Coding and Analyzing Ethnographic Records 227
DECONSTRUCTION   CONSTRUCTION CONFIRMATION (open) (axial) (selective)
Mary Ann's Lav/
You can always find what you're not looking for.
LEARNING OBJECTIVES
• To gain an overview of the patient work that data analysis requires. « To gain detailed knowledge of the three steps of data analysis,
■ To learn how to code the materials collected.
« To familiarize yourself with the analysis of documents and images.
• To understand how to test hypotheses and to deal with deviant cases.
• To survey the limits and possibilities of the main softwares for the digital analysis of textual materials.
13.1 Introduction
Ethnography handbooks and books are usually devoid of instructions on how to conduct data analysis. They prefer to dwell on questions concerning the episte-mology and philosophy of research, the relationship between the researcher and actors, or information collection techniques. Among the rare exceptions are the exponents of Grounded Theory (in particular Glaser and Strauss - see 3.5), who have made several original and significant contributions on the matter. I shall draw on these studies in describing this delicate phase of the ethnographer's work.
13.2 The analysis in three steps
It should be remembered that ethnographic data is never analyzed subsequent to its collection. According to many authors (among them Glaser and
SCA SCAASCA (story)
Figure 13.1 Coding
Strauss, 1967; Wiseman, 1970; Spradley, 1980; Strauss and Corbin, 1990; Charmaz, 2000}, the collection and the analysis of data (fieldnotes, documents and visual materials) are not strictly distinct phases. Instead, they are closely intertwined processes which proceed circularly in reciprocal interaction because the data analysis drives closely-focused sampling and information collection.
Strauss and Corbin (1990: 59) propose a data coding procedure divided into three progressive steps: deconstruction {open coding), construction {axial coding) and confirmation (selective coding). This is a sophisticated procedure which enables systematic and efficacious analysis. It generates a spiral reflexive process in which sampling (S), collection (C) and analysis (A) are repeated with a progressively narrower focus in each of the three phases (see Figure 13.1).
The collection of data (fieldnotes, documents and visual materials) therefore takes place in all three of these steps, but in each of them it serves a different purpose: in deconstruction the ethnographer seeks to uncover the conventions regulating the interactions observed; in construction he or she devises a story (theory) about the phenomenon observed; in the confirmation the information collected is used to document precisely and systematically the hypotheses contained in the story (theory).
The American sociologist Kathy Charmaz (2000) has argued that the Grounded Theory must be emancipated from its original subtly positivist premises, released from its objectivistic underpinnings, and purged of certain 'strong' assumptions (causality, operational acts, testing hypothesis, verifiability, confirmation and disconfirmation, prediction, reproducibility, and so on). Of course, Grounded Theory must evolve in a more constructivist direction. Nevertheless, as I have argued throughout this book, there is a risk of throwing the baby out with the bath water: numerous terms can be very well recovered from the positivist lexicon, rather than eliminating them within a constructivist rationale.
' 'i,2„l Beconstructiort (open coding)
Deconstruction is an exploratory phase in which the ethnographer examines the field for concepts (or categories) that can explain an observed phenomenon.
228   Doing Ethnography
As we saw in Chapter 10, the ethnographer proceeds unsystematicaily in this phase, paying attention to all interesting events. He or she adopts a listening stance, ready to change focus whenever a noteworthy action or interaction occurs. Then, after spending some time in the setting, and having taken a certain number of ethnographic notes (mainly observational, but also theoretical,: methodological and emotional), he or she sets about analyzing them.
There are three main strategies with which the ethnographer can begin the analysis of his or her observational notes:
(1) using a check list or conceptual grid to 'interrogate' them;
(2) using a framework; or
(3) classifying.
These three strategies involve different degrees of laboriousness. I shall therefore start with the simplest of them and finish with the most complex.
The check list
The first strategy has been suggested by Atkinson (1992: 455-9) and Silverman1 (1993: 39-41; 2006: 89-92). It consists of the use of a check list (coding sheet or coding form or conceptual grid) with a relatively small number of entries or items (around 20) eliciting information from the observational notes.1 As-Silverman (2006: 89) stresses, this coding form must not be drawn up before the fieldwork begins. Rather, it should be developed only after a sizable number of observations have been made and, if the research is being conducted with, colleagues or co-workers, after extensive discussion. This style of work entails that
the data are inspected for categories and instances. It... disaggregates the text (notes and transcripts) into a series of fragments, which are then regrouped under a series of thematic headings (Atkinson, 1992:455, quoted in Silverman, 2006: 92).
Coding and Analyzing Ethnographic Records 229
ASE STUD' ' pa'
In the early 1980s David Silverman carried, out a study on doctor-patient interactions in a pediatric cardiology unit. During the research the focus was restricted to 'how decisions (or "disposals") were organized and announced. If seemed likely that the doctor's way of announcing decisions was systematically relaled not only to clinical factors (like the child's heart condition) but to social factors (such as what parents would be told at various stages of treatments)' (2006: 89),
After observation of more than fen outpatient clinics, Silverman and his co-worker Robert Milliard developed the following coding sheet to record their observations, which is reproduced only partially (for the full coding sheet, go to www.sagepub.co. uk/gobo). ■.
■ 9 Stage of treatment:
First consultation
Pre-inpatient.......
Post-catheter (test requiring inpatient stay) ■ . ■ : Post-operation 10 Outcome of consultation:
Discharge or referral elsewhere .
Non-inpatienf follow up
Possible eventual catheter or surgery
Catheter
Surgery
No decision
14 Scope of consultation: ...
Prior treatment. -.: History
Extra-cardiac Physical states Child development . Child behavior.
Family's practicalities of treatment or attendance Doctor's practicalities of treatment or attendance Anxieties and emotional problems of family Social situation of family : .:
External treatment agencies ■:.
15 Family's presentation of a referral history
Family
Doctor
230   Doing Ethnography
Coding and Analyzing Ethnographic Records 231
16 Format of doctor's initial elicitation question (e.g. how is she? is she well?)
19 Diagnose statement:
(a)   Use of 'weli' (Dr/FarnJIy/Boih)
'(b). Use of 'normal' (Dr/Famlly/Both)
(c)   Possible diagnoses mentioned (0/1 />!)
20 Decisions:
(a) Passible disposals mentioned (0/1 />1) .
(b) Medical preference stated (Yes/No)
(c) Medical intention stated (Yes/No)
(d) Family assent requested (Yes/No) ■ .
(e) . Family allowed to make decision (Yes/No)
(f) Family wishes volunteered (Yes/No)
■  (9) ■ Family dissent from.doctor's proposed disposal (Yes/Mo) (Source: Silverman, 2006: 90-1, Table. 3.2) ■
This coding form enabled the researchers to identify some behavioral patterns:^
;. For instance, by relating Item 14 on the scope of the consultation to the.
decision-format (Item 20), we were able to see differences between.'
consultations involving Down's children and others. Moreover, it also turned..: : out that there were significant differences between these two groups both . : in the form of the elicitation question (Item 16) and the diagnosis statement; ;•. (Item ::: ■ ■■
Obviously, in making fieldnafes, one is not simply recording data but.: :  also analyzing them. The categories you use will inevitably be theoreti- : :  cally-saiurated - whether or not you realise it! So the coding form , , .': reflected my Interest in Goffman's (1974) concept of 'framing'. This meant that I tried to note down the activities through which the participants managed their identities. For Instance, I noted how long the . doctor and patient spent on social "small talk' and how subsequent. :'. appointments were arranged.
These concerns show how theoretically-defined concepts drive good ethnographic research v., They also demonstrate how one can develop analysis of field data after a research problem has been carefully defined (Silverman, 2006:92).    ■.
This strategy has two advantages and a shortcoming. The advantages are that it simplifies the analysis by reducing the complexity of the information, and that patterns can immediately be identified by cross-referencing information reported in the items. The shortcoming is that the coding form is a highly constrictive instrument from which it is difficult to escape (Atkinson, 1992: 459), so that the ethnographer is cognitively unable to grasp information and activities not immediately insertable into the grid. To remedy (partly) this drawback, it may be useful for the ethnographer occasionally to return to his or her fieldnotes and retrieve information on activities obscured or not considered by the grid (Silverman, 1993: 39). In the case of Silverman and Hilliard, moreover, they tape-recorded all interactions so that they could recode as their hypotheses developed.
