer a 2 x 2 tabic which reveals the staffs decision-making rules. This is set out in Table 9.1. Table 9.1: Casualty Department Rules Situation Actor COIWiMloiul Non-convcnllunnl 'Theoretic' *B«ľ patientl -inappropriate' patient* 'M»-*»**' ChiW.cn Naive' patient*______ Source: adapted from Dingwall and Murray: 1983 ?m Interpreting Qualitative Data So. in a convenlional situation, ů patient who docs not cooperate with staff is normally defined as 'bad'. Children, however, because they maybe perceived as non-theoretic, will not lind that such behaviour leads to this label. Similarly, in a situation offering no choice (i.e. 'non-conventional'), patients will be labelled as 'inappropriate' ('theoretic') or 'naive' ('non-theoretic'). Indeed, as Dingwall and Murray show, the attribution of deviance to a patient arises only within one of three 'frames' which shape the perceived clinical priority of a presenting patient as set out below: 1 A 'special' frame sorts out patients according to their perceived moral worth (e.g. as"bad', 'inappropriate', 'naive' or simply a child). 2 A 'clinical' frame judges patients simply by whether they constitute what staff perceive lo be an 'interesting' case. 3 A 'bureaucratic' frame operates >" terms of a conception of 'routine' patients, without perceived deviant characteristics or special clinical interest. Such patients get routine treatment. Just as Douglas discovered that the pangolin's anomalous characteristics were the key to unravelling the social organisation of the Lcle, so the anomaly created by children who break rules and yet are not treated as 'bad' patients shows the complexity of decision-making in a hospital setting. In both cases, the importance is revealed of avoiding single-element explanations and of focussing upon the processes through which the relations between elements are articulated. Rule 6: Understand the Cultural Forms through Which 'Truths' Are Accomplished In the Preface to this book, 1 referred to my preference for working with 'naturally-occurring' dala. This seems logical if your interest is in the practices through which phenomena like 'families', 'tribes' or 'laboratory science' are constructed or assembled. Despite this, however, many ethnographers move relatively easily between observational data and data that are an artifact of a research setting, usually an interview. In Chapters 5 and 7, I pointed out the difficulties this can create, especially where 'triangulation' is used to compare findings from different settings and to assemble the context-free 'truth'. However, there are two dangers in pushing this argument very far. First, we can become smug átxmt the status of 'naturally-occurring' data. I have already referred to Hammcrsley and Atkinson's (1983) observation that there are no 'pure' data; all data are mediated by our own reasoning as well as that of participants. So (o assume that 'naturally-occurring' data arc unmediated data is, sclf-evidently, a fiction of the same kind as put about by survey researchers who argue that techniques and controls suffice to produce data which arc not an artifact of the research setting, Six Rule$.qf Qualitative Research 209 'Hie second danger implicit in the purist response is that it can blind us to the really powerful, compelling nature of interview accounts. Consider, for instance, the striking 'atrocity stories' told by mothers of handicapped children and their appeal to listeners to hear them as 'coping splendidly' (see my discussion, in Chapter 5, of Baruch: 1981, Voysey: 1975). There are powerful cultural forms at work in such 'moral tales'. Consequently, the last thing you want to do is to treat them as simple statements of events to be triangulated with other people's accounts or observations. For the fact is that, as societal members, we cart see the 'good sense' of such tales. In many respects, an 'atrocity story' is no less powerful because there is no corroborating evidence. It reveals the 'moral work' involved in displays of 'responsible' parenthood, particularly, as in Baruch's study, where that responsibility had to be demonstrated in the context of potentially unintelligible, high-technology cardiac medicine. Such a perspective derives from two very different but equally neglected sources. Wittgenstein (1968) implies that we should not treat people's utterances as standing for their unmediated inner experiences- This is particularly striking in his discussion of statement« about pain (paras. 