The framework
The second strategy consists in using a framework to analyze organizational routines. As in the previous strategy, the framework is constructed only after several hours of observation have been conducted in the field.
In 10.6.2. I discussed research ihal I have been conducting for some years, togeiher wilh co-workers, at a number of Italian medical emergency dispatch ■centers.; After.a period of observation.we.devised an analytical framework to. interpret the operators' activities (see Gobo et at, 2008). It consisted of six entries:
: (a)  the operator's goal;
(b) interaction strategy among the actors to achieve the goal; .
(c) . operator's difficulties; .
(d) : operator's needs;
(e) ; inefficiencies of the computerized information system; and ■ .
(f) theoretical notes or outcomes.
It sometimes happened that a call-taker requested the dispatch of an ambulance when none were available because all the center's vehicles were already dealing with emergencies, In this case, the framework was compiled with observational notes, as follows:.
(a)   Ihe operator's goal:
This situation arises whenever all the vehicles are already dealing with emergencies and the call-taker receives an urgent request for assistance. The situation may. be extremely serious (classified as 'red code') but no ambulances are available,
- T
232   Doing Ethnography
If the call-taker's case is considered most urgent his goal is to get an ambulance J assigned to his case as quickly as possible. To do so. he must 'grab' an ambulance assigned to another case deemed less urgent (for example, one classified as 'green code').
(b)   Interaction strategies among the actors to achieve the goal
Strategy 1 - In a loud voice the call-taker asks a colleague call-taker to re-route-A the ambulance (which he had just dispatched) to the scene of the more serious emergency. This request is made when the call-taker realizes that the colleague;! has sent an ambulance to a location close to the one from which his high-priority « emergency call has been made. ■
Strategy 2 - Again in a loud voice the call-taker asks the operators handling vehicles redeployment (they sit at the ISDN-Radio Area stations In Figure 10.4, 10.6.2) whether there are ambulances soon becoming available but not yet on his monitor.?
There may sometimes be several simultaneous requests for ambulances. All of : them are shouted out loud, and they overlap to Increase the din in the operations;, room (Figure 13.2),
W3fff
Figure 13.2  Call-taker's request to the ISDN-Radio Area operators.
(c)   Operator's difficulties
m The call-taker has difficulty in communicating his request to the colleague ■. has Just dSatched the ambulance. The more physically distant the operator .
Coding and Analyzing Ethnographic Records 233
from each other, the more they must shout their requests, Because of the height of the desk backs and the monitors, the call-takers must rise to their feet to see the faces of their colleagues.
(ii) The call-taker has difficulty in communicating with the ISDN-Radio Area because of its distance away from him. He therefore has to shout to make himself heard.
(d) The call-taker's needs
• Ta rapidly attract the attention of the colleague call-taker.
0   To rapidly attract the attention of the ISDN-Radio operators,
• To get'a rapid reply about the availability of ambulances.
D   To know whether vehicles will soon be available, and which ones.
(e) Inefficiencies of the computerized information system
•■ The interphone system (with headsets) and its SMS (Short Message System) backup for computer-mediated communication among operators is considered inadequate for urgent communications;
8   consequently, the operators prefer face-to-face verbal communication;.
• but this is hampered by the layout of the work stations, . (?)   Theoretical notes or outcomes
The observational notes highlight that the 'old' or analogical technologies . (spoken, non-verbal, face-to-face) are anything but obsolete, Indeed, the operators prefer 'headsers-off communication (with shouted requests, expletives, and so on) to the interphone system with its SMS backup, because they find it more rapid and reliable. Indeed, In a previously-observed incident, this 'headsers-off ■type of communication - not foreseen by the computerized information system -. ;,enabled a nurse to intercept information decisive for successful conclusion of a. rescue operation that she was monitoring via radio.
■ Hence, the information system can only function properly with the so-called 'soft & technologies,' i.e. those with low formal and engineering content. And the design of systems and services based on the user's social knowledge, rather than the designer's, must necessarily envisage the integration of digital technologies with :; analogical ones, instead of excluding the latter.
234  Doing Ethnography
Coding and Analyzing Ethnographic Records 235
The classification
The third strategy is to classify the observational notes applying a similarity dissimilarity criterion. Assigned to each note is a code referring to a concept, and the same code is assigned to notes with similar content. However, the same notes-may refer to several concepts, so that they may receive other codes and labels as well, and be repeatedly classified under different headings (Hammersley and: Atkinson, 1983: 170).2 Consider, for example, the following observational note taken by David Rosenhan (1973: 133) when he was a pseudopatient in a psychP atric hospital. The doctor's behavior can be classified with diverse codes or labels.
Table 13.1
Actors_Verbal exchanges_Body movements    Codes or labels
Pseudopatient Pardon me, Dr. X.
Could you tell me when 1 am eligible for grounds privileges?
Physician Good morning, Dave.       (Moves off without    1. indifference .v::
How are you today?'        waiting for 2. rudeness
a response) 3. non-person
(considering patient as)
The aim of classification is to deconstruct the events and actions observed and segment them among a series of concepts. Depending on the researcher's prefer- "■. ences, there are three ways to select the codewords assigned to ethnographic notes:
(1) invent new terms;
(2) use terms taken from the literature; or
(3) use terms employed by the actors.
The first option is by far the most preferable because it stimulates the researcher's creativity and averts the danger of confusion with already-existing:; concepts or theories (Strauss and Corbin, 1990: 68). However, it should be borne in mind that assigning a code to an action is not a neutral act, for it immediately: attributes a function to the action, and assigns it a frame.
Deconstructing a set of observational notes is to denaturalize them, i.e. dissolve; their spatial and temporal wholeness. By means of classification, the researcher breaks up the natural flow of actions and events (still present in his or her ethno-:; graphic notes) in order to give them new sense.
This deconstructive activity can be performed in various ways, among them the use of counterfactual conditionals as discussed in 9.2.1.
13.2.2 Construction (cxial coding)
In the second (constructive) phase, the researcher reassembles the concepts, developed in the previous phase into a new pattern, the aim being to construct: a first coherent framework. He or she does so according to the model proposed:
by Strauss and Corbin (1990), which comprises the following five components: causal conditions, intervening conditions, context, micro-actions, and consequences. In logical terms, we have the following sequence:
(A) CASUAL CONDITIONS —>
(B) PHENOMENON (understudy) —>
(C) CONTEXT —>
(D) INTERVENING CONDITIONS —>
(E) ACTION/INTERACTIONAL STRATEGIES —*
(F) CONSEQUENCES
Figure 13,3
A first logical framework can thus be constructed for every phenomenon. On the basis of these preliminary results, the ethnographer conducts a second sampling procedure, but this time focused only on the concepts which he or she has decided to explore further.-' In this second phase, the data collection concentrates on each of the five aspects in Figure 13.3. The relative observational notes furnish the basis for constructive analysis which states the relations between the concept (usually of a ceremonial or ritual around which many other micro-actions rotate) and its attributes (see 5.6).
Fruitful use may be made in this phase of interviews and informal conversations conducted in the field (see Chapter 11). These aid the researcher to understand what he or she has observed, and to uncover meanings hidden within the rituals reported by his or her observational notes. This is an analytical procedure which has been used for some time. For example Humphreys (1970), in his already-mentioned study on men's homosexual activity (see 12.3.1), noted down the car license plates of the users of the city park tearooms, traced the names of the owners, contacted them, and then interviewed them. Notwithstanding the enormous ethical issues raised by Humphreys' behavior (which were much debated on publication of his book), his interviews gave him greater insight into the participants' practices. Indeed, as Brekhus, Galliher and Gubrium stress,
the contrast between the focal thin description of tearoom participants and the thick description of the participants' personalities, families, and neighborhoods outside of the tearoom gives the new theme empirical punch (2005:15).
13.2.3 Confirmation (selective ceding)
In the third phase (confirmation), the ethnographer documents/checks the hypotheses formulated during the constrictive phase and anchors them to a theory. This final phase involves two stages of materials analysis (see Figure. 13.1) which integrate the data at a level of generality higher than in the previous two phases. The ethnographer constructs a 'story, a descriptive narrative about the central phenomenon of the study' (Strauss and Corbin, 1990: 116) consisting of ten or so statements (though still taking the form of hypotheses) about the relations between the core concept and its indicators.
236  Doing Ethnography
Coding and Analyzing Ethnographic Records 237
In the mid-1980s, the American health sociologist Juliet Corbln carried out a study 'focused, on.the issue of how. 20 women with chronic illness managed their pregnancies. They were enlisted as subjects at the end of the first or beginning of the second trimester, and followed until six weeks after delivery' (Strauss end Corbln. 1990: 118),    :; J
In the advanced phase of his fleldwork, Corbin integrated.the various phases of her previous analysis to produce a story entitled (or based on the core category), 'management of risk factors associated with a pregnancy/illness':
. The main story seems to be about how women with pregnancies cornpli-r cated by chronic iliness manage the risks they perceive to be associated t with their pregnancies. Each pregnancy/illness can be said to be on-course,: :. .. indicating that the risks are being managed, or off-course, indicating that -.-J they are not. Women are managing the perceived risks in order to have a healthy baby. This desired outcome seems to be the primary force motivating them to do whatever is necessary to minimize the risks. However. :   they are not passive recipients of care but play a very important role in the management process. They not only are responsible for monitoring their ■    illnesses and pregnancies at home, but also make very active decisions .. about the regimes I hey are told to follow. Sn the latter case they consider I he harm that might come to the baby from procedures like amniocentesis or :. from taking high doses of certain medication while pregnant. They carefully
weigh the risks and make judgments about the right thing to do. If they think : . . ■ the doctor is wrong, then they do what they (the women) think should be done (1990: 119-20). .
Having told his or her story, the ethnographer returns to the field and/.: undertakes a final sampling procedure in order to collect further information with which to test his or her statements. The relationships, documented in a satisfactory manner, assume the form of distinct patterns representable in a non-numerical matrix (see Table 13.2,13.7.4).
Strauss and Corbin (1990) stress that the distinction between the deconstructive, : constructive and confirmatory phases is only analytical. In practice, the three phases interweave. Or they may be repeated at the end of the research, if the ethnographer realizes that some concepts are still unclear or some relations are still poorly defined (Strauss and Corbin, 1990: 5S; Silverman, 1993: 46). Analysis strategies obviously differ according to the researcher's degree of knowledge about his or her topic of study. As noted in the section on conceptualization (5.6), the ethnographer may already have a well-defined theory {constructed through previous research or drawn from another author), so that he or she can start his or her research from the second phase (construction) straight, or the third one (confirmation). Besides the virtuality of each phase, one should not
forget that the analysis of observational notes serves to deconstruct the data, conceptualize them, and then reassemble them in new form (Strauss and Corbin, 1990: 57), striking an appropriate balance among creativity, rigor, consistency and, above all, theoretical sensitivity. As Duranti paints out,
an annhfsis is, after all, /; selective process of representation of a given phenomenon xcith the aim of highlighting some of its properties. An analysis that tried to reproduce a perfect copy of its object would not be an analysis, it would give it back to us the way it was. Analysis implies transformation, for some purpose (1997a: 114).
13.3 Using documents
As already said in 1.2.4, a researcher engaged in ethnographic inquiry should examine the documents collected, not in and of themselves, but rather in light of his or her observation plan. In other words, fieldnotes are the core of ethnographic research, and all other materials (documents and images) are ancillary or supplementary to them. For two reasons: first because we are dealing with an ethnographic methodology, not a documentary methodology (see 2.4); second because the documents have been produced, not by the ethnographer with his or her phenomenological and denaturalizing gaze, but by an organization or a group. They therefore have a status different from that of fieldnotes and must be deconstructed beforehand. How is this done?
First, the deconstruction can be performed by not treating the document in the usual or natural manner (i.e. by reading it from beginning to end). The document's unitariness can be 'fragmented': for example by analyzing pages extracted at random, or by photocopying the document and then mixing up the pages, or by reading it in reverse from the end to the beginning.
Second, an important exercise consists in answering the following series of questions proposed by Hammersley and Atkinson (1983:142-3):
(1) 'How are texts written?
(2) How are they read?
(3) Who writes them?
(4) Who reads them?
(5) For what purposes?
(6) On what occasions?
(7) With what outcomes?
(8) What is recorded?
(9) What is omitted?
(10) What is taken for granted?
(11) What does the writer seem to take for granted about reader(s)?
(12) What do readers need to know in order to make sense of them?'
Answering these questions enables the researcher to deconstruct the document and bring out its ideological structures and rhetorical narratives, thereby avoiding simple content or textual analysis.
238   Doing Ethnography
Coding and Analyzing Ethnographic Records 239
Third, the researcher must refrain from the 'autism' intrinsic to narrative analysis. As Gubrum has argued:
much of [narrative] analysis has centered on the internal organization of stories. Less attention has been paid to their production, distribution, and • circulation in society ... I have found that the internal organization of stories, while important to understand in its own right, does not tell us very much about the relation of stories to the worlds in which they circulate (2005: 169).
As Lindsay Prior (2004) has rightly emphasized, not only do we do things with words (to paraphrase the title of Austin's famous book of 1962), we also, do, things with documents. Put otherwise: just as words are not separate from deeds, so documents are not separate from action. They themselves are actions:; in that they intervene in reality to modify and influence it. Seeing documents as, non-human agents (see the ANT theory in 10.5) allows one to escape from purely linguistic analysis and switch to a pragmatic approach.
13.4 Analyzing images
The above recommendation on the use of documents applies to images as well:, their analysis should serve to integrate observational notes, for otherwise, the.: research will expand to such an extent that it becomes never-ending.
An integrative approach has been used by Mehan, Hertweek and Meihls (1986) in their study on decision-making in elementary school students' educational careers (see Case Study in 11,2). Thelrspecific purpose was to grasp the teachers' interpretations of students' behavior. After long periods of observation of the 'referrals' of students in different elementary-school classes (on the basis of psycho-aftitudlnal tests, intelligence tests, medical diagnoses, etc.), Mehdn and colleagues showed the teachers various episodes that they had video recorded in the classroom. An example follows:
Similar behavior was displayed by a referral student and a non-referra! student In another classroom. 'Eddie' has been referred for many reasons, one of which was 'hitting other people'. At the opening of the videotaped reading lesson he began hitting the student seated beside him. During the viewing session the teacher stopped the tape at the point and cited this as an example of the behavior for which Eddie had been referred. Later in the same reading lesson, another student who had not been referred struck , Eddie on the head. The teacher did not stop the tape or comment on the incident (1986: 80). :
This episode revealed to the researchers that: ■ :.
teachers are not reacting to discrete pieces of information: they do not seem to be separating student's behavior (for example hitting a child or saying.
'it's too hard for me') from the circumstances surrounding It. Instead of attending to behavior in isolation, teachers are attending to action in context. The context, in turn, includes the student, the task, the iesson, and the situation In which the action transpires , . , a piece of behavior is not the same when if is conducted by different contexts. 'Johnny hitting Mary during math' is not the same as 'Mary hitting Billy at recess' ... In this way a 'slap in the face' takes on different meanings when embedded in different classroom contexts (1986: 80-1).
Precisely because they preserve the participants' actions in time, video recorded materials lend themselves better to intersubjectivity than do observational notes, and they are particularly useful for team research. Indeed, as Duranti remarks,
such a record can be viewed by different people and subjected to analysis in ways that are quite different from the ways in which a narrative by an observer of the same event would allow (1997: 119).