244-246, 448-449). Wittgenstein asks: what does it mean when I say I'm in pain? And why ís it that we feel unable to deny this assertion when someone makes it? In our community, it seems, we talk about pain as if it belongs to individuals. So, in understanding the meaning of someone saying 'I'm in pain' we reveal what our community takes for granted about private experience (but not private experience itself). So Wittgenstein makes the point that, in analysing another's activities, we arc always describing what is appropriate to a communal language-game'. Just as I have argued that 'the phenomenon always escapes', so, for Wittgenstein, there is no direct route lo what we might choose to call 'inner experience', A second source for understanding (he public sense of interview accounts is to be found in Mills' (1940) discussion of 'vocabularies of motive'. Mills reminds us that, for sociological purposes, nothing lies 'behind' people's accounts. So when people describe their own or others' motives, the appropriate questions to ask are: when does such talk get done, what motives arc available and what work does 'motive talk' do in the context in which it arises? As Gilbert and Mulkay (1983) were to argue, many years later: 'the goal of the analyst no longer parallels that of the participants, who are concerned to find out what they and others did or thought but becomes that of reflecting upon the patterned character of participants' portrayals of action' (1983, 24). Conceived in this sort of way, interview data become a fascinating topic for analytically sensitive case-study worR. As I argued in Chapter 8, with a little lateral thinking, it is also possible to derive from this approach practical as well as analytic insights. For instance, given the cultural compunction for parents, particularly mothers, to display their 'responsible parenthood', can this be incorporated into medical consultations? In the study of the paediatric cardiology unit (PCU), it would have been 210 Interpreting Qualitative Data tempting to Ibllow other researchers (e.g. Byrne and Long: 1976) and to suggest thai parenls' reported problems derive from doctors' inadequate* communication skills. Our analysis suggested, however, thai the constraints of the setting and of the task at hand (speedy diagnosis and treatment) meant that the first outpatients clinic had no space for some parental concerns and that, in any event, many parenls needed lime to come (o terms with what they were being told. If time was allowed lo pass (when, for inslance, parents had faced the questions of other anxious relatives and had consulted popular medical manuals or the family physician) and the family was invited to revisit the hospital, things might turn out differently.» Such a clinic wasf indeed established at the PCU and the constraints further altered by informing parents in advance that their child would not be examined this lime. An evaluation study indicated that, in the eyes of the participants, this was a successful innovation (Silverman: 1987. 86-103). Yet at no point had we set out to teach doctors communication skills. So the sociological truism 'change the constraints of the setting and people will behave differently' had paid off in ways that we had not foreseen. People responded to the new setting by innovating themselves, parents bringing their children along to sec the playroom and lo discover that the ward was not such a frightening place after all. Conclusion I hope that the discussion of the policy input of one qualitative study has introduced a positive note into these observations. Reviewing my first five rules, I could not fail to notice the uniformly negative form in which they are couched - as if research were all a matter of what you must not do. Of course, I intended throughout to convey a sense of the good things that research can do. I tried lo convey this in the examples of successful case-studies and, above all, in my implicit appeal to lateral thinking. If, as I heard somebody say the other day, the world is divided into two sorts of people - those who make such a statement and those who don't - then I am firmly with the latter group. Perhaps, as Douglas implies, we have something to learn from the Lele. Part of what we might learn is living with uncertainty. Curiously, the critics of such apparently disparate theorists as Garlinkel and Saussure and his heirs have one argument' in common. If everything derives from forms of representation, how can wc find any secure ground from which to speak? Arc we not inevitably led lo ;in infinite regress where ultimate truths are unavailable (see Bury: 1986)? Three responses suggest themselves. First, isn't it a little surprising that such possibilities should be found threatening when the natural sciences, particularly quantum physics, seem to live with them all the time and adapt Six Rules of Qualitative Research 211 accordingly, even ingeniously? Second, instead of throwing up our hands in horror at the context-boundedncss of accounts, why not marvel at the elegant solutions that societal members use to remedy this? For practical actors, the regress becomes no problem at all. Finally, like societal members, why not use practical solutions to practical problems? For instance, as I argued in Chapter 7, even sophisticated qualitative analysis can find practical solutions to the problem of validity (counting where it makes sense to count, using the constant comparative method, and so on). The worst thing that contemporary qualitative research can imply is that, in this post-modern age, anything goes. The trick is to produce intelligent, disciplined work on the very edge of the abyss. *