As for written documents, so for images, the first step in their analysis is deconstruction. Watching a video recording like a film may certainly be useful, but it does not fulfill the functions of a phenomenologically focused gaze. De-naturalizing a video requires fragmentation of its wholeness: partial use is made of the recording, and with timelines different from those of continuous and integral viewing. There are various ways to de-naturalize the social practices preserved in a video recording. An example provided by Christian Heath and Paul Luff, well-known English representatives of workplace studies, is 'transcribing the talk and then mapping out the visual conduct' (1993: 325). Or the researcher can read tine transcripts before listening to the soundtrack; or again, he or she can switch off the video (or cover it) and conduct a first analysis listening only to the sound, and vice versa. It is in fact difficult to analyze a document in the real time of its production, because too much information simultaneously impacts upon the researcher, who must follow both the soundtrack and tire video footage. Instead, slowing down or breaking up the natural rhythm of the document facilitates analysis and aids deconstruction. Moreover, to bring out the latent conventions regulating behaviors, a sequence-by-sequence analysis can be performed by viewing the same episode again and again (or, in the case of recorded interviews, listening and re-listening to the same episode). Only after repeated exposure does the researcher begin to feel a sense of estrangement and uncover tacit knowledge. As Duranti points out:
the introduction of recording machines such as the tape recorder and the video camera (or camcorder) . . . has a number of advantages over the traditional method of participant-observation based on the researcher's skills at listening, seeing, and (most importantly) remembering - whether or not aided by written notes. The ability to stop the flow of discourse or the flow of body movements, go back to a particular spot and replay it allows us to concentrate
240  Doing Ethnography
Coding and Analyzing Ethnographic Records 241
on what is sometimes a very small detail at the time, including a particular sound or a person's small gesture (Duranti, 1997a: 116).
Duranti concludes:
Although at this point a video tape is, albeit limited, the best type of record we can have if we are interested in the integration of speech with body movements and, more generally, with visual communication, we are still trying to learn how to take advantage of such a tool (Duranti, 1997a: 115).
13.5 Documenting hypotheses
Murphy's Law of Research
Enough research will tend to support whatever theory.
Before we examine how hypotheses are checked, there is an epistemological issue to clarify. One often hears or reads the expression 'verifying a hypothesis.' Yet it pertains to an outmoded form of scientific reasoning (the 'inducfivism' devised by the British philosopher Sir Francis Bacon) which maintained that hypotheses should be proved to be true (for details on this topic, go to www.sagepub.co.uk/gobo). Other expressions as well, like 'testing' or 'hypothesis checking' (although less neo-positivist) are not always suited to a constractivist perspective because they assume that it is possible to get at the truth.
Most recently, amid postmodernism and the reflexive turn, the epistemologi- ~ cally most suitable expression seems to be 'documenting a hypothesis.' Whatever the case may be, one should never use the expression 'verifying a hypothesis.'
13,5,1 logical procedure for documenting a hypothesis
Many years ago, Becker and Geer (1960) suggested a logical procedure for
hypothesis testing. It moved through four phases:
(1) compare the hypothesis within different groups of subjects;
(2) ensure that the hypothesis covers all events observed and deemed relevant;
(3) pay maximum attention to the 'deviant cases' or exceptions not adequately explained by the hypothesis; and
(4) conduct a statistical test (if the events have been counted) to calculate the extent of the deviance.
Mehan (1979: 21) and Silverman (2000: ISOff) have supplemented this procedure by suggesting that a hypothesis should first be tested on a restricted corpus of instances. If it does not work, it should be revised in order to encompass the deviant instances. According to the British methodologists Nigel Fielding and Jane Fielding, in qualitative studies
there is no random error variance. All exceptions are eliminated by revising hypotheses until all data fit. The result of this procedure is that [in qualitative
research] statistical tests are actually unnecessary once the negative cases are removed (19S6: 89).
These authors therefore adopt a more radical position by maintaining that all exceptions must be eliminated (see 13.S).
Hypothesis checking is a mix of inductive and deductive processes. It is facilitated if the ethnographer is able to formulate his or her hypotheses in a quasi-experimental form. This is what Rosenhan (1973) did when he designed a counter-research whose outcomes, for the hypothesis to be valid, had to be the reverse of the results of the previous research (see 5.8).
An alternative is to document hypotheses by conducting small-scale tests in the field, formulating beforehand (because, as the well-known proverb says, 'it is easy to be wise after the event') the consequences that should ensue if the hypothesis is valid.
: Some years ago three undergraduates conducted a small-scale ethnographic
;'- exercise at an art gallery. After some hours of observing the behavior of visitors,
'■ they formulated their hypothesis: "when tess expert visitors (i.e. those without a
1 specialized knowledge of art) enter a room, they first head for the paintings which
i the largest number of people are looking at' (theoretical note). We may call this,
j a 'hypothesis cf conformisrn.'
j To document this hypothesis, in a room chosen at random, the students positioned
i themselves in front of one of the pictures (also chosen at random) painted by a
j littie-known artist. They Inspected it carefully and with manifest interest, pretending
j to exchange their impressions and to fake notes.
j After a while a middle-aged couple entered the room. Without hesitating, they
; approached the picture of such apparent.interest to the students and read the
; explanatory caption to find out the painter's name. They seemed disappointed,
j They even fried to peek at the students' notes, but being unable to do so, walked
[ away. Then two further visitors entered the room. They too Immediately went up to
j the picture that the students were studying. They gazed at it for a few seconds,
| and then one of them, with a bemused expression on his face, asked the other:
| 'Do you know him?' The answer was 'No.' They inspected the picture for a few
| more seconds and then strode off.    . .
\ The logical procedure followed by the students can be summarized as follows:.
I 1. Having checked the statuses on the two attributes considered, by. the | . hypothesis (level of art knowledge;, density of visitors looking at a painting),. ■ :
242  Doing Ethnography
2. having checked for the possible presence of intervening variables" explaining the visitors' behavior (for example, courting, the students' physical; attractiveness, etc.),
3. having observed a sufficient number of individuals or instances, and >
4. having also considered possible exceptions,
5. the hypothesis of conformism can be considered adequately documented.
6. A possible counter-test might be to observe 'expert' visitors, noting whether ! their behavior was less conformist, more self-directed.
(To see a complete research project conducted by one of Europe's best-known-' ethnologists, with detailed reconstruction of his hypothesis-testing procedure, go to www.sagepub.co.uk/gobo.)
The biggest risk run by a researcher when checking hypotheses is that he or she.; may reproduce his or her prejudices and pre-assumptions {see 5.5). To partly.;; avert this risk, the researcher may find it useful to deliberately disseminate, a 'cognitive trap' in order to slow down and make cautious her or his reasoning ■ (Gobo, 1993), for example, by reversing the data chronologically, 'mixing' the observational notes, and so on. This has the function of 'blocking' the researcher's mental connections, and forestalling too hash' conclusions which;;; may prevent him or her from expecting the unexpected.
13.6 Building a theory
The difference between 'hypothesis' and 'theory' is anything but clear in the literature. This is certainly not the place to deal with this intricate issue. For our practical purposes we may define a theory as a set of assertions which rest.;; on concepts that have not been given operational definitions (see 5.4), and.; consequently have not been directly observed. From this it follows that theories, :; contrary to hypotheses, are not directly testable.
A theory therefore consists of a series of hypotheses which have been checked ;| and documented. A theory has only a mediated (by the hypotheses) relationship ; with the empirical level. The relationship between theory and hypotheses is ; depicted graphically by the following Figure 13.4.
How can one build a theory? A concrete answer was provided in 13.3, when constructing a story was discussed. We may say that a story (which is a set 7 of hypotheses) is a theory in the making. The main difference is that a story is ; a set of hypotheses still to be tested. But once they have been tested, the story ';! becomes a theory.
Hyp.1
(relationship between two or more variables)
Coding and Analyzing Ethnographic Records 243 THEORY
Hyp.2	Hyp.3	Hyp.n
(relationship	{relationship	(relationship
between two or	between two or	between two or
more	more	more
variables)	variables)	variables)
Figure 13.4  The relationship between theory and hypotheses
13.7 A research example: The organizational construction of blindness
In order to exemplify the procedures described thus far, I shall report the data analysis strategies used in a study on the organizational construction of blindness which I co-ordinated some years ago (Gobo, 1997). The research was conducted in a third-year class (16 pupils) of an Italian elementary school to which a sightless girl (named Yasmeen, aged 8) had been assigned. The initial (and therefore still generic) aim of the research was to study the difficulties of integration encountered by a disabled person in the school system.
13.7.1 DeconsirrucHcri (open coding)
After some weeks of closely observing the sightless girl's interactions with her classmates, and having collected and collated my ethnographic notes, I began to analyze the latter by classifying their content. This analysis yielded a preliminary list of categories (concepts) of rituals. I gave a label to each category: 'arrival at school,' 'getting ready to enter die classroom,' 'going to the washroom,' 'questioning pupils,' 'turn-taking strategies' to speak in class, 'choosing a playmate,' etc. For example, I assigned to this last category an episode - recorded by an observational note - when a boy asked the teaching assistant why Yasmeen (the sightless girl) could not join in a game. The teaching assistant answered as follows: 'because she can't see and so might hurt herself.' In a related theoretical note, I put the hypothesis that sightlessness is not nearly the self-evident phenomenon as one might believe. The fact that the boy had asked such an apparently obvious question suggested that children have mental schemas of sightlessness which differ profoundly from those of adults (and also researchers). At the beginning of the school year, Yasmeen's classmates probably had no precise social representation of a sightless person.4 In other words, they did not know what a sightless person was (apart from the banal notion that he or she is someone who cannot see), and above all what a sightless person can or cannot do. This was evidenced by the fact that they initially asked Yasmeen to join in the same games that they proposed to other children. At this stage of the research, I therefore hypothesized that sightlessness (as regards its behavioral features) was primarily a 'process,' a conceptual product constructed in everyday
244  Doing Ethnography
interactions at school among pupils, teaching staff and non-teaching staff. I recalled a study carried out many years previously by Sudnow (1967) on the social organization of death in hospital, where he documented how death is a social phenomenon before it becomes a biological one. The death of the patient occurred well: before his physical demise, when the hospital staff decided - more or less correctly - that he was dying. A series of organizational practices reserved for dying were then enacted, so that it was very difficult for the patient to 'save himself.'
13.7.2 Construction (axial coding)
Having formulated this hypothesis, based on the ethnographic material and suggested by Sudnow's study, I moved on to tire next phase (construction.) and collected further materials documenting organizational practices and environmental constraints in the school that contributed decisively to constructing a specific notion of disability, and especially the sightless girl's identity. I used the five components of Strauss and Corbin's (1990) model to construct a preliminary framework: causal conditions, intervening conditions, context, micro-actions and consequences. Thus tire initial and generic research objective was made more precise. If the pupils formed their conception of sightlessness mainly through the organizational practices and rituals performed around the sightless child, then systematic observations would have to be made in the school premises indicated by my observations during the deconstruction phase as most significant: the entrance lobby, the staircases, the toilets and the classroom. I therefore assembled a more specific sample for tire new category: the help practices (ethno-methods) enacted for Yasmeen by tire teaching assistant in these five places. For each of them I identified the most recurrent actions, which were then sampled.
For example, during recreation, the children were accompanied to the toilets"; before the next lesson began:
Observational note 23
Yasmeen usually took more time than her friends to finish her snack, and on occasion was urged by the teaching assistant to eat more quickly: 'Come on I You 're always the last' Consequently, Yasmeen got to the washrooms a little after the others. While her classmates were in the toilets, the class teacher waited at the door, leaving the children alone inside. By contrast, the teaching assistant always accompanied Yasmeen into the toilets. . ■ <■
Theoretical note 6
It is likely that this behavior altered the children's concepts of 'private' and 'body' with respect to Yasmeen.   :.
Coding and Analyzing Ethnographic Records 245
Observational note 46
Whereas the children often playfully splashed each other with water while they were in the washrooms, they never played the same joke on Yasmeen. On leaving, the children held hands and filed back into the classroom. Yasmeen. and the teaching assistant were left behind in the toilets and only got back to the classroom when the other children had settled down at their desks. .
For me to state that the children 'never' played jokes on Yasmeen, or that she was 'always' accompanied into the washroom, I must have collected a certain number of substantially similar episodes. In this case, from the sampling frame of 216 recreations (as many as there were in the school year) I extracted a sample of 43 cases, which represented 20% of all episodes of the 'going to the toilet' ritual; Samples were similarly constructed of the most significant rituais (units of analysis) performed in the other places in the school. '■'■/■■[■■
At Yasmeen's elementary school, the cadence of organizational time was set by. the ringing of a bell. At 8:20. the bell rang for a first time to tell the children to get ready to go up to the classrooms. At 8:30 it rang again: the children formed pairs and then filed into the classrooms to begin their lessons, During these organiza- : tionai activities there occurred a series of behaviors which constructed Yasmeen's 'difference': (a) whereas the other children came to school on their own or were accompanied by their parents, Yasmeen was brought by an escort provided by the social services; (b) whereas the children had to arrive at the school before the first bell rang, lateness by Yasmeen was tolerated, and she sometimes arrived even : after the second bell; (c) whereas the children played and joked among them-selves while waiting for the second bell fa ring, Yasmeen was taken aside by the teaching assistant, and they talked together separately from the class and the teachers; (d) whereas the children formed pairs to go up to the classroom, . Yasmeen and the teaching assistant either preceded or followed them. It appears likely from these observational notes that Yasmeen's identity (and in particular what it means to be sightless in cognitive terms) was constructed organizationally,
The children's work tables were arranged facing the teacher's desk. Yasmeen sat in the back row, next to the teaching assistant's desk. The children changed places during the school year so that they: could interact with other classmates.
246   Doing Ethnography
Coding and Analyzing Ethnographic Records 247
But Yasmeen stayed at the same work table throughout the year, because fi ■ next to the teaching assistant's desk. I also noted that the children who showet: '■ greatest friendliness towards Yasmeen were those who sat at the fable close ■ her: physical proximity bfed affection. A further difference was apparent fron ■' differing use of deference rituals (Goffman, 1956a): whereas the children addressed' the class teacher as 'Miss', Yasmeen was allowed to use the teacher's first name
The various subjects taught during the school day may have heightened the differences between Yasmeen and the class. During a geography lesson, for oomplo the class teacher described the morphological features (mountains, plains, rivers) of an Italian region, The pupils stood around the teacher's desk on which the map was displayed. Yasmeen stretched out her hands to touch the map. The teacher told her'; that In so doing she was preventing the other children from following the descrlc ■ ■. and that in any case the map did not have 'enough relief for Yasmeen to be able ■ to understand by touching it. The same thing happened while: the. ted explained how the compass worked; Yasmeen tried to open the instrument." v
As the teacher was explaining a point, Yasmeen would sometimes ra^se herharid" to ask questions or to make remarks, thereby interrupting the lesson and irritating her classmates. On other occasions she raised her hand to ask for the turn, but spoke before being explicitly permitted to do so by the teacher. This event was rather common: the convention that a pupil had to wait for a s'gncl from the teacher before he or she spoke was often breached, because iho children co-ordinated themselves differently; they looked at who had raised Iheir hand first, or if there were still hands raised by pupils wanting to speak, they acled without waiting for the teacher's signal; These cognitive and social skills were obviously not available to sightless Yasmeen. If was grotesque for the teacher to reprimand her by saying: 'Wait your turn! Someone put their hand up before you did.' .
Also the weekly lessons In religious (Catholic) education and English unjustifiably-' constructed a difference between Yasmeen and the class. The religious education teacher (a priest) adopted a protective and permissive attitude towards Yasmeen which he never showed to the other pupils, The English teacher was likewise over-Indulgent with Yasmeen: indeed, a pupil exclaimed one day: 'Why does-Yasmeen always get "very good" and we almost never do?' The teacher's explanation was that Yasmeen had to be helped because she was handicapped.
13,7,3 A framework
Using the suggestions of Strauss and Corbin (1990), I then constructed the::; following framework, with five aspects (see 13.2.2).
1, Causal conditions = The professional model used by the teaching assistant to interact with disabled pupils
The main features of the teaching assistant's professional model were the principles on which it is based. These included: never leave disabled pupils
on their own; always keep them beside you; give them affection; do not let them : feel aione; support them; foster their relations with the other children; guide
them in their movements; do not let them feel too diverse, etc.
As Strauss and Corbin (1990: 100) note, in reality a ritual rarely has only one ■cause: the etiology is usually more complex. For example, the type of training "received by the teaching assistant, his or her religious beliefs and biography may "■'be included among the causal conditions.
i   In my observational notes (given below in 5. Action/interactional strategies), ■ causal conditions are signaled by the conjunction 'whereas,' which frequently occurs in the descriptions.
2. Phenomenon (main concept) = Teacher's supporting practices
I was interested in the following properties of the teaching assistant's support:
(a) the main aim of her supporting practices; ■(b) the rhythm with which they were repeated; and (c) their timing.
The indicator of the attributes 'aim' was the degree of independence that Yasmeen was able to acquire in the course of the school year. Given that this was an excessively general indicator, I concentrated on one of its (sub)indicators: the amount of initiative granted to Yasmeen by the teaching assistant.
The indicator of the attributes 'rhythm' was the frequency of tine supporting action.
The indicator of the attributes 'time' was the duration of the supporting action.
The choice of the indicators had to be carefully thought-out, and it had to concentrate on highly significant aspects of the practice examined, because the indicators would subsequently be used as variables in the non-numerical matrices.
For the degree of independence indicator I selected a four-mode variable comprising none/low/medium/high. For the frequency of support indicator I used a five-mode variable: never/ rarely /sometimes/ often/ always. Finally, for the duration of support indicator I restricted the variable to three modes: no action/brief action/prolonged action.
3. Intervening actions = The classroom teacher's mental model (or conception of sightlessness)
This model may have helped or hindered the micro-actions.
4. Context * Environmental constraints, the behavior of the other teachers, the school staff and Yasmeen's classmates as the school year proceeded, and of the children's parents
5. Action/interactional strategies = (a) Arrival at school, (b) Control of punctuality, (c) Waiting in the entrance lobby, (d) Forming the line
For example, I identified, sampled and systematically observed four micro-actions in the entrance lobby (my observational notes are in brackets):
(a) Arrival at school (whereas the children came to school individually or were brought by a parent, Yasmeen was accompanied by an escort);
248   Doing Ethnography
(b) Punctuality ritual (whereas it was compulsory for the children to arrive at the school before the first bell rang - at S.20 - lateness by Yasmeen was tolerated-indeed, she sometimes arrived after the second bell at 8.30);
(c) Waiting in the entrance lobby (whereas the children played and joked amon^i themselves in the lobby, Yasmeen was met by the teaching assistant, and they talked together - separately from the class and the teachers - while waiting for the second bell to ring); and
(d) Forming the line (whereas the children lined up in pairs before going to the classroom, Yasmeen and the teaching assistant either preceded or followed; the class).
6. Consequences = Concept of handicap acquired by Yasmeen's classmates
As said, Yasmeen's identity - and in particular what sightlessness signifies, m cognitive terms - was organizationally constructed. The causal chain may have: been as follows:
(1) the teaching assistant's professional model guided her supporting practices, which in their turn; (2) were the cause of Yasmeen's lack of independence; and (3) formed her classmates' conception of disability.
The three assertions were very strong hypotheses which had to be tested and documented with equally strong empirical evidence. The axial coding therefore formulated assertions about the concept (usually related to an action) and its properties.
13.7.4 Confirmation (selective coding)
The third phase began with tire construction of a story (Strauss and Corbin, 1990k 119) consisting of a restricted number of hypothetical statements (around ten) which, once tested, constituted the framework of the theory. The aim of the: research, in fact, was to produce a theory, not just descriptions. The story was"
Professional model
Yasmeen's lack of independence
Formation of a particular conception of disability among Yasmeen's classmates
Figure 13.5  Professional model
Coding and Analyzing Ethnographic Records 249
entitled 'supporting practices' (the main category) and its hypothetical assertions were:
the teaching assistant had learned an assistive professional model;
2.  this model tended (unconsciously?) to keep Yasmeen in a state of dependency; -
in fact, towards the end of the school year Yasmeen had still not developed a significant degree of independence:
Yasmeen's independence was obstructed not only by the teaching assistant but also by environmental constraints in the school; : .
the supporting practices which rotated around Yasmeen contributed crucially to forming the concept of visual disability among her classmates, especially as regards what a sightless person can and cannot do;
the same function was performed (albeit to a lesser extent) by the sanctions and rewards distributed by the class teacher, and by the teachers of religion and English; ■
7.  during the school year Yasmeen's classmates changed the way in which they:, related to her;     .: .
towards the end of the school year Yasmeen had not formed any meaningful or close relationships with her classmates;
9.  her only close relationship was with the teaching assistant
Having outlined the story, die next step was to return to the field and sample (for the third time) the actions and events associated with each assertion, thereby collecting further information with which to document their validity. Obviously, the data collected previously was not neglected: given that ethnographic research is a long and laborious process, it is vital that the etlinographer exploit all information to the maximum extent, and especially information that has already been collected. I therefore returned to my 'old' notes with a new objective, that of confirming or disproving the assertions of the story.
Each assertion (which was no more than a statement) had to be supported by an accurate description. In other words, it had to be enriched with details and episodes which gave it solidity in the eyes of the reader (concept of 'thick description'). I thus expanded die story, constructed a complex model, and gave greater sophistication to my theory. In some cases, the relations confirmed were
250   Doing Ethnography
Table 13.2 Patterns of behavior
DURATION FREQUENCY
of the of the
leaching teaching
DEGREE OF INITIATIVE assistant's assistant's
SETTING      MICRO-ACTIONS   granted to Vasmeen     action_action
	Arriving at school	none	—	—
Entrance	Controlling	high	—	—
lobby	punctuality			
	Waiting	none	long	always
	Forming the line	none	long	always
	Turn-taking	high	no action	_
Classroom	Participating in activities	high	no action	—
	at the teacher's desk			
	Question-asking	high	no action	
	Rotating pupils among	none	no action	—
	work tables			
	Writing compositions	high	brief	rarely
	Entering the toilets	low	prolonged	always
Washrooms	Wafer	—	—	—
	splashing			
	Leaving the	low	prolonged	always
toilets
Legend: — = not relevant,
';-:Si:Sl|iiHi|:
outright patterns which could be represented in non-numerical matrices (see::fi§^i Table 13.2), like those proposed by componential analysis (cf. Spradley, 198(1: 130ff) or produced by cross-classification.
The matrix furnishes a succinct representation of the pattern of relations. However, like all typologies, taxonomies and classifications, it is static. The researcher must consequently make it dynamic by describing processes. The information contained in the matrix must therefore be connected with actions so that the reader can perceive the Lndexicality and meaning of the interaction lying behind the succinct representation furnished by a matrix.
13.7.5 Operational proposals
Obviously, it is not necessary for a research study to finish with practical sugges- ] tions besides the theoretical conclusions (which are obligatory). However, it docs i no harm if young researchers propose applications which exploit the richness of | their ethnographic observations. The point is very simple: if research has no j social utility, should the public pay for it? This is obviously not to argue that ail social research should be applied research; anything but. Yet conceiving social research only as pure basic research is deleterious to both research or society.
As regards the case examined here, I began to think of measures that could be designed to encourage the sightless girl's integration into the school. From this ■ point of view, a variety of organizational changes could be proposed. First, individual intervention by the teaching assistant should have been reduced because it had harmful effects: by creating excessive dependency in Yasmecn,
Coding and Analyzing Ethnographic Records 251
it hampered her progressive achievement of independence. Moreover, it did not encourage the development of relationships between Yasmeen and her classmates: rather than fostering solidarity and social integration, the teaching assistant vvas a factor in Yasmeen's social exclusion.
Second, the teaching assistant should have taught the whole class, and not Yasmeen singly. Functional differences can be designed for this purpose while avoiding overlaps between the roles of the two teachers.
Third, as regards the specific problem of blindness, a collective distribution of responsibility for Yasmeen could sensitize her classmates in practical terms. The emphasis on practical aspects is important. A study conducted by Siperstein and Bak (1980) in England, where children at an elementary school were given lessons to heighten their awareness of blindness, showed that purely technical instruction produces improvements in terms of politeness and carefulness towards sightless classmates, but it does not create friendship. The responsibi-lization of Yasmeen's classmates should instead have been developed on practical tasks: for example, many of the activities described in previous sections could have been entrusted to the children themselves, thus stimulating greater solidarity in the class.
Fourth, Yasmeen should have been given greater freedom of action so that she could develop intellective and psychomotorial autonomy. She could have gone to the toilets by herself, finding her way by touching the corridor walls, orienting herself by means of the voices and noises in the school, and by using the banisters on the stairs. Many other ergonomic devices could have been installed for her. It was also important for Yasmeen to learn certain falling techniques (forwards, sideways, etc.) in order to increase her autonomy.
13.8 Dealing with deviant cases (instances)
Mater's law
if the facts do not conform to the theory, they must be disposed of,
Ethnographers often come across observations or instances that conflict with their hypotheses or theory (for a case study by the ethnologist de Martino, go to www.sagepub.co.uk/gobo).
Mehan (1979) and Silverman (2000: 181) argue that a theory must be reformulated so that all deviant cases are removed. This task can be made easier if you collect video or tape-recorded data; so you can return to them to check on apparently deviant cases which is very difficult if you only have fieldnotes.
This is certainly to be recommended, and researchers should make every effort in this direction. However, it is not always possible to resolve the discrepancy between hypothesis and instances. As Strauss and Corbin (1990: 14.0) point out, there may be numerous reasons for the gap: the presence of observations that differ markedly from the majority of cases (outliers); observations which represent transition states (both entry and exit) in the behavior studied; or the differing influences that intervening conditions ma}' exert on the states of an attribute, without this invalidating the relation highlighted. Illuminating in this regard is
252  Doing Ethnography
Coding and Analyzing Ethnographic Records 253
the analog}' proposed by Jerry Jacobs (to justify the fact that his formulation -relative to the phenomenon of suicide - was valid for almost all the 112 notesleft by suicides with the exception of ten):
consider the statement 'Light travels in a straight line,' except when it encounters an opaque object, except in the case of refraction, except in the case of diffraction, and so forth. One does not say of these exceptions that they tend to negate the principle of the rectilinear propagation of light. They simply work to narrow its scope and set its limits ... To the extent that one is able to explain the 'exceptions' in such a way that the explanations are consistent with the evidence, the sum total of these explanations constitutes a more detailed arid inclusive understanding of light {Schwartz and Jacobs, 1979:159).
13.9 Using computers to analyze ethnographic data
The Law of Computer Programming
Any given program, when running, is obsolete.
As I have already mentioned, recent years have seen the spread of software programs which can be of great assistance in the analysis of materials. To draw a rather rough distinction, we may say there are two types of software on the market: qualitative (CAQDAS) and statistical.
13.9.1 Computer-Assisted Qualitative Data Analysis Software Denoted with the acronym CAQDAS (Computer-Assisted Qualitative Data Analysis Software), programs as NVIVO, ETHNOGRAPH, ATLAS, QTJALPKO and so on have been developed mainly for the analysis of texts (interviews and documents in particular).
They were initially electronic 'cut and paste' tools, but with the passage of: time some of them (for example, NVIVO) have extended their capabilities to the; analysis of visual materials as well, starting from the semiotic principle that: images, too, are nothing other than texts.
Learning these programs obviously requires time and resources, but the investment is worthwhile.6 Moreover, there are several excellent overviews. (Seale, 2000; Fielding, 2001; Kelle, 2004) which explain the use of them. I only provide the references for these publications here, given that there is insufficient: space to discuss them thoroughly.
What, bearing in mind that fieldnotes are nothing other than texts, are the. principal advantages of these programs?
13.9.2 Five advantages of CAQDAS
First, they help researchers when sorting their ethnographic notes. It should be remembered that the analysis of ethnographic notes is not done in a day; instead it is a process that lasts for months. It is therefore easy to forget the details of;
work done previously. Manual sorting is always a complex operation: the codes previously used, or their original meanings, are forgotten; and the assignation of the codes to the fieldnotes is not always consistent. This gives rise to problems of reliability (see 14.3) in the classification, and to confusion in building a consistent theory.
With these software programs, instead, the researcher has a constantly present overview of the list of codes that he or she has invented (the master list), and their meanings. Moreover, ethnographic notes sorted under a particular code can be instantly retrieved and, if necessary, changed, renamed or revised. However, these capabilities and the relative ease with which codes can be assigned may be harmful, because they give rise to delusions of omnipotence. Various authors have warned that these programs (if used recklessly) may generate a plethora of codes difficult to handle in the subsequent phases of analysis (Bryman, 1988: 83-7; Silverman, 1993: 47). Hence the well-known Dykstra's Law - 'if de-bugging is the process of removing bugs, then programming must be the process of putting them in' - applies to these programs as well.
Second, starting from the basic codes, more refined CAQDAS can be used to enumerate codes (and therefore produce statistics as defined in 12.5), create co-occurring codes showing relationships among categories, and develop hierarchical category systems. These category systems are constituted by nodes which can be browsed.
Third, this way of working increases the transparency of the fieldnotes classification process and heightens intersubjectivity. This facilitates team research: the researchers can exchange codes and examples of sorting, and agree on the interpretation of certain incidents. All this considerably improves reliability.
Fourth, the more refined CAQDAS aid the process of 'building a theory.' The meaning of this expression must obviously be defined, for the theory is not automatically churned out by the program when it has sorted the ethnographic notes. If only this were so! What the software can do is assist the researcher in the more rigorous checking of hypotheses. For example, NVIVO has nine Boolean operators (AND, OR/XOR, NOT, IF, THEN, EXCEPT and so on) which, once the observational notes have been imported into the software, enable the hypotheses to be checked automatically by scanning the content of the observational notes. The outputs from this scanning are portions of text (in this case of ethnographic notes) assembled by the action of the Boolean operators and immediately visualized.
Finally, hypotheses and theories can be graphically depicted by means of diagrams.
Since the beginning of the 1990s, widespread use has been made of CAQDAS, fueled by an unconscious collective hope that at last software is available for the automatic and standardized treatment of textual data - rather like the statistical packages that have made such a major contribution to the success of quantitative research. Now, however, these hopes have faded: there is nothing miraculous about CAQDAS, and those who use such programs must be diligent and patient. Indeed, several criticisms have been brought against them:
conventional approaches to CAQDAS have tended to naturalise divisions between data management, analysis and representation, and thus there has
254   Doing Ethnography
Coding and Analyzing Ethnographic Records 255
been a tendency to neglect the potential of digital technology to transform all stages of the research process . . . And while some software packages now enable visual data to be included in the dataset, there remains little scope to integrate across data types within an interpretative analysis (Coffey el al. 2006: 18).
13.9,3 Statistical computerized systems for data
collection and analysis :j| Recent technological advances have also led to the development of a number of semi-automated systems for collecting real-time observational data. They are especially widespread in ethnographic psychology, which since its foundation by the American social psychologist Robert R Bales (1916-2004), has always played an important niche role in psychology.
These systems supposedly facilitate observation by improving the reliability!: and accuracy of recording compared to cumbersome paper and pencil methods, ; and by making data calculation and graphing more efficient. The use of computers! to record and analyze data has consequently acquired ever greater importance in clinical work as well as in research.
These softwares are mainly used to analyze video materials. For instance, we can video-record the interaction between a mother and her child, or among | children at play, and then classify what happens in an interval of time í three seconds, for instance). The operation is made easier by the fact that the video I material and the software interact in an integrated manner.
These software programs enable the user to record up to 100 different j responses during a session. They have the capability of recording response ? frequency, duration, intervals (variable duration), time samples, latency, inter- 5 response time (IRT), antecedent-behavior-consequences (ABC), and discrete ; trials. A text feature enables the recording of notes for unique or atypical events. Moreover, a pause feature allows the interruption of observation sessions if nec- j essary, and entry errors made while recording can be edited. The data analysis program gives tire user the option of calculating response frequency (total I number and rate), duration, latency, IRT, percentage of intervals, percentage of í trials, and conditional probabilities. The user can also define subgroups that; contain various combinations of responses (A or B, A and B).
The software also offers the option of calculating statistical measures as central tendencies (mean and median), variability (range and frequency distribu- : tion), and statistical significance (z score transformations). There are more than : 20 in circulation: most notably The Observer, Behavioral Evaluation Strategy and Taxonomy (BEST) and Behavior Observer System (BOS). For a complete overview of their features see Kahng and Iwata (1998) and for a demo of the program 7k Observer, go to http://www.noldus.com/site/doc200401012.
13,10 Concluding remarks
This chapter has shown how to conduct the analysis of ethnographic materials. Such analysis can be done manually, or by means of digital technologies, which are the cause of great optimism. However, we should not forget what happene
in quantitative methods during the 1960s. As the Italian quantitative methodologist Alberto Marradi wrote:
before the age of the computer, data analysis required large investments of time and ability. Researchers therefore saw accuracy in data collection as a guarantee that their efforts would not be wasted on distorted and unreliable data. Now that it is possible to produce hundreds of cross tabulations or correlation matrices in a few hours .. ., the psychological need for this guarantee has gone. [Indeed] on reading research reports, one gains the impression that increasingly sophisticated statistical techniques are applied to increasingly inferior or inadequate data (1980: 4).
It is therefore clear that we have only just begun to experiment with digital technologies, and that we are still far from understanding their potential. But it is certain that, as Duranti stresses,
the invention of new tools that can be used for storing, replaying, manipulating, and reproducing information about human interaction not only offers new solutions to old problems, it also opens up the possibility of new analytical questions (1997a: 115).
KEY POINTS
:<> Data analysis moves through three main steps: deconstruction (open coding), construction (axial coding) and confirmation (selective coding). "In the first .-"step (deconstruction) three alternative strategies can be employed: (a) using a coding sheet; (b) using a framework; or (c) sorting the ethnographic notes.      : .:
• Constructed in the second step is a framework consisting of five elements: (causal conditions '-» phenomenon (under study) -> confer -»intervening conditions h> action/interactional strategies -» consequences) which
:: begin to outline the main features of what will subsequently be the story.
• The third step (construction) ends with the writing of a story which contains the main hypotheses generated by analysis of the materials and forms the core of the final theory,
8 Analysis of the ethnographic notes can be usefully integrated with a: analysis of documents, and of audio or video materials. "An essential part of the analysis is' formulating hypotheses and
documenting them with thin and thick descriptions, 8 Theories consist of sets of hypotheses which have been adequately
documented.
• Not all the incidents observed are appropriately explained by the hypothesis; so some are not covered by it and assume the status of deviant instances.
8 The ethnographer must make the maximum imaginative effort to revise hypotheses so that the majority of deviant cases can be recovered, ■.
8 CAQDAS may be useful tools for sorting textual materials, constructing hierarchical category systems, testing hypotheses, and building theories more rapidly and accurately.
256  Doing Ethnography
Coding and Analyzing Ethnographic Records 257
KEY TERMS
Analysis
(see pp. 234-40}
Axial coding
(see pp. 234-5)
CAQDAS
(see pp. 252-4)
Coding sheet
(see pp. 228-31)
Deviant instances
(see pp. 251-2);
Framework
(see pp. 231-3)
Open coding
(see pp. 227-34)
Selective coding
(see pp. 235-7) :.;
Story (see pp. 235-7)
Theory (see pp. 242-4)
A selective process whereby a phenomenon Is represented so that certain attributes can be emphasized."-The aim of an analysis is not to produce a perfect copy of its topic, instead, it involves a fransformatiory1 of, a cognitive intervention in, the reality observed.
Construction is the phase when the ethnographer begins reassembling in different forms the concepts deve-"-' loped during the previous open coding stage, the aim being to construct a first consistent framework.
Computer-Assisted Qualitative Data Analysis Software' used for textual analysis.
A conceptual grid with a relatively small number of enf- ■ ries or items (around) eliciting information from the observational notes.
incidents and observations at odds with the hypothesis ■■ or the theory which seeks to encompass and explain them.'
A consistent schema made up of a restricted set of elements on which observational notes are collected,
Deconstructioh is an exploratory phase in whichthe ethnographer examines the field for concepts (or cat- . ; cgories) which can explain a phenomenon observed,
The confirmation phase,; in which the hypotheses:.
emerging from the axial coding, are tested and : strongly; documented, and then anchored to a : theoiy.
An account centered on the main research concept or topic. It consists of ten or so assertions (though still taking the form of hypotheses) about the relations between the core concept and Its'indicators;'.'The story is a mock-up of the full theory. •.
: A set of assertions resting on concepts that have not been given operational definitions and consequently-have, not been directly observed.: Theories, unlike hypotheses,, are not directly, testable. Variously mutually consistent and appropriately confirmed hypotheses constitute a theory. ■:
RECOMMENDED READING
For undergraduates: Seale. Clive (2000)
For graduates:. Fielding. Nigel (2001)
For advanced researchers: Emerson, Robert M. (2004)
SELF-EVALUATION TEST
s'Are>youTeady.forthe next :ehapter?..eheckyour:knowledge::by.answering the following open-ended questions:
1. Three strategies are proposed for the analysis of observational notes (coding sheet, framework and sorting). Which of the three do you
.   prefer? Why?
2. What Is the main difference between a theory and a hypothesis?
3. Should an analysis, and consequently a theory, seek to achieve a perfect match with the reality observed?,
4. What is meant by the '.integrative use' of documents and visual . materials?
5. Why is the expression 'documenting a hypothesis' preferable to the more common expression Verifying a hypothesis'?
6. Are CAQDAS a genuine alternative to the traditional manual analysis of ethnographic data? Can you see any risks if they take over from the latter?
Notes
1 This strategy is very widespread in psychology and it lias authoritative precursors in for examples, Bales (1951) and Sears, Rau and Alport (1965), who drew up a short list of behavior categories and then determined their presence or otherwise (their importance and frequency) in incidents concretely observed.
2 Available for some time are software programs which provide very useful assistance with simple or multiple classification. They are discussed in more detail in 12.9.
3 Lofland (1971: 123) has called this process the 'drama Df selection,' because the researcher is forced to restrict his or her focus. This entails that some of the materials must be discarded, perhaps those that the researcher is most attached to because they are rich with interesting details. The researcher must, however, make drastic choices, lest the topics proliferate and a consistent procedure be lost.
4 Goffman noted a similar phenomenon in the psychiatric hospital where he conducted his research: 'children of resident doctors were the only non-patient category I found that did not evince obvious caste distance from patients; why I do not know' (1961: 218, note 2}.
258   Doing Ethnography
5 After a lesson at the town museum, the teacher told the class to write compositions about their visit. Yasmeen introduced hers as follows: 'My composition is perhaps not as good asi: those by my classmates because I could not see the things in the museum very well because' they were enclosed in the showcases.' Evident in Yasmeen's cognitive system is the equation1 'see — touch.'
6 If you are interested in software planning seminars, free interactive introductions to several software programs, introductory workshops, special customized events, follow-up work-; shops, advisory helplines, CAQDAS bibliography, web links to other resources and so on, goto http://caqdas.soc.surrey.ac.uk.
I 5 A-
First Law of Debate
Never argue with a fool - people might not know the difference.
LEARNING OBJECTIVES
• To gain an overview of the debate on the legitimation of qualitative research.
• To understand why it is important to address the problem of evaluating and legitimating your research,
» To identify the main evaluation standards applicable to an ethnographic study.
• To review these standards from a 'practical' perspective.
• To learn how to generalize and defend your generalizations against criticism.
14.1 Introduction
Of all the themes addressed thus far, that of legitimation is probably the most complicated. It has provoked the most heated controversies, and found least agreement among experts on qualitative methodologies. But it is also perhaps the most important of issues, for upon it depends the reputation of ethnography, of qualitative research, and more generally, of the social sciences themselves. I shall seek to show why with a straightforward example.
A group of three of my students conducted an exercise. The first student was from Sicily (Southern Italy), the second from Florence (Central Italy), and the third from